Blog

Cron Expression Syntax: Fields, Operators, and Dialects

Cron expression syntax is built from five space-separated fields — minute, hour, day of month, month, and day of week — that tell a scheduler when to run a job. Each field can hold a value, list, range, step, or *, so 0 9 * * 1-5 runs at 09:00 every weekday.

What Is Cron Expression Syntax?

Cron expression syntax is the compact string format a scheduler uses to decide whether the current clock time matches a schedule. A cron expression describes a set of times, not a countdown. The scheduler compares the current minute, hour, day, month, and weekday against the expression; when every field matches, the command runs.

That difference matters: cron is not an interval timer. Writing */5 in the minute field does not mean “wait five minutes and then run again.” It means “match minute values 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, and 55.” Cron has no memory of the last run and does not measure elapsed time, so the expression is a pattern-matching rule, not a stopwatch. Wikipedia makes the same point plainly: cron is a mere pattern-matcher, not a frequency counter (Wikipedia).

A standard cron line has five time-and-date fields followed by the command to execute. The field order never changes: minute, hour, day of month, month, day of week. System crontabs add a sixth column for the user, and some non-Unix schedulers put seconds first or a year last, but the traditional five-field form is still the baseline. This article focuses on writing the expression correctly. Execution failures, permissions, logging, and notifications are separate operational concerns that come into play only after the schedule itself is right.

Anatomy of `0 9 * * 1-5`: every token, decoded

The expression 0 9 * * 1-5 breaks down from left to right. The first field, 0, means minute zero. The second field, 9, means the ninth hour on a 24-hour clock, or 09:00. The third field, *, means every day of the month. The fourth field, *, means every month. The fifth field, 1-5, means Monday through Friday, assuming Sunday is 0 or 7 in the dialect being used. The job fires at 09:00 on weekdays only. In a system crontab, a user column would sit between the fifth field and the command, which is one reason copy-pasting crontab lines between files can shift the meaning.

The Five Fields: Minute, Hour, Day of Month, Month, Day of Week

The five fields are the core reference for cron expression syntax. Each field accepts a limited set of values, and some also accept three-letter names.

Position	Field	Allowed values	Name aliases	Notes
1	Minute	0–59	None	Smallest unit in standard cron
2	Hour	0–23	None	24-hour clock, no AM/PM
3	Day of month	1–31	None	Some dialects add `?`, `L`, `W`
4	Month	1–12	`JAN`–`DEC`	Names may not be supported everywhere
5	Day of week	0–7 or `SUN`–`SAT`	`SUN`–`SAT`	Sunday numbering varies by dialect

In standard Unix cron, both 0 and 7 mean Sunday, while 1 through 6 mean Monday through Saturday. That convention is not universal. Quartz, AWS EventBridge, Cloudflare, and some other dialects use 1 for Sunday. A copied expression can still parse but run on a different day.

Cron has no concept of a default value. You cannot leave a field blank, and a blank field does not inherit a fallback. The wildcard * is how you say “every value in this field.” Quartz-style schedulers add ? to mean “no specific value,” which helps avoid the day-of-month and day-of-week conflict described later. That difference is often the first sign that a five-field expression is being read by a six-field parser.

Field-order mistakes are costly because they usually produce a valid expression instead of an error. Swap the month and day-of-month fields, and a monthly job can become a daily job on the wrong date. Swap the hour and day-of-month fields, and a 09:00 task can become one that runs on the ninth day of each month. The scheduler does not warn you. It just matches a different pattern.

Value ranges, name aliases, and the values each field rejects

Each field rejects values outside its range, but when that happens varies. Standard cron usually rejects an out-of-range number when the crontab is loaded, while some managed schedulers validate only after deployment. Month names such as JAN and DEC, and weekday names such as MON and SUN, are accepted by many Vixie-derived implementations and are case-insensitive. They are not part of the portable baseline. Vercel’s documentation says Vercel cron jobs do not support alternative expressions like MON, SUN, JAN, or DEC (Vercel, last updated September 16, 2026).

Day-of-week aliases are also where Sunday numbering becomes visible. Cloudflare Workers documentation lists weekdays as 1-7 with case-insensitive three-letter abbreviations such as MON and fri (Cloudflare, last updated September 4, 2026). Quartz and AWS EventBridge also start the week at 1 for Sunday. Unix crontab, Kubernetes CronJob, Jenkins, GitHub Actions, and Spring accept 0 for Sunday and usually also accept 7. A name alias is safer than a bare number when the dialect supports it, but not when the platform rejects aliases entirely.

Field-order mistakes that silently shift your schedule

A field-order mistake rarely causes a syntax error. It causes a different schedule. 0 0 1 * * means midnight on the first day of every month. Reorder it to 0 0 * 1 *, and it means midnight every day in January. The scheduler has no way to know which one you intended.

Two habits prevent most of these errors. First, read the expression aloud with the word “of” between fields, in this order: minute, hour, day of month, month, day of week. Second, check the plain-English translation in a cron expression parser before deploying. A parser that shows the next run times catches a shifted field immediately, because the next-run list no longer matches the cadence you wanted.

The Four Operators: * , – / and How They Combine

Four characters do nearly all the work in standard cron expression syntax: *, ,, -, and /. The asterisk means every allowed value. The comma builds a list. The hyphen defines an inclusive range. The slash defines a step.

Combinations are where the compact syntax becomes powerful. 1-30/5 applies a step of five to the range 1 through 30, producing 1, 6, 11, 16, 21, 26. 0,15,30,45 in the minute field equals */15 in that same field, because both select minutes 0, 15, 30, and 45. 5/15 starts at 5 and steps by 15 through the rest of the field, producing 5, 20, 35, 50. The slash does not create a timer. It selects points from the allowed values.

The highest-frequency standard schedule, * * * * *, runs every minute. According to JasperBernaers.com, that is 1,440 runs per day and 10,080 runs per week. Those numbers are a useful sanity check before deploying anything that runs every minute. A job that sometimes takes longer than sixty seconds can overlap with itself, which is why high-frequency schedules often need a lock or a concurrency policy.

Steps are not intervals: why `*/7` is not “every 7 minutes”

The step operator is the most common source of false expectations. */7 in the minute field selects minutes 0, 7, 14, 21, 28, 35, 42, 49, and 56. The next run after minute 56 is minute 0 of the following hour, only four minutes later. JasperBernaers.com documents this short-gap behavior: any step that does not evenly divide the field range produces an inconsistent final interval. Minutes divide evenly by 2, 3, 4, 5, 6, 10, 12, 15, 20, and 30. A step of 7 does not divide 60 evenly, so every hour has a short gap. If you need a true seven-minute cadence, standard cron cannot express it; you need a scheduler with second-level precision or an application-level loop.

Which Cron Syntax Are You Actually Writing? A Dialect Matrix

“Cron” is not one language. Before writing an expression, identify the scheduler. The same string can be valid, invalid, or valid with a different meaning depending on the platform.

Platform	Fields	Seconds	Sunday	`?`	`L` / `W` / `#`	Time zone default
Unix / Vixie crontab	5	No	0 or 7	No	No	Host local or `TZ=`
Kubernetes CronJob	5	No	0 or 7	No	No	UTC unless `spec.timeZone` set
Quartz	6 or 7	First	1	Required in one day field	Yes	JVM default unless set
Spring `@Scheduled`	6	First	0 or 7	Allowed	No	JVM default unless set
AWS EventBridge	6	No; year last	1	Required	Yes	UTC
Azure NCRONTAB	6	First	0	No	No	Service configuration
Cloudflare Workers	5	No	1–7	No	Yes	UTC
Jenkins	5	No	0 or 7	No	No; has `H`	Controller time zone
GitHub Actions	5	No	0 or 7	No	No	UTC
Vercel	5	No	0–6	No	No	UTC

The practical rule is to treat Linux/Vixie five-field cron as the baseline, then check how the target platform deviates. The biggest deviation is Sunday numbering. In Quartz and AWS EventBridge, Sunday is 1 and Saturday is 7. Copying 0 0 * * 1 from a Linux crontab into Quartz does not fail; it changes the run from Monday to Sunday. On a managed platform, that mistake can run for months before anyone notices.

Is Sunday 0, 1, or 7 in your dialect?

Unix crontab, Kubernetes CronJob, Jenkins, GitHub Actions, and Spring treat Sunday as 0, and most also accept 7. Quartz, AWS EventBridge, and Cloudflare treat Sunday as 1. Azure NCRONTAB uses 0 for Sunday and rejects 7. So the safe answer is not a number at all: if the dialect supports names, use SUN, MON, and so on. If it does not, confirm the numbering from the platform documentation before deploying. Vercel is a notable case where names are not supported, so a numeric day-of-week value must be checked against Vercel’s own 0–6 range (Vercel).

The six-field, seconds-first family: Quartz, Spring, Azure NCRONTAB

Quartz is the clearest example of a different dialect. It is a Java scheduler that uses six or seven fields, with seconds first and an optional year last. Broadcom’s Scheduler documentation, which uses Quartz cron, describes special characters *, ?, #, and #-#, and states that the day of week starts at 1 for Sunday (Broadcom TechDocs, last updated September 17, 2026). Spring @Scheduled uses six fields with seconds first. Azure Functions NCRONTAB also uses six fields with seconds first. Pasting a five-field expression into any of these is not a harmless mistake: the parser either rejects it or shifts every value one position to the right, which can make a daily job run sixty times more often.

What changed recently: OCPS 1.0 (2025) and 2026 vendor documentation

In 2025, the Open Cron Pattern Specification published OCPS 1.0, a formal specification of the Vixie cron dialect that codifies the five-field pattern and rules for previously ambiguous edge cases (Wikipedia). The specification is designed as a backward-compatible superset, with future versions planned to add predefined schedules, second-level precision, and advanced calendar modifiers. Its current role is clarification, not replacement. OCPS is a specification effort, not a runtime that a scheduler automatically follows. Platform behavior still comes from each vendor’s documentation.

That is why vendor documentation dates matter. Vercel’s cron jobs page was last updated September 16, 2026, and Cloudflare’s Cron Triggers page was last updated September 4, 2026. Broadcom’s Quartz-based Scheduler cron syntax page was last updated September 17, 2026. When a platform says Sunday is 1, or that names are unsupported, or that only one day field may be set, the current documentation is the authority for that platform.

Common Cron Schedules You Can Copy — and What They Actually Run

These are standard five-field schedules that work in Vixie-style cron and most managed schedulers that accept five fields. The “runs per cycle” column is the check that catches a misread.

Expression	Meaning	Runs per cycle
`/5 * * *`	Every 5 minutes	12 per hour
`0 0 * * *`	Every day at midnight	1 per day
`0 9 * * 1-5`	Weekdays at 09:00	5 per week
`0 0 1 * *`	First day of every month at midnight	1 per month
`0 4 * * 0`	Every Sunday at 04:00	1 per week
`0 0 1 1 *`	Once a year on January 1	1 per year

A real operational schedule shows why spacing matters more than guessing. RE:NODE describes four ordered tasks for a game server: announce a restart at 04:40, save and back up at 04:50, restart at 05:00, and take a weekly locked backup on Sunday at 04:00. The ten-minute gap between the backup start and the restart is intentional. A panel “delay” is a wait time, not a completion check. A backup of a large world can take minutes to finish, so a restart scheduled sixty seconds after the backup begins can interrupt the archive. Two schedules ten minutes apart are easier to reason about than one schedule with a guessed delay.

Copying any of these schedules still requires a dialect check. A line containing 0 for Sunday is safe on Unix and Kubernetes but means something else in Quartz or AWS. A line containing a name alias is readable on Linux and unsupported on Vercel. The expression is portable only after the dialect is known.

Shorthand macros: @daily, @hourly, @weekly, @reboot — and where they don’t work

Some cron implementations support macros in place of the five fields. @yearly or @annually equals 0 0 1 1 *; @monthly equals 0 0 1 * *; @weekly equals 0 0 * * 0; @daily or @midnight equals 0 0 * * *; @hourly equals 0 * * * *. @reboot runs once when the daemon starts, which usually corresponds to a machine boot, and it has no five-field equivalent (Wikipedia).

Macros are convenient in a Linux crontab and unreliable in hosted schedulers. Kubernetes CronJobs work with five-field Vixie syntax and the @daily-style macros, but they do not support @reboot because a cluster has no single boot moment (JasperBernaers.com). GitHub Actions and AWS EventBridge do not support the macros in the same way. If portability matters, write the five fields out. They are longer by a few characters and they work in more places.

The Day-of-Month OR Day-of-Week Rule (The Most Expensive Misreading)

In standard Vixie cron, if both the day-of-month field and the day-of-week field are restricted — that is, neither contains * — the job runs when either field matches. The rule is OR, not AND. This is the most expensive misreading in cron expression syntax because it silently multiplies the number of runs.

The classic example is 0 0 13 * 5. Many people read it as “midnight on Friday the 13th.” In Vixie cron it means “midnight on every 13th, and also midnight on every Friday.” JasperBernaers.com estimates that produces about 60 runs a year rather than one or two. A monthly report written this way can start arriving weekly. A maintenance job written this way can run four or five times a month.

There is no way to express a true AND in standard five-field cron. The workaround is to leave one of the two day fields as * and test the exact condition inside the command. For example, schedule 0 0 13 * * and have the script check whether the day is Friday before proceeding. Some managed schedulers avoid the problem by disallowing both fields at once. Vercel’s documentation states that you cannot configure both day of the month and day of the week at the same time; when one has a value, the other must be * (Vercel). Quartz and AWS EventBridge require a ? in whichever day field is not being used, so only one day condition is ever active.

The self-check is simple: if both date fields are set to something other than *, stop and confirm that OR semantics are actually intended.

Schedules Five Fields Cannot Express: First Monday, Last Day, Friday the 13th

Some schedules look like cron problems but cannot be expressed in five fields. “The first Monday of the month,” “the last day of the month,” and “Friday the 13th” all require a cross-field condition. The first Monday needs a weekday plus a day-of-month range. The last day needs to know the length of the current month. Friday the 13th needs both day-of-month and day-of-week to be true at the same time, which standard cron cannot express because of the OR rule.

Quartz-style extensions add characters for some of these cases. L means last: in the day-of-month field it is the last day of the month, and after a weekday it is the last of that weekday. W means the nearest weekday to a given day, without crossing the month boundary. # followed by a number selects the nth weekday of the month. ? means “no specific value” and is used in one of the two day fields to keep the other active.

Availability is not universal. Quartz, AWS EventBridge, AWS Systems Manager, and Cloudflare support L, W, and # to varying degrees. Standard Linux cron does not. On Linux, the correct approach is to keep the five-field expression as a coarse filter and perform the exact date check inside the command. That also avoids the OR trap, because the expression no longer tries to encode a cross-field rule.

A common mistake on Linux is 0 9 1-7 * 1. It looks like “the first Monday of the month,” but because both day fields are restricted, Vixie cron reads it as “days 1 through 7, OR every Monday.” The job runs far more often than intended. The safer form is a daily or weekly coarse schedule plus a date test in the shell.

L, W, # and ? on Quartz, AWS and Cloudflare

Cloudflare Workers documentation supports five fields with Quartz-like extensions, including L and W in the day-of-month field and L and # in the weekday field (Cloudflare). Its examples include 0 18 * * 6L for 18:00 UTC on the last Friday of the month, and 59 23 LW * * for 23:59 UTC on the last weekday of the month. In Cloudflare’s weekday numbering, 1 is Sunday and 7 is Saturday.

AWS Systems Manager supports L, W, and # in its cron expressions, and its documentation gives cron(15 10 ? * 6L *) as 10:15 AM on the last Friday of every month and cron(0 2 L * ? *) as 02:00 AM on the last day of every month (AWS Systems Manager). The same documentation states that you cannot specify values in both the day-of-month and day-of-week fields; one of them must be ?. Quartz itself requires a ? in one of the two day fields, so only one day condition is active.

The Linux workaround: escaped `%` and a date test inside the command

On a standard Linux crontab, the workaround for “first Monday” or “last day” is to schedule a superset and filter inside the command. For a first Monday check, schedule the job every Monday at the desired hour and test whether the day of the month is 7 or earlier. For a last-day check, schedule it late in the month and test whether tomorrow is the first day of the next month. The shell expression must be written carefully because % has a special meaning in crontab files.

In a crontab command, an unescaped % turns into a newline, and everything after the first % is sent to the command as standard input. That is why date +%Y-%m-%d inside a crontab can behave incorrectly unless every percent sign is escaped as \%. The cleaner alternative is to put the logic in a script file and call the script from cron. The script runs under a normal shell, where % has no special meaning, and the crontab line stays short. If the command must stay inline, escape each % and test the expression with a parser or a manual run before trusting it.

Does Cron Use UTC or Local Time? Time Zones and DST

Cron schedules wall-clock time in the time zone of the host or service that runs it. The expression itself has no time-zone field. That means two servers with identical cron expressions can run the same job at different real-world moments if their system time zones differ. It also means a hosted scheduler that forces UTC will interpret 0 0 * * * as midnight UTC, which may be 19:00 or 20:00 in the user’s local time.

Several platforms force UTC. Vercel’s cron jobs always use UTC (Vercel). AWS EventBridge cron() rules are always UTC and offer no time-zone option. GitHub Actions scheduled workflows are always UTC. Kubernetes CronJobs default to UTC unless spec.timeZone is set, and that field only exists from Kubernetes 1.27 (JasperBernaers.com). Cloudflare Cron Triggers execute on UTC time (Cloudflare).

Daylight saving time adds two predictable failures. When clocks move forward, an hour disappears. A job scheduled inside that skipped hour does not run at all that day. When clocks move back, an hour repeats. A job scheduled inside that repeated hour runs twice, roughly an hour apart. In the United States, the affected local window is usually 01:00–03:00. A midnight reconciliation job may be safe from the skipped hour, but a job scheduled at 01:30 local can be skipped or duplicated depending on the direction of the clock change.

The skipped hour and the repeated hour

The spring-forward skip is silent. A job at 30 2 * * * simply has no matching wall-clock time on the transition day, so it does not run. No error is raised and no log line is guaranteed. The fall-back repeat is louder in its consequences. A billing run, an email send, or a counter increment that runs twice can produce duplicate charges or duplicate messages. The expression did not change; the clock did.

Mitigations: shift to 03:00, guard with a lock, or pin to UTC

Three practical mitigations reduce DST risk. The first is to move critical jobs outside the 01:00–03:00 local window, with 03:00 or later as a common target. The second is to make the job idempotent and guard it with a lock file so a repeated run does nothing harmful. The third is to schedule in UTC semantics and accept that a “9 AM” job drifts by an hour relative to local office hours twice a year. For machine-facing batch work, UTC is often the more predictable choice. For user-facing notifications, local time may still be worth the complexity, but the schedule should be reviewed against the DST calendar.

crontab -e: Where the Expression Actually Lives

The expression itself is only half of a working cron job. On Linux and Unix-like systems, the expression lives in a crontab file. User crontabs are edited with crontab -e and listed with crontab -l. System crontabs live in /etc/crontab or /etc/cron.d and are managed by system administrators. The behavioral history behind these files is long: cron dates to May 1975, Vixie cron was written in 1987, Vixie cron 4.1 was renamed ISC Cron in January 2004, Red Hat forked it to cronie in 2007, and anacron 2.3 was merged into cronie in 2009 (Wikipedia). Those names matter because “cron” behavior on a modern Linux distribution often comes from a Vixie-derived implementation plus later patches.

Cron’s environment is not the same as an interactive shell. The PATH is usually minimal, often /usr/bin:/bin, and shell profile files are not read. A command that works when typed into a terminal may fail with “command not found” under cron. Use absolute paths for binaries and scripts, or set a PATH= line at the top of the crontab. Output is another common silent failure point. Cron may try to mail stdout and stderr to the crontab owner, and if no mail system is configured, that output disappears. Redirect output explicitly with >> /var/log/job.log 2>&1.

To test the environment cron will use, run the command through env -i /bin/sh -c '...'. That reproduces the stripped environment and catches missing variables, relative paths, and profile-dependent assumptions. Plain cron also does not catch up on runs missed while the machine was off. If the machine was asleep or powered down when the scheduled minute passed, that run never happens. Anacron and systemd timers with Persistent=true exist to cover that gap, but they are not the same as standard cron.

System crontab’s extra user column

System crontabs add a user field before the command. A line in /etc/crontab or /etc/cron.d has six columns: five time-and-date fields, the user to run as, and the command. User crontabs do not have that column. Copying a five-field user crontab line into a system crontab shifts the command into the user field and can make the line fail or run as the wrong user. The reverse mistake is just as common: copying a six-column system line into crontab -e makes cron treat the username as the start of the command.

Before You Deploy: A Sanity-Check Checklist

A cron expression that parses is not necessarily a cron expression that does what you meant. Deployment checks should verify the next run time, the time zone, the overlap behavior, and the missed-run policy.

First, confirm the dialect and the forced time zone. A five-field expression in a six-field scheduler is either rejected or shifted. A UTC platform changes what “9 AM” means. Second, parse the expression and inspect the next several run times rather than only checking that the parser accepts it. Tools such as crontab guru and an online cron parser can show the next fire times, which is where field-order errors and */7 short gaps become visible. Third, check both day fields. If day-of-month and day-of-week are both restricted, confirm that OR behavior is intended. Fourth, decide whether the job can overlap and protect it if it can.

Verify per platform: runnable commands

Different platforms have different verification paths.

Linux crontab: test the command environment with env -i /bin/sh -c '/path/to/job.sh'. This catches missing PATH entries and profile dependencies.
systemd timer: run systemd-analyze calendar 'OnCalendar=...' to print the next elapse time before enabling the timer.
Kubernetes CronJob: run kubectl create job --from=cronjob/<name> manual-run-1 to start the job immediately without waiting for the schedule.
Cloudflare Workers: use wrangler dev and call the local scheduled route at /cdn-cgi/local/scheduled to test the handler. The documentation shows curl "http://localhost:8787/cdn-cgi/local/scheduled" and supports a ?cron= parameter to simulate a pattern (Cloudflare).
Parser check: crontab guru or cron-parser can translate the expression and list upcoming runs. Verify the next run time, not just whether the expression parses.

Stop overlapping runs and spread load: flock, Jenkins H, RandomizedDelaySec

If a job can take longer than its interval, prevent overlapping runs. On Linux, wrap the command in flock -n /tmp/job.lock, which exits immediately if the lock is already held. On Kubernetes, set concurrencyPolicy: Forbid on the CronJob so a new run does not start while the previous one is still active. In application code, an advisory lock or a lock file can do the same job.

High-frequency schedules also create load spikes when many jobs fire at the same minute. Jenkins uses H to substitute a hashed value, so each job runs at an unspecified but invariant time within the hour instead of all jobs starting at minute 0 (Wikipedia). systemd timers use RandomizedDelaySec to jitter starts across a fleet. Both approaches spread load without changing the underlying cadence. For missed runs, remember that cron does not catch up. Anacron and systemd timers with Persistent=true do, but their catch-up timing differs, so the behavior should be documented alongside the schedule.

Two final checklist items belong at the top: confirm whether the platform is local time or UTC, and confirm whether both date fields are restricted. Those two checks prevent a large share of “the job ran at the wrong time” and “the job ran too often” incidents.

Kubernetes CronJobs and GitHub Actions: Same Syntax, Different Rules?

Kubernetes CronJobs and GitHub Actions scheduled workflows both accept five-field Vixie-style cron expressions, but the syntax is where the similarity ends. The differences are operational guarantees, not field definitions.

Kubernetes CronJob syntax matches crontab for the five fields and supports the @daily-style macros, but it does not support @reboot because a cluster has no single boot moment. The default time zone is UTC unless spec.timeZone is set. concurrencyPolicy controls whether a new run can start while the previous one is still active; Forbid prevents overlap. startingDeadlineSeconds controls how late a missed run may still start. If a CronJob misses too many schedules, the controller can give up on it, and the reason is recorded as an event. Those are runtime behaviors layered on top of the expression.

GitHub Actions scheduled workflows are best effort. They can be delayed by 5 to 30 minutes, especially at busy times such as the top of the hour. They run in UTC only. Intervals shorter than five minutes are not honored. A scheduled workflow is also disabled automatically after 60 days without repository activity (JasperBernaers.com). A correct expression therefore does not guarantee an on-time run. For jobs that must happen on time, trigger them from a system that guarantees delivery and use Actions as the runner rather than the scheduler.

Kubernetes CronJob: UTC default, concurrencyPolicy, startingDeadlineSeconds

The Kubernetes CronJob spec accepts the standard five fields. The default schedule interpretation is UTC, and spec.timeZone is only available from Kubernetes 1.27. concurrencyPolicy: Forbid is the built-in way to stop overlapping runs, which is often preferable to a lock file inside the container. startingDeadlineSeconds defines the window after the scheduled time in which a missed run may still start. If the controller misses too many schedules in a row, it stops creating jobs and records an event. The expression can be perfectly correct while the runtime still skips or delays a run.

GitHub Actions: best-effort timing, UTC only, disabled after 60 days

GitHub Actions scheduled workflows use five-field cron syntax, but they are explicitly best effort. Delays of 5 to 30 minutes are normal, and runs can be dropped at peak times. Schedules shorter than five minutes are not honored. The time zone is always UTC. Scheduled workflows are disabled after 60 days without repository activity, so a dormant repository can silently stop running its schedule. These are platform policies, not syntax errors.

Managed Schedulers Beyond Linux: Vercel, AWS, and systemd Timers

When moving from a Linux crontab to a managed scheduler, check three things before copying the expression: whether name aliases are supported, whether both day fields can be set, and whether the time zone is forced. Those three checks catch most migration failures.

Vercel uses five fields but does not support MON, SUN, JAN, or DEC. It does not allow both day-of-month and day-of-week to be configured; one must be *. It always uses UTC. Each invocation includes an x-vercel-cron-schedule header containing the cron expression that triggered it, which helps distinguish schedules when multiple cron jobs share the same path (Vercel).

AWS EventBridge uses six fields with a year field last, requires ? in one of the two day fields, treats Sunday as 1, and always uses UTC. AWS Systems Manager has two different cron modes. Maintenance windows require six fields and support the fastest rate of five minutes. Associations require five fields and must use an interval of at least 30 minutes; AWS states that associations cannot use cron or rate expressions shorter than that. Maintenance windows support L, W, and #, while associations also support # and L in their documentation examples (AWS Systems Manager).

systemd timers are the server-side alternative to cron. They use OnCalendar, support Persistent=true for catch-up, and support RandomizedDelaySec for jitter. systemd-analyze calendar verifies the schedule before the timer is enabled. The trade-off is real: a managed platform may give stronger delivery guarantees but removes some syntax freedom, while a server crontab keeps more syntax freedom but leaves overlap protection and catch-up behavior to the operator.

Vercel Cron Jobs: no name aliases, one day field, always UTC

Vercel supports five fields with minute, hour, day of month, month, and day of week. It does not support name aliases. It does not allow both day-of-month and day-of-week to be restricted. It always uses UTC. The invocation header x-vercel-cron-schedule tells the function which expression triggered it. Those constraints make Vercel simpler in one way — the OR trap cannot occur — and stricter in another: an expression copied from a Linux crontab that uses MON or JAN will not work.

AWS EventBridge and Systems Manager: six fields, `?`, Sunday = 1

AWS EventBridge cron() expressions use six fields with a year at the end, require ? in exactly one of the day fields, and use Sunday as 1. They are always UTC. AWS Systems Manager maintenance windows require six fields and support a minimum rate of five minutes. Systems Manager associations require five fields and support a minimum interval of 30 minutes, and they do not currently support specifying months. Both AWS paths support advanced day syntax to different degrees, so the correct reference is the specific AWS service documentation, not a general cron guide.

systemd timers: OnCalendar, Persistent, RandomizedDelaySec

systemd timers use a different calendar syntax, not cron syntax. OnCalendar defines the schedule, Persistent=true runs a missed job after the machine comes back, and RandomizedDelaySec adds jitter so a fleet does not start every job at the same second. systemd-analyze calendar prints the next elapse time and is the verification step before enabling the timer. One semantic difference matters when migrating: cron treats day-of-month and day-of-week as an OR when both are restricted, while systemd treats them as an AND. A schedule that uses both fields can fire on fewer days as a timer than it did as a crontab line.

Conclusion

Cron expression syntax comes down to five fields and four operators, but the errors that cost real time usually come from outside the grammar. Sunday numbering changes between dialects, the day-of-month and day-of-week fields combine with OR semantics in Vixie cron, and time zones and daylight saving can skip or repeat a run without any syntax mistake. Before writing the first expression, identify the scheduler and the forced time zone. Before deploying, use a cron parser or a platform-specific command to verify the next run time, then protect the job with a lock or concurrency policy if it can overlap. The expression is only one part of a reliable schedule.

FAQ

Is Sunday 0 or 7 in a cron expression?

Unix, Kubernetes, Jenkins, GitHub Actions, and Spring accept 0 for Sunday, and most also accept 7. Quartz, AWS EventBridge, and Cloudflare use 1 for Sunday, where 0 is either invalid or means something different. The safe approach is to confirm the dialect before using a number, or use a name alias such as SUN when the platform supports it. Vercel does not support name aliases, so its numeric range must be checked directly.

Does cron use UTC or the local time zone?

The expression itself carries no time zone. Cron matches the wall-clock time of the host or service running it. Vercel, AWS EventBridge, GitHub Actions, and Kubernetes CronJobs default to UTC. A Linux crontab uses the server’s local time unless a TZ= line overrides it. If the job matters, confirm the time zone before converting local intentions into an expression, or deliberately schedule in UTC and accept the seasonal drift.

Why does my cron job run more often than I expected?

The most common cause is the day-of-month and day-of-week OR rule. When both fields are restricted, Vixie cron runs the job when either matches, so 0 0 13 * 5 fires on every 13th and every Friday. The second common cause is a step that does not divide its range, such as */7 in the minute field, which creates a short gap at the end of each hour. Also check for duplicate crontab entries or repeated deployments.

Can cron run a job every 30 seconds?

Standard five-field cron cannot. Its smallest unit is one minute, and there is no seconds field. Schedulers that support seconds use a six-field format with seconds first, including Quartz, Spring, and Azure NCRONTAB. On a plain Linux server, the usual workaround is a one-minute job that loops internally with sleep, or a systemd timer with OnUnitActiveSec. GitHub Actions does not honor intervals shorter than five minutes.

What does `/5 * * *` mean, and what does `5` mean instead?

*/5 in the minute field selects minutes 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, and 55 — twelve runs per hour. A plain 5 selects only minute 5 of each hour — one run per hour. The difference is that */n steps through the allowed values rather than measuring an interval, so a step that does not divide 60 evenly produces an inconsistent final gap.

How do I schedule “the first Monday of the month” or “the last day of the month”?

Five-field cron cannot express those cross-field conditions directly. Quartz, AWS, and Cloudflare support # for the nth weekday and L for the last day. On Linux, schedule a coarse run and test the date inside the command, escaping % as \% if the crontab line uses date formatting. Avoid 0 9 1-7 * 1 on Vixie cron, because the OR rule makes it run on days 1–7 and every Monday, not only the first Monday.

2026年10月13日

Rep Counter: How Automatic Counting Works (2026 Guide)

A rep counter automatically tracks your repetitions during a set. In 2026, the leading options are camera-based iPhone apps such as RepCount and Push Up Counter. They follow your body on-device, count only full reps, and need no wearable or sensor.

If you actually searched for “Rep.” seat counts, the CSS counter-reset property, or a WoW reputation boost, jump to the disambiguation section below—this guide points each of those intents to where it belongs.

What Is a Rep Counter—and What Is It Actually Counting?

In a fitness context, a rep counter is a tool that automatically records how many repetitions you complete in a set. In 2026, the mainstream form factor is a phone app that uses the device camera rather than a wearable, a sensor, or a smart mat. Push Up Counter describes its own product the same way: it “turns your iPhone into a hands-free rep counter,” with “no wearables, no tapping between reps.”

The core question a rep counter answers is narrow and specific: how many did I just do? That matters because counting in your head drifts, and a partner counting out loud is not available to most people training at home. Moving the count to software returns attention to the set itself—pace, depth, and whether your hips held position.

Three things define the category as it stands today:

A primary movement is counted. Camera apps in the current market are built around one flagship exercise, most often the push-up.
Counting is automatic. The app advances the number when it sees a completed repetition, not when you tap.
Form is part of the count. A repetition that does not meet a depth or alignment threshold is typically rejected rather than added to the total.

“Rep Counter” vs. Three Other Meanings of “Rep”

The phrase “rep counter” competes with at least three unrelated intents, and it is worth routing them explicitly.

Politics. In political coverage, “Rep.” is the abbreviation for Representative, a member of the U.S. House. A “rep counter” in that context is a tally of House seats. According to the House Press Gallery, the 119th Congress stands at 218 Republicans, 214 Democrats, one Independent, and two vacancies—the vacancies following the resignations of Rep. Tony Gonzales (R-TX) on April 14, 2026 and Rep. Sheila Cherfilus-McCormick (D-FL) on April 21, 2026.

CSS. The counter-reset property creates and initializes named CSS counters; it has been baseline-available across browsers since July 2015, per MDN Web Docs. Counterintuitively, it is counter-set, not counter-reset, that resets the value of an existing counter.

Gaming. A “rep boost” in World of Warcraft refers to paid reputation farming. One vendor, Gamingcy, sells a Consortium reputation boost from $19 with a stated ETA of 7–21 days.

This guide covers only the fitness meaning.

How Does Automatic Rep Counting Actually Work? Camera, Sensor, and Manual Compared

Automatic rep counting is not one technology. There are three distinct paths, and the differences between them explain almost everything about which one works where.

Camera-Based Pose Tracking: Landmarks, Joint Angles, and a State Machine

The camera path has three stages. First, a pose model finds body landmarks in each video frame. Second, the system computes joint angles from those landmarks. Third, a state machine decides when a repetition has started and finished.

The landmark models are the same ones used across mobile computer vision. RepCount uses the iPhone’s front camera together with Apple Vision to detect body movement in real time—its product description states it “automatically counts your exercise reps so you can focus on your workout—not the numbers,” with no manual input required.

An open-source project, PushForm, documents the same pipeline in more detail in a GitHub specification. It runs a MediaPipe Tasks Vision pose landmarker on-device—the lite model, with a GPU delegate and CPU fallback, targeting 30 frames per second—and streams 33 landmarks per frame to a backend that computes angles in 2D. Elbow angle is computed for both sides each frame, median-filtered over five frames, and passed through a hysteresis state machine: the phase enters DOWN below 95 degrees and leaves DOWN above 155 degrees, with no transition inside that band. A rep is the interval from entering DOWN to leaving DOWN.

Sensor and IMU Counting: Why It Works on Barbells and Fails on Cables

The sensor path relies on an inertial measurement unit. Acceleration data is integrated to estimate a driving impulse, and thresholds decide when a repetition has occurred. This works on barbells because barbell travel is geometrically predictable.

The failure mode is documented. A BarSpeed issue reports that the impulse counter’s thresholds were “fitted at travelRatio 1.0 and are not converted through the pulley ratio, so on a stack they are applied in the wrong frame.” The measured collapse on paced machine captures was stark: a leg curl set of 10 reps produced 2 calls, a leg curl set of 12 produced 0, a single-leg press set of 8 produced 0, and a pull-up set of 8 produced 0.

Worth noting: the same issue reports that the alternative detector was measured at 108 calls against 253 over the same corpus—so narrowing the gate is not a free win either.

Manual Counting: Still the Only 100% Reliable Option (and Why That’s Fine)

Manual counting remains the correct choice in two situations: movements the camera cannot see clearly, and test sets where you need absolute certainty. A one-tap online rep counter does that job without any camera at all. It is not a failure of the technology. It is the honest answer when the sensor geometry does not match the exercise.

What Counts as a Full Rep? Depth, Lockout, and the Partial Rep Problem

Every automatic counter has to make a judgment call: is this a complete repetition, or is it a half-range wobble? The rules that decide this are where products differ most, and where a counter’s number either means something or does not.

Hard Rules First, Classifier Second: A Layered Form Gate

One public example shows how a layered gate can be built. In the PushForm specification, deterministic hard rules are evaluated at the end of each rep, in a fixed order of precedence:

Rule	Threshold	Fault label
Hip sag	mean signed hip deviation below −20°	hip sag
Hip pike	mean signed hip deviation above +20°	hip pike
Partial range of motion	minimum elbow angle exceeds 110°	partial ROM

Alongside these, the specification defines timing behavior: a DOWN phase shorter than 400 ms is a Bounce—the phase still changes, but no rep event is emitted. If the smoothed elbow angle stays inside the hysteresis band for more than 1.5 seconds, a stall event fires once and clears when the angle leaves the band. Tracking loss is declared when the tracked side’s elbow landmarks fall below 0.6 visibility for more than five consecutive frames; the machine freezes in its current phase, and the rep resumes when tracking returns.

If no hard rule fires, a trained classifier evaluates the rep’s feature vector—but a fault label is applied only when the predicted probability is at least 0.7. Below that confidence floor, the rep counts. These thresholds come from an individual open-source project, not an industry standard, and are useful as a reference point rather than a prescription.

The commercial version of the same philosophy is simpler to state. Push Up Counter says a rep “has to reach depth,” that anything shallower “is recorded as a partial rather than counted,” and that partials “form the denominator of your form score—so the number you finish with is the number you actually earned.”

Why You Can’t Judge Depth From a Front-Facing 2D Pose

A front-facing camera cannot reliably determine how far you lowered your chest—there simply is not enough depth information in a 2D pose from that angle. Push Up Counter addresses this directly: the side view is the position that checks depth, and from a front-facing 2D pose the app “counts reps but leaves the depth verdict to a side view.” If you want depth gating, camera placement is not optional.

Which Exercises Can a Rep Counter Handle—and Which Ones Break It?

This is the question the current market does not answer well. Product pages rarely state which exercises a counter handles and which ones it quietly fails on. The following decision table consolidates what the available evidence supports.

Exercise × Equipment Decision Table

Exercise / equipment	Recommended counting	Known failure risk	Basis
Push-up (bodyweight)	Camera	Needs a side view for depth; knee push-ups out of scope in PushForm v1	Push Up Counter, PushForm spec
Barbell squat / bench / deadlift	Sensor or IMU	Thresholds fitted to barbell travel; tempo’d sets may not be counted at all	BarSpeed #303
Pull-up (bodyweight + bar)	Unreliable—verify before trusting	8 performed reps produced 0 calls in measured data	BarSpeed #303
Leg curl (machine)	Manual	2/10 and 0/12 correctly detected	BarSpeed #303
Single-leg press (machine)	Manual	0/8 correctly detected	BarSpeed #303
Cable row, face pull, Pallof press	Manual	No rep marks exist in the public corpus to score them	BarSpeed #303

Why Cable and Machine Sets Are the Hardest Case

Cables and machines break automatic counting for a structural reason, not a tuning reason. An impulse counter’s thresholds are numbers in a drive frame fitted at a travel ratio of 1.0—the geometry of a barbell moving freely. On a stack, the cable and pulley ratio changes the relationship between how far the handle moves and how far the load moves. Apply barbell-frame thresholds to a stack and they land in the wrong frame.

The pull-up result is the most instructive row in the table. It is a bodyweight movement on a bar, which sounds like an easy case, yet 8 performed reps produced 0 detected calls in the measured set. “Bodyweight plus bar” does not automatically mean a camera counter will handle it either—visibility, occlusion, and the movement path all matter.

The practical conclusion is unglamorous: for cable and machine work, keep logging manually — a manual rep counter is still the reliable tool — until the technology catches up.

Does a Rep Counter Record or Upload Your Workout Video?

For camera-based apps, the privacy dividing line is simple to state: does the video ever leave the device?

The leading products in this category say no. Push Up Counter states that all processing happens on device, that no video is ever recorded or uploaded, that there are no accounts and no ads, and that counting runs offline. RepCount similarly performs detection on-device using the front camera and Apple Vision.

This is a structural advantage of the camera path rather than a marketing choice. Pose inference is light enough to run locally, so there is no technical reason to upload frames to a server for the core counting function. A cloud-dependent design would introduce both latency and a privacy surface that the on-device approach avoids.

Before installing any rep counter app, ask four questions:

Is video ever stored on the device?
Is any video or landmark data uploaded?
Is an account required?
Does counting still work with the network off?

A camera-based counter that answers “no, no, no, yes” is behaving the way the category’s leading products do.

Sets and Reps: How a Rep Counter Fits Into Your Training Log

The terminology matters because automatic counting has to understand all three terms to be useful.

Set — a continuous block of repetitions performed without rest.
Rep — one complete repetition within that block.
Rest — the interval between sets.

A counter that only knows reps produces a number with no structure. A counter that understands sets can close a set when you stop to rest and open a new one when you start again. Push Up Counter describes exactly this behavior: “Stop to rest and the app closes the set. Start again and it opens a new one—no buttons in the middle of a workout.” The PushForm specification uses the same vocabulary in a defined glossary, distinguishing a Set, a Connection, and a Rep as separate concepts.

Why does set structure matter beyond convenience? Training volume is calculated as sets × reps × load. If set boundaries are wrong, volume is wrong, and every downstream metric built on volume inherits the error. Getting sets and reps correctly segmented is the precondition for any strength-tracking logic layered on top—including estimated one-rep max.

Rep Counter Apps Compared: RepCount, Push Up Counter, and PushForm

Three products illustrate the range of approaches available in 2026.

RepCount is an iOS app that uses the iPhone’s front camera and Apple Vision to detect body movement in real time and count reps automatically, with no manual input. Its framing is about attention: count the reps so you can focus on the workout, not the numbers. Source: App Store listing.

Push Up Counter: Muscles is an iPhone and iPad app built specifically around the push-up. It offers three phone placements—facing you, beside you for a side view, or flat on the floor below your face—counts only reps that reach depth, records shallower attempts as partials that form a form score denominator, groups sets automatically, shows live pace in reps per minute, and includes a six-week program that opens with one max-effort set to set your level. Pricing is a free download with a 7-day trial, after which either a monthly subscription or a one-time lifetime unlock applies. Source: pushupcounter.app.

PushForm is an open-source Python web app documented in a public specification. Version 1 supports standard push-ups only; knee push-ups and any exercise-mode selector are explicitly out of scope, as are front-on or overhead orientations and multi-person tracking. Its value here is transparency: the full decision logic, thresholds, and acceptance criteria are readable. Source: GitHub issue #1.

Pricing models differ meaningfully. Free trial plus subscription suits users who want a maintained product without a large upfront cost. A one-time lifetime unlock suits users who dislike recurring charges. Fully open source suits developers who will self-host. Note that PushForm and BarSpeed are individual GitHub projects—they are not industry standards, and their specifications should be read as documented engineering decisions rather than consensus rules.

How to Choose: Four Questions That Decide It for You

Which exercise do you actually train most? If it is the push-up, a purpose-built counter will be more accurate than a general one.
Can you place your phone for a side view? Depth gating depends on it, so if your training space does not allow it, expect counting without depth judgment.
Does the app promise on-device processing? Confirm this in the privacy policy before installing.
Which pricing model fits how you train? A subscription is only worth it if you will still be using the app in six months.

Turning Rep Counts Into Progress: e1RM and RPE Baselines

Most of the search results in this space split into two camps: rep counting apps that never mention estimated one-rep max, and 1RM calculators that never mention rep quality. That split is a mistake, because rep quality directly determines how much you should trust an e1RM.

An e1RM is a prediction of the most weight you could lift for a single repetition, based on a submaximal set. Barbell Medicine’s 1RM calculator incorporates RPE into the estimate and notes that “lower-repetition sets, especially sets of 1 to 5 reps performed at higher RPEs, tend to produce more useful 1RM estimates.” The calculator also cautions that e1RM is a ballpark figure, not a prescription: fatigue, technique, range of motion, and day-to-day performance all affect the relationship between a training set and a true 1RM.

The dependency is direct. A set of 5 reps at RPE 9 only estimates your maximum usefully if those 5 reps were full-range. Five partial reps at the same RPE would overstate or understate your capacity depending on how the depth shortfall was scored. This is where rep counting and 1RM estimation connect: accurate counting is the input, e1RM is the output.

For context on what strong looks like, a 2024 analysis of more than 800,000 raw, drug-tested powerlifting competition results reported the following averages, as summarized by Barbell Medicine:

Lift	Men	Women
Squat	2.17× bodyweight	1.64× bodyweight
Bench press	1.5× bodyweight	0.95× bodyweight
Deadlift	2.51× bodyweight	1.98× bodyweight

For health rather than performance, a dataset of nearly 9,000 men aged 20 to 82 found that a leg press 1RM of 1.9× bodyweight and a bench press 1RM of 1.1× bodyweight were associated with the lowest risk of mortality (underlying study indexed at PubMed 18595904). No equivalent dataset exists for women, but extrapolating from data showing women average roughly 50% of male upper-body strength and 60% of lower-body strength yields targets of about 0.55× bodyweight on bench press and 1.14× on leg press.

The operating order is: get counting accurate and depth-compliant first, then use your cleanest qualifying sets to estimate e1RM, then track e1RM over time.

How to Build Your Own Rep Counter (Developer Path)

For developers, the search results in this space offer scattered GitHub issues and no consolidated path. The following six-step sequence reflects the pipeline documented in the PushForm specification.

Extract pose landmarks. Run a pose landmarker on-device—in this example, MediaPipe Tasks Vision’s lite model with a GPU delegate and CPU fallback, targeting 30 fps and streaming 33 landmarks per frame.
Compute joint angles. Calculate elbow angle for both sides every frame. Select the tracked side as the one with higher mean landmark visibility over the last 10 frames, and re-evaluate that choice only while in the UP phase.
Build a hysteresis state machine. Enter DOWN below 95 degrees, leave DOWN above 155 degrees, and do not transition inside the band. Defaults should be overridable between sets, never during one.
Apply hard rules. Evaluate deterministic faults in a fixed order—hip sag, hip pike, then partial range of motion—so the system produces form feedback even with no model loaded.
Add a grey-zone classifier. When no hard rule fires, classify the rep’s feature vector and apply a fault only at or above a confidence floor. The documented floor is 0.7: a stub classifier at 0.69 probability counts, at 0.70 it rejects.
Deploy on-device. Normalize rep features by torso length so they do not drift with camera distance, and hold the last good angle for five frames when landmark visibility drops before declaring tracking lost.

For measurable acceptance criteria, the same specification targets a macro-F1 of at least 0.85 on stratified cross-validation and at least 0.75 on a held-out collection day, over a dataset of roughly 40 reps per class collected across at least three sessions, two lighting conditions, and both facing directions.

These thresholds come from one individual open-source project. They are a concrete starting point, not a standard you are obliged to meet.

Conclusion

A rep counter in 2026 is genuinely useful—camera-based, on-device, and accurate on full reps—but its value depends entirely on knowing what it can and cannot see. Pick one camera-based app, set your phone at a side angle so depth can actually be judged, and accept that only full reps count toward your total while partials feed your form score instead. Then use your cleanest low-rep, high-RPE sets to estimate e1RM and track real strength progress. For cable and machine work, keep logging manually until the technology catches up.

FAQ

Do I need a wearable or smart mat to use a rep counter?

No. In 2026 the leading rep counters are camera-based phone apps—RepCount and Push Up Counter both work with just your phone’s camera. No wearable, sensor, or smart mat is required. The trade-off is that you need a clear camera angle and enough space in frame.

Does a rep counter count partial reps, and what happens to them?

Most quality apps deliberately do not count partial reps toward your total. Push Up Counter, for example, records anything shallower than its depth threshold as a “partial” rather than a completed rep. Those partials still show up—they form the denominator of a form score—so you can see how many attempts fell short.

Is my workout video recorded or uploaded to a server?

With the leading camera-based apps, no. Processing happens entirely on-device: no video is recorded, nothing is uploaded, no account is required, and the app works offline. Always verify this in the app’s privacy policy before installing, but on-device pose inference is now the norm for this product category.

Can a rep counter track cable, machine or leg-press sets?

Not reliably today. Sensor-based counters tuned to barbell geometry have measured near-total failure on cables and machines—one public dataset showed 2/10, 0/12, 0/8, and 0/8 reps correctly detected across four machine sets. Camera-based counting needs a clear side view and may struggle when the body moves with the cable path. For now, manual logging — or a simple online rep counter — remains the safe choice for machine work.

Is a rep counter app free, or do I need a subscription?

It varies. Push Up Counter offers a free download with a 7-day trial, then either a monthly subscription or a one-time lifetime unlock. Open-source projects like PushForm are free but require self-hosting and currently support only standard push-ups. No single pricing model dominates the category yet.

What does “rep counter” mean when it refers to Congress?

In political coverage, “Rep.” is short for Representative—a member of the U.S. House. A “Rep counter” in that context means a tally of House seats. As of the 119th Congress, the House stands at 218 Republicans, 214 Democrats, one Independent, and two vacancies (source: House Press Gallery). This is unrelated to fitness rep counting.

2026年10月9日

Luau Formatter: What It Changes and How to Use It

A Luau formatter rewrites Luau source into a consistent layout—spacing around operators, braces, keywords, and type annotations—without changing what the code does. As of October 2026, the Sparkdown formatter runs on save in its VS Code extension and web editor, and it must return identical output when run again.

What a Luau Formatter Actually Changes (and What It Never Touches)

A Luau formatter only rewrites the layout layer of a file: the space around operators, the space inside braces, line breaks and indentation after keywords, and spacing inside type annotations. The program’s behavior stays the same.

That boundary comes from the parser. A formatter edits whitespace that has already been classified into grammar nodes. It does not rename identifiers, reorder expressions, or rewrite the contents of strings. If a formatting diff changes anything other than whitespace, that is an implementation defect, not a feature.

In this ecosystem, the formatter is Sparkdown‘s document formatter, exposed as getDocumentFormattingEdits under packages/sparkdown-language-server/src/utils/providers/. The VS Code extension and the web editor share that one provider, so both surfaces produce the same output. Only the trigger—a save event versus Mod-S—and the surrounding UI differ.

Every rule follows a simple model: whitespace falls into two categories. Whitespace the parser judges to be extra is deleted. Whitespace judged to be an optional separator is normalized to exactly one space. In the ImpowerGames/impower tracker, most formatting bugs trace back to a single grammar rule putting a token in the wrong category.

The practical test for this class of formatter is the diff itself: it should show whitespace changes only. If tokens move, either you are running a different tool or the build is broken.

The two whitespace categories behind every Luau formatting rule

The difference between ExtraWhitespace and OptionalWhitespace explains behavior that otherwise looks arbitrary. ExtraWhitespace collapses to nothing. OptionalWhitespace acts as a separator: if it is empty, one space is inserted; if it contains several spaces, it shrinks to one.

Issue #1108 shows both categories hitting the same source position. In {[k] = v}, the space after ] is the first capture of LuauAssignmentOperation‘s begin, and that capture is classified ExtraWhitespace, so the formatter deletes it. In the rewritten {[k]= v}, that same capture is empty, and the leading capture of LuauAssignmentOperator becomes a zero-width OptionalWhitespace between ] and =, which the formatter widens to one space. The two classifications take turns undoing each other.

FormattingAnnotator: marking the whitespace a formatter is allowed to edit

FormattingAnnotator, in packages/sparkdown/src/compiler/classes/annotators/, labels whitespace nodes so the formatter knows what it may touch. A mid-line OptionalWhitespace is marked as a separator, which normalizes it to one space and inserts one when it is empty.

The annotator also keeps the explicit exceptions. It forces a trailing space for the keywords listed in KEYWORDS_REQUIRING_TRAILING_SPACE and for the concat operator ... That keeps the forced-space rules in one table instead of scattering them across the formatter. When a keyword or operator is missing from that table, the generic character-level rule wins and the space disappears—the source of several bugs below.

How to Run a Luau Formatter on Save in VS Code and the Web Editor

Getting format-on-save working correctly takes three steps, and the order matters. The third step is the one that protects your code.

Open a .luau file in VS Code with the Sparkdown extension installed, or open your project in the web editor. The extension and the web editor are both labeled surfaces on the same Sparkdown provider, so the rules are identical in either place.
Point the language’s default formatter at Sparkdown’s document formatter, then enable editor.formatOnSave. If you prefer not to auto-save, Mod-S in the web editor triggers the same formatting path manually. That is the recommended way to start.
Format once manually and read the diff before you enable automatic saving. Confirm the change is whitespace-only. If you turn on format-on-save first, the very first run can touch valid code before you have verified what the build does.

Steps 1 and 2 are easy. Step 3 exists because editor-side behavior is treated as a separate verification item in this project, not as a given. Issue #1226 and issue #1227 both list editor verification in their acceptance criteria: a screenshot of Mod-S in the web editor before and after, and format-on-save checked in a desktop development host or explicitly disclosed as unchecked. The rule layer can be validated in Node while the editor path has not been driven at all. Several issues state plainly, “Reproduced in Node; not checked in either editor.”

What a safe Luau formatting diff should look like

A safe diff only adds or removes spaces and blank-line whitespace—nothing else. The specific shapes to expect are the ones the tracker pins: a space added inside a one-line table’s braces, a trailing space restored after return, or several spaces reduced to one around an operator.

If you see a keyword merged into a following brace (return{1}), a type’s spacing changed (Array <number >), or a bracketed key whose = spacing flips on the next save, stop. Treat it as an implementation problem. Do not edit your source to match the output.

Confirming you are running the Sparkdown document formatter

Several extensions can register a formatter for the same language, so when output looks wrong, check which provider actually ran. Set Luau’s default formatter explicitly to Sparkdown’s document formatter instead of picking from a list each time. That way, formatOnSave cannot silently route to a different tool.

The second check is the surface. The language server’s document formatter is the shared one. If format-on-save produces a result that the same file does not produce when formatted manually through that provider, the problem is in the editor integration path, not in the spacing rules.

Luau Formatter Spacing Rules: The Before-and-After Reference

These rules are scattered across the acceptance criteria of individual issues. Gathered into one table, each row gives a string you can paste in, the output a correct formatter should produce, and the issue that pins it.

Input as typed	Expected output	Rule	Source
`{hp = 8, items = {"potion", "apple"}}`	`{ hp = 8, items = { "potion", "apple" } }`	One space after `{` and one before `}` in a one-line table	#1226
`{k = v}`, `{ 1, 2 }`, `{ [k] = v }`, `{ x: number }`, `{ number }`	`{ k = v }`, `{ 1, 2 }`, `{ [k] = v }`, `{ x: number }`, `{ number }`	Same rule for table constructors and table types	#1226
`{}` / `{ }`	`{}` / `{}`	An empty table stays tight	#1226
`{{1}, {2}}`	`{ { 1 }, { 2 } }`	Nested one-line tables	#1226
`"HP: {hp}"`, a backtick string’s `{x}`, `#value={volume}`	unchanged	Interpolation and binding braces are not tables and stay tight	#1226
A multi-line table	unchanged	Contents keep their own lines and indentation; trailing-comma rule unchanged	#1226
`return { 1 }`, `in { 1, 2 }`	space kept	A keyword before a table keeps its space	#1226
`f{ 1 }`, `a[1]`, `f(x)`	tight	A call keeps no space	#1226
`Array<number>`, `Map<string, number>`	unchanged	Generics stay tight against the type name	#1095
`number?`, `local u: number? = 1`	unchanged	Optional `?` glued to the type; `=` keeps one space each side	#1065
`local x = y :: number`	unchanged	`::` keeps one space on each side	#1099
`local x: number`, `obj:method()`	unchanged	The annotation and method-call colon stays tight	#1099
`s ..= "b"`	unchanged	`..=` keeps its leading space	#1114
`x += 1`, `x -= 1`	unchanged	The other compound assignments keep their space	#1114
`local s = 'x' .. 'y'`	`local s = "x" .. "y"`	Single-quoted strings normalize to double quotes; the operator space survives	#1097

One-line Luau tables and table types: inner brace spacing

Tables should end up as { hp = 8, items = { "potion", "apple" } }: one space after the opening brace and one before the closing brace. This applies to one-line table constructors (nodes LuauTable, LuauTypeTableStruct) and one-line table types. { a = 1 }, { 1, 2 }, { [k] = v }, { x: number }, and { number } all follow the same rule. A nested table such as {{1}, {2}} becomes { { 1 }, { 2 } }.

The rule has to be implemented by node rather than by character because { and } are the same characters used in interpolation braces. The formatter historically decided spacing by character, with { in NO_SPACE_AFTER and } in NO_SPACE_BEFORE at getDocumentFormattingEdits.ts:23. That is why { a = 1 } collapsed to {a = 1}, and why a define table and an animation block on neighboring lines could be spaced differently.

Tight suffixes: generics, optional `?`, and interpolation braces

Three suffixes stay glued to what comes before them: generic angle brackets, the optional type ?, and interpolation or binding braces.

For generics, the expected output is unchanged input. Array<number> and Map<string, number> keep no space before < or before >. At bfb1197a6, the formatter instead produced Array <number > and Map <string, number >. The cause: LuauTypeName and LuauPrimitiveType end with a trailing _LUAU_BINARY_OPERATOR_AHEAD_ whose lookahead includes the comparison operators <=? and >=?. As a result, the < and > of a generic looked like a comparison that could follow a value. The output was stable on a second pass, so a formatted script kept the misspelled type.

The same mechanism produced number ? for the optional suffix. When a value followed (local u: number? = 1), the space before = was also removed, yielding local u: number ?= 1—a line that reads like a ?= operator (issue #1065, reported at c51ea0bfa). A type without the suffix, such as local s: string, was left alone.

Operators that keep their spaces: `::` and `..=`

Two operators are easy to get wrong because the character-level rule looks past the operator.

The cast operator :: should keep one space on each side, the same as y + 1 or a .. b. At bfb1197a6, the formatter emitted local x = y:: number. The space before :: is an OptionalWhitespace passed to shouldInsertSpaceBetween; the next character is the first : of ::, and : sits in NO_SPACE_BEFORE, so the separator collapsed. The space after :: is a separate node and survived, giving an asymmetric result (issue #1099). The tight colon in local x: number and obj:method() must stay tight under the same fix.

The ..= compound assignment should keep one space before it, as += and -= already did. At 96ebb1a1d, the formatter produced s..= "b", because the character following the separator is the first . of ..= and . is in NO_SPACE_BEFORE (issue #1114). Single-quoted string normalization interacts with the same area: local s = 'x' .. 'y' was normalized to double quotes, but the first pass also dropped the operator space to "x".. "y", and only a second format restored it (issue #1097).

Formatting Idempotency: Why Formatting a Luau File Twice Must Match Once

A formatter must be a fixed point of its own output: running it on an already-formatted file must reproduce that file character for character. This is not an optimization target—it is the acceptance bar. Issue pages in this project say so directly, requiring that “formatting twice gives the same text as formatting once for every format fixture.”

Idempotency matters because format-on-save runs the formatter on every save. If the formatter is not a fixed point, a whitespace position can flip back and forth on each save. That creates a permanent noise diff and makes code review useless—reviewers cannot tell a real change from the formatter toggling a space.

The clearest failure case is issue #1108, reported at 34406f45f. local t = {["k"] = "v"} becomes {[k]= v} on the first pass, and the second pass restores {[k] = v}. The root cause is the category mismatch described earlier: the space after ] is captured as ExtraWhitespace and deleted in one spelling, while in the other spelling the capture is empty and only a zero-width OptionalWhitespace remains, which is widened to a space. Each pass undoes the last.

A second, more dangerous class exists: output that is stable but wrong. A second-pass check cannot detect that a misspelling has been frozen in place. The generic and optional-type cases are exactly this—Array <number > and number ? were both verified to be stable on a second pass, so no round-trip test would have caught them. Idempotency is necessary, not sufficient.

A two-pass idempotency test you can run on any Luau formatter

The test is mechanical. You do not need to read the formatter’s source or trace the maintainer’s code paths.

Save a copy of the file you want to check.
Format it once and save the result as pass 1.
Format pass 1 and save the result as pass 2.
Diff pass 1 against pass 2. Any byte of difference means the build is not a fixed point.

In this repository, the same check runs in-process. formatSource in packages/sparkdown-language-server/src/tests/formatter/formatSource.ts drives getDocumentFormattingEdits over a source string, which is how the reproductions above were measured. deltaFormatEquivalence.test.ts passes only when the invariants hold. You can exercise a single file with node scripts/test-suite.mjs run packages/sparkdown-language-server src/tests/formatter/<your-test>.test.ts --wait 1500.

Two habits make the test more useful. First, run three passes instead of two when a quote-normalization change is involved—the 'x' .. 'y' case needed a second format before the operator space reappeared. Second, treat “pass 1 equals pass 2” as only half a result. Inspect the text itself for spellings nobody would write.

When a Luau Formatter Rewrites Good Code Into Odd-Looking Spelling

The most visible symptom is a keyword glued to the brace or string that follows it. return {1} becomes return{1}, for k in {1, 2} do becomes for k in{1, 2} do, if c then {1} else {2} becomes then{1} else{2}, and return [[s]] becomes return[[s]] (issue #1110, reported at 34406f45f).

The code still runs, but the formatter has rewritten common, well-formed Luau into a spelling that reads like a function call—f{1}—which is not how anyone writes the language. The reported cases cover four keywords: return, in, then, and else. The result is stable on a second pass, so nothing self-corrects.

The second symptom is type-layer output that is stable and wrong: Array <number >, number ?, and y:: number from the previous section. All three survive a round-trip check.

A related trap is confusing brace bodies with tables. The { ... } bodies of layout, style, animation, theme, component, morph, and screen blocks are indented by block depth, and a closing } lines up with its header. They are not Luau tables, so the one-line table spacing rule does not apply. A one-line block gets { a = 1; b = 2 }, with one space after each ;, one space before a block’s {, and one on each side of an optional =. Issue #1226 explicitly leaves those brace blocks out of its scope.

Is it my code or the formatter build? A short diagnostic order

Work through these in order, and stop at the first one that matches.

Is the diff whitespace-only? If so, the formatter is behaving within its design. If tokens moved, were renamed, or a string’s contents changed, stop—that is a defect, and you should not commit the result.
Does the output change on a second run? Format the result again and diff. A difference means a broken fixed point; record the input, the build, and the two outputs.
Is the output stable but spelled unusually? If return{1} or Array <number > comes back unchanged the second time, the problem is the implementation’s spacing rules, not your source. Rewriting your code to match it will only make the file harder to read.

The ImpowerGames/impower root cause and umbrella issue #1104

Most of these cases converge on one underlying defect class. Issues #1108, #1110, #1095, #1099, and #1114 are each filed under the umbrella item #1104 and fixed on pull request #1107. They are different surface symptoms of the same whitespace-classification problem, found by successive review rounds of that PR.

The structural reason they cluster is that the formatter historically decided spacing by looking at adjacent characters, while the classification lives on parse-tree nodes. FormattingAnnotator forces a trailing space for keywords in KEYWORDS_REQUIRING_TRAILING_SPACE, but only when the next character is (. A keyword followed by { or [ therefore fell through to the call-gluing rule, and else was absent from the set entirely. The ..= operator was missing from the forced-space table that already covered ...

The tracker also carries the proposed regression net for this class. Issue #996 describes a layout-variant oracle: take Luau code that already runs green, mechanically rewrite it into one-line, glued-keyword, semicolon, and comment layouts that Luau defines as equivalent, run each rewrite through the conformance harness, and require the same output text, the same returnedOK, and no new diagnostic. The motivation is that 22 upstream conformance fixtures passed end to end while ten layout-specific Luau bugs filed between 2026-09-26 and 2026-09-27 existed, because those fixtures are all written in one tidy layout.

Version and Status Caveats: What’s Fixed and What’s Still Open

Every rule in this article is tied to a specific issue and commit, because formatting defects are fixed continuously and different builds can behave differently. The commits cited above are 34406f45f (the bracketed-key spacing toggle and the keyword gluing), bfb1197a6 (generics, ::, single-quoted strings), c51ea0bfa (optional ?), 96ebb1a1d (..=), and 871116e (the line numbers cited for the table-spacing rule).

Before treating any row of the reference table as current behavior, re-test it against the build you actually have. The behavior descriptions here reflect the state of the code at those commits, and the issue pages show the related work landing on pull requests #1077, #1107, and #1245.

Two items were still shown as open at the time of writing in early October 2026. #1227, which formats brace bodies by brace depth across the seven block constructs in the Sparkle UI system, is open with a linked pull request. #996, the Luau conformance layout-variant oracle, is also open.

One scope limit applies to everything here: the whole corpus comes from the tracker of a single repository, ImpowerGames/impower. These rules describe the conventions of one implementation—Sparkdown, as shared by the VS Code extension and the web editor—not an official standard of the Luau language.

Conclusion

A trustworthy Luau formatter changes only layout, never semantics, and it must be a fixed point of its own output. You can verify both properties yourself: check that the diff contains nothing but whitespace, and check that a second run produces a byte-identical result to the first. Start by formatting a scratch file with the constructs you use most—one-line tables, optional types, generics, ::. An online Luau formatter is the zero-setup scratchpad for that first pass: paste the script, get back cleanly indented code with type annotations preserved, then compare the spacing decisions against the reference table above. Only then turn on format-on-save. When a file keeps changing on the second run, record the issue and the build version instead of rewriting your code to match the formatter.

FAQ

Why does formatting my Luau file twice give a different result than formatting it once?

The formatter is not a fixed point: a whitespace position is deleted on one pass and restored on the next. The typical cause is that the same position is classified as extra whitespace in one spelling and optional whitespace in another (issue #1108). Diff the two outputs and report it to the maintainers rather than editing your code.

Does running a Luau formatter change what my code does?

By design, it does not. It changes whitespace, line breaks, and space inside braces, and never token order or string contents. When the tool is working correctly, a formatting diff is always a whitespace-only diff. If you see keywords reordered or a string rewritten, that is an implementation defect—do not commit it.

What spacing should a one-line Luau table use inside its braces?

One space after { and one before }, giving { a = 1 }, { 1, 2 }, { [k] = v }, and { x: number }. An empty table stays {}, and { } collapses to {}. Interpolation and binding braces stay tight—"HP: {hp}", a backtick string’s {x}. Multi-line tables and the trailing-comma rule are unchanged (issue #1226).

How do I run a Luau formatter on save in VS Code?

Point Luau’s default formatter at Sparkdown’s document formatter so no other extension competes for the language, then enable editor.formatOnSave. In the web editor, Mod-S triggers the same path manually. Format one file by hand and read the diff before you enable automatic saving, so the first run cannot touch valid code unsupervised. And if you just need one snippet cleaned up with no editor configuration at all, an online Luau formatter does the job in the browser—free, no registration, with Roblox style-guide defaults.

Why did formatting turn `return {1}` into `return{1}`?

The separator rule glues a call-like opener—(, [, or {—to the preceding word character so that f(x), a[1], and f{1} stay tight. FormattingAnnotator exempts certain keywords with a forced space, but only when the next character is (. else is also missing from that keyword set, so then{1} and else{2} glue the same way (issue #1110). The code still runs, but it reads like a function call; this is an implementation defect, not a problem with your source.

2026年10月9日

Date Calculator Between Two Dates: Days & Workdays

A date calculator between two dates gives you the exact span between any two calendar dates—total days, weeks, months, and years. Enter a start date and an end date, choose whether the end day counts, and it handles leap years for you.

How to Use a Date Calculator Between Two Dates (Step by Step)

A date calculator between two dates turns two calendar dates into a precise duration. It does the math that manual counting often gets wrong: different month lengths, leap years, weekends, and whether the final day belongs in the total. Most tools return the result in several formats at once, so you can pick the unit that fits the task.

Step 1: Enter Your Start Date & End Date Inputs

Open the tool and find the two required fields: Start Date and End Date. Put the earlier date in Start Date and the later date in End Date. If your tool accepts dates in any order, the result is still an absolute span. Use ISO-8601 format when you can, such as 2026-10-01, because 03/04/2026 means April 3 in most countries and March 4 in the US. CalcOpenly notes that ISO dates avoid that day-and-month ambiguity.

Step 2: Decide Whether the End Date Counts

Look for an “include end date” or “count end date too” toggle. Most calculators default to exclusive counting: end date minus start date, so the final day is not counted. Turn the toggle on when both the first and last day count, such as leave from Monday to Friday or an event that runs through its closing day. This one setting changes the answer by exactly one day.

Step 3: Read Total Days, Weeks, Months, and Years

Read the output in four forms: total days, total weeks and leftover days, calendar-style years/months/days, and business days. For example, CalcOpenly shows that 1 October 2026 to 30 December 2026 is 90 total days, 0 years 2 months 29 days, 12 weeks and 6 days, and 64 business days. Use that worked span as a sanity check. If a tool returns a different total for the same dates, check its end-date and holiday settings before you trust it. The Gregorian calendar is the default basis for these counts.

Inclusive vs Exclusive Counting: Does the Calculator Count the End Date?

The default convention in most date calculators is exclusive: the calculator counts the first date but not the ending date. As PK, DQYDJ notes: “it counts the day of the first date but not the ending date.” That is the same convention as subtracting one date from another in a spreadsheet, which is why adding one day to the end date often produces the “missing” day.

Four Common Day-Count Conventions Compared

Four conventions cover most real-world uses. Subtraction, the calculator default, counts the start date and leaves out the end date. Inclusive counting counts both the start and end dates. Clear days exclude both ends. US federal court deadlines use a fourth rule: exclude the trigger day, count every day including weekends and holidays, and include the last day unless it falls on a weekend or holiday. CalcOpenly shows that the same 3–10 March 2026 pair can be 7, 8, or 6 days depending on the convention.

Why ‘3 March to 10 March’ Can Be 6, 7, or 8 Days

Take Tuesday 3 March 2026 to Tuesday 10 March 2026. Subtracting the dates gives 7 days. Counting both ends gives 8 days. Counting neither end gives 6 clear days. Contracts, leave policies, and statutes each pick a convention, and choosing the wrong one moves a deadline by a day. The fix is to state the convention in writing, not to assume the calculator’s default matches the rule you are working under.

US Federal Court Deadline Rule: FRCP 6(a)

US federal courts follow Federal Rule of Civil Procedure 6(a). The rule excludes the day of the triggering event, counts every day including weekends and holidays, and includes the last day—but if that last day is a Saturday, Sunday, or legal holiday, the period runs to the next day that is not. A 10-day period triggered on Wednesday 4 March 2026 would end on Saturday 14 March, so it moves to Monday 16 March. This is a non-default convention, which is why a standard calculator’s exclusive-end result should not be used for court deadlines without adjustment.

How Leap Years and Unequal Months Affect Your Result

Leap years and uneven month lengths are the two biggest reasons manual date math fails. A leap year adds one day to any span that crosses February 29. Unequal months mean a month is not a fixed number of days, so month-based output depends on where the span starts.

Why February 29 Adds One Day

The Gregorian calendar adds February 29 in years divisible by 4, except century years not divisible by 400. According to datecalculatorbetweentwodates.com, January 1, 2024 to January 1, 2025 is 366 days, while January 1, 2025 to January 1, 2026 is 365 days. A span that crosses February 29 gains one day; a span that does not cross it does not. That single day can change a payroll period, an interest calculation, or a contract deadline.

Month-End Clamping: Jan 31 + 1 Month = Feb 28 or 29

Month addition uses end-of-month clamping. If the target month does not have the same day number, the result becomes the last day of that month. CalcOpenly gives the example of 2024-01-31 to 2024-03-01: the total is 30 days, which reads as 1 month and 1 day because one month after 31 January is 29 February in a leap year. In a non-leap year, 31 January plus one month becomes 28 February. The same rule explains why 31 January to 1 March 2024 is not a clean two months.

Gregorian Calendar: The Default Basis for Every Count

Standard date calculators use the Gregorian calendar for all dates, including dates before the calendar was widely adopted. The Gregorian calendar divides the year into 365 days, with 366 in a leap year, and splits it into 12 months of 28–31 days. Its leap-year cycle is why a calculator can return an exact result while manual counting drifts by a day or more.

Calendar Days vs Business Days: What Changes?

Calendar days count every date on the calendar, including Saturdays, Sundays, and holidays. Business days, also called working days, usually count Monday through Friday only, and may exclude public holidays. The distinction matters when you are measuring elapsed time versus time available for work.

How Business Day Counts Treat Weekends

By default, business day counts treat Saturday and Sunday as non-working days. A tool may also let you change the weekend definition—for example, Friday–Saturday—but the common US setting is Saturday–Sunday. CalcOpenly notes that whole-day counts are unaffected by time zones and daylight saving time; only hour-level counts change when a clock change falls inside the span. If you need Monday–Friday only, use the business-day output rather than the total-days output.

Do Business Day Calculators Exclude US Public Holidays?

Most basic calculators exclude weekends only and leave public holidays to the user. File Toolbox states plainly that its business-day count skips Saturdays and Sundays only and does not remove public holidays. Some tools offer a US federal holiday option or a custom holiday list; HowLongAgoGo says its business-day calculator includes public holidays for multiple countries, including the US. Check whether observed holidays are included before relying on the result for payroll or legal deadlines.

The Weekend Holiday Double-Count Trap

A holiday that falls on a weekend should not be subtracted twice. If a Saturday holiday is observed on Friday, you subtract one working day, not two. CalcOpenly illustrates the principle with England and Wales: over Christmas 2026, 26 December is a Saturday and the Boxing Day bank holiday moves to Monday 28 December, so listing 25, 26, and 28 December removes three days, not four. For US dates, the same logic applies to any federal holiday whose observed date shifts to a weekday.

Weeks, Months, and Years Breakdown: Why 90 Days Isn’t Always 3 Months

The calendar-style output expresses total days as X years, Y months, and Z days. That format is useful for human-readable durations, but it is not a fixed fraction of a year. Months have 28–31 days, so the same number of total days can produce different month-and-day labels depending on the start date.

Reading the Calendar-Style Output

For the days between two dates such as 1 October 2026 to 30 December 2026, the result is 90 total days, 0 years, 2 months, and 29 days, or 12 weeks and 6 days. CalcOpenly also shows 2,160 hours for that span. Additional outputs may include minutes and seconds, calculated as days × 1,440 and days × 86,400. Read the calendar-style line as a convenience label, not as a substitute for the total-day count.

When to Trust Total Days Instead

Use total days when the deadline, billing period, or contract is written in days. In the CalcOpenly example, two spans with the same number of days can show different months-and-days figures because of month-end clamping. If the agreement says “30 days,” use 30 days. If it says “one month,” use the month figure and check the end-of-month case. Total days is the safest unit for legal deadlines, payroll, and interest calculations.

Add or Subtract From a Date: The Companion Tool

The add/subtract side of the date calculator answers a different question: what date is X days, weeks, months, or years from a starting point? It is the companion to the between-dates calculator, not a replacement for it. The between-dates calculator measures the span between two known dates; the add/subtract tool projects a new date from one known date and a duration.

Add/Subtract Years, Months, Weeks, or Days

Enter a start date, choose add or subtract, and enter the amount and unit. The tool uses the same Gregorian calendar and end-of-month clamping rules as the between-dates calculator. For example, 2024-01-31 plus one month equals 2024-02-29 in a leap year, not March 2 or March 3. When the target month lacks the start day, the result falls on the last day of that month.

Related Tool: Age Calculator

An age calculator applies the same date arithmetic to a birth date and a reference date. It returns exact age in years, months, and days, and may also show total days alive, weeks, and the next birthday. Event countdown tools use the same span logic in the opposite direction, counting down to a future date such as a holiday, wedding, or product launch. Keep the between-dates calculator focused on the two-date span and link out for add/subtract, age, and countdown queries.

US Scenarios: Payroll, Court Deadlines, and Project Planning

US users often need more than a simple day count. Payroll, court deadlines, and project schedules depend on business days, holiday observance, and a clearly stated counting convention. The same two dates can produce different answers in each context if the convention is not defined.

Federal Court Deadlines

Federal court deadlines follow FRCP 6(a): exclude the trigger day, count weekends and holidays, include the last day, and roll forward if the last day is a weekend or holiday. A 10-day period triggered on Wednesday 4 March 2026 ends on Monday 16 March 2026 because the natural end date is Saturday 14 March. A standard calculator set to exclusive-end counting would return 10 days but would not perform the roll-forward, so use a court-deadline tool or apply the rule manually.

Payroll and Billing Cycles (Weekly, Biweekly, Semi-Monthly)

Weekly, biweekly, and semi-monthly payroll cycles depend on business-day counts and holiday observance. Use business-day output to estimate pay periods and billing periods, then subtract any US federal holidays that fall on weekdays. If the holiday is observed on a different date, adjust the count so the holiday is not double-counted. The same logic applies to invoice due dates and notice periods.

Project Planning with Business Days

Project schedules should use a holiday-aware business day count, not a simple Monday–Friday filter. For 2026, start from 261 weekdays and subtract US federal holidays that fall on weekdays; verify observed dates for Saturday and Sunday holidays. CalcOpenly reports 261 weekdays in 2026 and, for England and Wales, 253 working days after 8 bank holidays, which shows how much a jurisdiction’s holiday list changes the result. State the counting convention—inclusive, exclusive, or clear days—in the project plan so no one has to guess.

How to Verify a Date Calculator’s Result Yourself

A date calculator is only as reliable as its counting convention and its date arithmetic. You can verify the days between two dates without writing a full date library by using serial day numbers, spreadsheet functions with caveats, or public reference algorithms. The goal is to confirm the total-day figure first, then check the business-day and holiday adjustments.

Manual Check: Serial Day Numbers

Convert each date to a serial day number, then subtract. CalcOpenly describes the method as counting days since 1 January 1970 in the proleptic Gregorian calendar: 1 October 2026 is day 20,727 and 30 December 2026 is day 20,817, so the difference is 90 days. This method handles leap years and month lengths automatically. If your manual result differs, check whether you counted the end date and whether the dates were entered in ISO order.

Excel DATEDIF and NETWORKDAYS Caveats

Excel’s DATEDIF can return incorrect results for specific edge cases. As Dave Bruns, Exceljet, notes: “The DATEDIF function can return incorrect results with date pairs that touch specific edge cases.” Exceljet suggests a custom lambda function as a workaround. NETWORKDAYS has a different trap: it counts both the start and end dates by default. According to Microsoft’s NETWORKDAYS example, 1 October 2012 to 1 March 2013 returns 110 working days, while a calculator with a default exclusive end date returns 109 for the same dates because 1 March 2013 is left out. NumPy’s busday_count also excludes the end date.

Open Standards and Reference Implementations

For independent verification, use a published standard and a public algorithm. ISO 8601-1:2019 defines date and time representations for information interchange. Howard Hinnant’s days_from_civil algorithm provides a chrono-compatible low-level date calculation that many implementations use. A calculator that publishes its last-reviewed date and states its default counting convention gives you the two pieces of information you need to trust or challenge its result.

Conclusion

A date calculator between two dates is only as reliable as the counting convention behind it. Total days, inclusive versus exclusive rules, leap years, and business-day definitions all change the answer. Before you rely on a result for payroll, court, or project planning, confirm whether the tool includes the end date, excludes US federal holidays, and uses the Gregorian calendar. Run one worked example—2026-10-01 to 2026-12-30 = 90 days and 64 business days—to verify the tool’s behavior before you use it for a real deadline.

FAQ

Does the calculator count both the start date and the end date?

Most date calculators default to exclusive counting: end minus start, so the end date is not counted. An “include end date” toggle adds one day, making the count inclusive. Check the convention before using the result for contracts or legal deadlines.

How does the date calculator handle leap years and February 29?

It uses the Gregorian calendar, which adds February 29 in years divisible by 4, except century years not divisible by 400. A span crossing February 29 gains one day; for example, January 1, 2024 to January 1, 2025 is 366 days. Month additions that land on February 29 are valid only in leap years.

How many working days are there in 2026?

2026 has 261 weekdays, Monday through Friday, according to CalcOpenly. Subtract US federal holidays that fall on weekdays to get the working-day count; verify observed dates for weekend holidays. The final number depends on the jurisdiction and whether observed holidays are excluded.

Do business day counts exclude US public holidays?

Most basic calculators only exclude weekends and leave holidays to the user. Some tools offer a US federal holiday option or a custom holiday list. Always verify whether observed holidays—such as a Friday observance for a Saturday holiday—are included in the exclusion.

What is the difference between calendar days and business days?

Calendar days count every date, including weekends and holidays. Business days typically count Monday–Friday only, and may exclude public holidays. Use calendar days for elapsed time; use business days for project schedules, payroll, and legal deadlines.

Can I calculate the number of days between two dates in the past or the future?

Yes. The calculator works for any two valid Gregorian dates, past or future. The result is always an absolute span; direction only matters for add/subtract tools. For future dates, confirm whether daylight saving or time zone changes matter—whole-day counts do not.

2026年10月9日

Resize GIF Without Losing Quality, Speed, or Transparency

To resize a GIF, drop the animation into a browser-based GIF resizer, type your target width in pixels with the aspect-ratio lock on, and export. Downscaling cuts file size while frame delays, transparency and looping stay intact — but shrinking never adds detail, so always start from your largest source file.

How Resizing an Animated GIF Actually Works — and What Stays Intact

Resizing an animated GIF changes exactly one thing: the pixel canvas. Everything else in the file — how long each frame stays on screen, how many times the animation loops, which colour index is marked transparent — gets copied through untouched by a competent resize operation.

Understand that, and most of what follows in this guide falls into place. AREM Labs states it directly: a standard pixel-dimension resize adjusts only the canvas width and height while leaving frame delay values and looping metadata untouched. As long as the file passes through a tool that preserves frame timing arrays, playback speed stays identical to the original.

The complication is that you’re not resizing one image. You’re resizing a stack of them.

Why a GIF Is Not Just a Bigger PNG

A GIF stores every frame as its own separate, palette-indexed image. According to ShotEdit, a three-second screen recording at 15 frames per second is 45 stored images. That’s why file size grows with duration, frame rate and pixel area all at once — and why a short clip that would be a 400 KB MP4 turns into a 6 MB GIF.

Each of those stored frames carries its own 256-colour palette. GIF supports a maximum of 256 colours per frame, and no amount of scaling changes that ceiling. When a resizer redraws a frame at a new size, it has to re-quantise the colours to fit that palette again. This is the mechanical reason gradients and dither patterns are where visible banding shows up, even when you’re only shrinking.

What Survives a Resize: Frame Delay, Loop Count, Transparency

A resize operation doesn’t touch the frame delay array, the loop/NETSCAPE extension, or the transparent colour index. Those are separate data structures in the file, and a pixel-level resize has no reason to rewrite them.

That’s why playback speed, loop count and transparency survive a correctly executed resize. It’s also why they break when a tool takes a different route — extracting every frame as a separate still, resizing the stills, and rebuilding the animation from scratch. Rebuilding is where timing drifts, loops reset to defaults, and the transparent colour index gets dropped or replaced with a background colour.

How to Resize a GIF in a Browser, Step by Step

This is the primary workflow, and the order matters. Skip a step and you get a specific, predictable failure.

Step 1 — Start from the largest, cleanest source you own. Resizing never adds detail, so your source file defines the ceiling on quality. If you have a video export and a small GIF, resize the video export down rather than upscaling the GIF.

Step 2 — Open a browser-based GIF resizer and load the file. EZGIF and similar no-signup tools cover resize, optimize and crop in one place, so you can complete the whole workflow without moving between applications or formats.

Step 3 — Enter width in pixels with the aspect-ratio lock ON. Let height calculate itself. Typing both numbers manually is how people stretch GIFs. AREM Labs lists unlocking the aspect ratio during scaling as the first common mistake: entering custom width and height values without the lock stretches characters, distorts circular products, and ruins brand presentation.

Step 4 — Pick a resampling method deliberately. Use smooth interpolation — Lanczos, bicubic or bilinear — for footage, reaction clips and photos. Use nearest neighbour for pixel art, UI captures, emoji and anything with hard one-pixel edges. ConvertICO labels these as “Photo and video” (Lanczos), “Smooth pixel art” (xBR/hqx) and “Crisp pixels” (nearest neighbour), and notes that crisp-pixel scaling uses whole-number scales so every original pixel grows by the same amount.

Step 5 — Preview before exporting. Check the busiest frame in the animation for banding, then compare the before and after file size. The busiest frame is where palette re-quantisation shows first.

Step 6 — Only if the export still misses the cap, move to the GIF compression levers. Resize stays first because it’s the most predictable reduction for the least visible quality cost.

Know the ceilings. Browser tools impose input limits — ConvertICO’s GIF upscaler accepts GIFs up to 25 MB and 500 frames with a maximum of 4000 px on any side, and other tool pages in the same category cite comparable figures. Oversized sources need pre-trimming before they will load at all.

The Order That Matters: Resize First, Compress Second

Resize, then compress. Not the other way around.

AREM Labs is explicit about why: compressing a file first introduces compression artifacts that get magnified and blurred during dimensional downscaling, and applying colour optimisation as the final step avoids that. ShotEdit agrees on the ordering logic, calling resize “usually the most reliable reduction when a GIF must hit a hard limit” and noting it costs the least visible quality per kilobyte saved.

Three Mistakes That Ruin the Output

Unlocking the aspect ratio. Custom width and height values without the lock stretch the subject. Keep the lock on and type one number.

Relying on browser CSS for downscaling. Uploading a 2000 px wide GIF and setting its display width to 400 px via HTML or CSS forces every visitor to download the full payload. AREM Labs puts the figure at roughly 15 MB in that example. Resize the source file itself.

Resizing without checking the frame count first. Scaling a 500-frame animation still produces an oversized file even at low resolutions. Trim unneeded frames before scaling, not after.

How Much Does Resizing Actually Shrink a GIF?

The pixel math is the whole story, and it’s simpler than most guides make it sound. Halving both dimensions removes 75% of per-frame pixel area. That’s the primary source of the weight drop.

AREM Labs puts it plainly: resizing from 1200×800 to 600×400 cuts the total pixel count per frame by 75 percent, which drastically reduces overall byte size. The weight doesn’t fall by exactly 75% — palette data, headers and per-frame metadata don’t scale with area — but the pixel reduction is what drives the result.

Lossy re-encoding stacks on top of that, but it plateaus fast. According to ShotEdit, a 228 KB clip dropped a further 7% at lossy 30, 15% at lossy 80, and 25% at lossy 200. In other words, tripling the lossy strength from 80 to 200 bought roughly ten additional percentage points of savings — and that’s the point at which dithering artifacts typically become visible.

The Pixels-Are-the-File Rule

GIF file size is driven by frame count multiplied by pixel area multiplied by palette complexity. Resizing touches only the middle term.

This is why the same resize produces dramatically different results on different files. A GIF that was already reduced to a small palette and a low frame rate has little left to give. A GIF freshly exported from a video with a full 256-colour palette per frame has a lot. Two GIFs at identical dimensions can still differ several-fold in weight, so dimensions alone never predict whether you’ll clear a file cap.

Why a Second Round of Compression Barely Helps

GIF has no true inter-frame compression. ShotEdit’s explanation is the clearest available: GIF stores only the changed rectangle of each frame beyond that, so once a recording has gradients, dithering or camera motion, almost every pixel changes between frames and the optimisation stops helping.

That leaves very little redundancy for a second optimisation pass to find. ShotEdit states that palette reduction and lossy smoothing are not cumulative in a useful way, and that re-running the same compression settings returns almost the same file. The practical advice that follows: change a setting instead — resize the frames or lower the colour count — rather than repeating an identical pass.

One exception worth knowing: flat UI recordings and screen captures of interfaces with solid fills barely shrink at any lossy setting, because there’s no dither noise left to smooth away. ShotEdit notes that a two-colour loading spinner is already close to the smallest a GIF can be.

Target Sizes and File-Weight Limits by Platform, Checked October 2026

One table, several channels, with the caveats that most single-source guides leave out. Platform limits drift, and Discord is the clearest current example.

The One-Page Size + Weight + Format Table

Platform / Use Case	Target Dimensions	Target Weight	Transparency	Source & Check Date
Shopify / WooCommerce product body	600×600 to 800×800 px	Under 2 MB	Supported	AREM Labs, 2026-09-16
Amazon A+ Content (full-width banner)	970×600 px	Under 3 MB	Supported	AREM Labs, 2026-09-16
Email newsletters (Klaviyo / Mailchimp)	480×480 to 600×400 px	Under 1 MB	Supported	AREM Labs, 2026-09-16
Etsy product descriptions	750×500 px	Under 1.5 MB	Supported	AREM Labs, 2026-09-16
X / LinkedIn inline feed	800×450 to 1200×675 px	Under 5 MB	Supported	AREM Labs, 2026-09-16
UI / software documentation tooltips	400×300 px	Under 800 KB	Supported	AREM Labs, 2026-09-16
Discord chat (free account)	—	8 MB or 10 MB — sources conflict	Supported	Wondershare, updated 2026-09-23; ShotEdit, 2026-09-05
Discord chat (Nitro Classic / Nitro)	—	50 MB / 100 MB	Supported	Wondershare, updated 2026-09-23
Discord animated emoji	128×128 px	Under 256 KB	Supported	Wondershare, updated 2026-09-23
Slack channel	—	Uploads fine, but over ~2 MB loads slowly	Supported	ShotEdit, 2026-09-05
GitHub issues and pull requests	—	10 MB per file	Supported	ShotEdit, 2026-09-05
Email attachment (whole message)	—	Commonly 20–25 MB	Supported	ShotEdit, 2026-09-05

The Discord conflict deserves to be stated plainly rather than smoothed over. Wondershare’s guide, updated 2026-09-23, lists 8 MB for free accounts — while the same article’s own body text still refers to itself as a 2025 guide. ShotEdit, checked 2026-09-05, lists 10 MB for a free Discord account. Two sources, both recent, disagree by 2 MB. Treat the figure as tier- and date-dependent, and verify against Discord’s own help centre before you commit a workflow to it.

For Web Pages: Why GIF Weight Is a Core Web Vitals Problem

An inline e-commerce product animation is best served at 480–800 px wide, per AREM Labs. That range keeps details sharp on mobile screens and desktop monitors while keeping the payload light enough to maintain fast Core Web Vitals and quick checkout loading.

This matters more for GIFs than for stills because a GIF is downloaded in full by every visitor. Oversized GIF dimensions slow mobile page loading, consume customer data plans, and cause layout shifts on responsive themes — and a layout-shift problem is a Core Web Vitals problem, not merely a storage problem.

AREM Labs frames the fix directly: resize the GIF to the exact pixel dimensions of its container rather than uploading a large file and scaling it down in CSS. If the container is 640 px wide and the animation is 1920 px wide, the answer is a 640 px asset, not a 1920 px asset with a width attribute.

Discord Emoji at 128×128 and 256 KB: The Strictest Target

Animated emoji is the tightest real-world constraint in this table. Wondershare states that animated emoji must be 128×128 pixels, and that regardless of dimensions, all emoji uploads — static or animated — must be under 256 KB.

Meeting both at once is a two-stage job: get the dimensions right first, then attack the weight. This is exactly why resize-then-recolour is the reliable path — 128×128 alone will often not clear 256 KB for a detailed animation, so colour reduction and frame dropping follow the resize. Note also that animated server emoji require Nitro to display as animated, per the same source.

Does an Online GIF Resizer Upload My File?

It depends entirely on the tool’s architecture, and the two architectures are easy to tell apart once you know what to look for.

Server-side tools upload your file, process it on a remote machine, and return the result. Client-side tools process everything on your device — the pixel data never leaves the browser. ImageLean’s GIF resizer is built on the client-side model: resizing runs in your browser, so the animation is not uploaded anywhere. ShotEdit describes its own architecture explicitly: compression runs locally with gifsicle compiled to WebAssembly, and the file never leaves your device.

The practical stakes aren’t abstract. Unreleased product recordings, client footage under NDA, internal dashboards, and anything governed by a company data policy should not be handed to an unknown server. ShotEdit names unreleased product recordings and customer data as the specific cases where local processing matters.

How to Verify a Tool Really Processes Locally

You can test this in about thirty seconds rather than trusting a privacy badge.

Load the tool, start a resize or compression on a reasonably large GIF, then disconnect your network — pull the Wi-Fi or unplug the ethernet. If processing is genuinely client-side, the job finishes normally. If it’s server-side, the job dies when the connection does.

Before uploading anything sensitive, also check three things: the tool’s privacy page, whether its output is re-encoded through a native engine or re-drawn from extracted frames, and whether it states a retention window. ConvertICO, for example, states that uploads and results get random names and are deleted automatically within four hours — a server-side tool with a disclosed retention policy rather than a no-upload tool.

Resize vs Compress vs Crop vs Trim: Pick the Right Lever First

Four terms, four different variables in the file. Confusing them is why people crop a GIF, see no size change, and conclude the tool is broken.

Resize changes canvas dimensions. Compress changes the colour or lossy budget. Crop changes the visible area. Trim changes duration.

The recommended execution order is: resize → tune lossy strength → reduce colours → drop frames. Stop the moment the file clears the cap. ShotEdit gives the reasoning behind the ordering — resize first because it’s the most predictable reduction and costs the least visible quality per kilobyte saved, then lossy strength until artifacts appear in flat areas, then colour count while watching gradients for banding, then frame dropping last because choppy motion is usually the first thing a viewer notices.

Cropping deserves a specific warning. It can leave file weight almost untouched if the frame count and per-frame palette stay the same, which is why the experience of “I cropped it and it’s still too big” is so common. Crop to fix framing; don’t expect it to fix weight.

Trimming is the opposite: it cuts weight roughly in proportion to the frames removed, making it the bluntest but most effective lever for long recordings.

The stop-loss rule. When a long recording still misses the target at acceptable quality, change format rather than degrading further. ConvertICO reports that in its tests a 2x animated WebP came in at under one third of the size of the equivalent 2x GIF while holding more detail. Animated WebP keeps full colour and smooth transparency; MP4 is smallest of all but has no alpha channel, so transparent areas become white. ShotEdit’s blunt version of this advice: if a GIF still won’t fit after the strongest compression setting, the clip is too long for the format, and the same recording as MP4 is routinely five to ten times smaller at better quality.

Keep GIF output only where the destination demands a .gif file — chat clients, forums, some email clients, and template-bound CMS embeds.

A Decision Table: Which Lever Fixes Which Problem

Problem	Lever	Why
Wrong display size on the page	Resize	Changes the canvas; also the largest predictable weight win
Over a hard upload cap, dimensions already correct	Compress (lossy strength)	Reduces palette rounding noise; biggest win on video-derived GIFs
Gradients banding after compression	Reduce colours (less aggressively)	Palette reduction is what causes banding
Too long a duration for the format	Trim / drop frames	Size falls close to proportionally with frames removed
Unwanted borders or desktop chrome visible	Crop	Fixes framing only — expect little or no weight change
Still over the cap at acceptable quality	Change format	Animated WebP keeps colour and transparency; MP4 is smallest but has no alpha

Batch Resizing GIFs: Desktop Tools, CLI and Code

When you have fifty emoji to prepare or a product catalogue of animated swatches, single-file resizing stops being practical. Three paths cover almost every batch scenario.

Desktop Batch Tools vs gifsicle vs sharp

Desktop batch path. Light Image Resizer version 7.6.6.178, released 2026-09-09, lists animated GIF among its supported formats and runs on Windows 11, 10 and 8.1. Its v7 feature set includes shell integration for the Windows context menu, pre-defined profiles including file-size targets such as 256 KB, 1 MB and 2 MB, and a batch converter that processes folders and subfolders.

CLI path. gifsicle’s --resize with an explicit resampling method is scriptable, deterministic and preserves frame delays — the right choice for dozens of emoji or asset variants where you want the same operation applied identically every time. Deterministic output matters more than speed once you’re generating a set that has to look consistent.

Code path. The sharp npm package resizes GIF alongside JPEG, PNG, WebP, AVIF and TIFF in Node.js, and at version 0.35.5 more than 10,000 other projects in the npm registry depend on it. That dependency count is a useful proxy for how battle-tested the pipeline is — but when your own code re-encodes, verify frame-delay and palette handling yourself rather than assuming the defaults match your source.

Mobile and one-off batches. Where no desktop or script is available, ImgPlay and Wondershare UniConverter handle batch work; Wondershare’s Discord guide describes using UniConverter’s converter interface to batch resize a set of GIFs to 128×128 in a single pass, with a frame-rate setting available to help stay under a 256 KB emoji ceiling.

Batch tip. Resize all variants from one master file and one preset. Emoji sets and product swatches stay dimensionally consistent only if they all come from the same source at the same settings.

The Extract → Resize → Reassemble Pattern (and Its Risks)

Historically, the only way to resize an animated GIF was to take it apart. A BleepingComputer forum guide from November 2006, with replies running as late as 2022, documents the classic workflow: IrfanView’s “Extract all frames” to split the animation into individual files, batch conversion to resize and reformat them, then unFREEz to reassemble the frames into a new animated GIF with frame delays measured in hundredths of a second — an entry of 1 giving a delay of 1/100 of a second, so 100 equals one second.

That thread is kept here explicitly as a dated reference, not as a current tool recommendation. The 83×83 px avatar figure in it was a 2006-era forum rule, and the tools are two decades old.

The pattern itself is still a useful mental model, and it’s also the failure mode this article’s diagnostic checklist keeps returning to. Extracting and reassembling frames is exactly the route that breaks things: frame order can scramble if you don’t grab the first file when dragging, timing drifts unless you manually tune the frame delay against the original, and — as one 2008 reply in that thread describes — transparency had to be re-selected frame by frame during reassembly, once on extraction and again on save. A true in-place resize avoids all three problems by never taking the animation apart.

Enlarging a GIF: What Upscaling Can and Can’t Do

Upscaling interpolates; it doesn’t invent detail. Expect softness, and expect hard edges and small text to suffer most. ConvertICO states this directly — the process uses proven image-scaling algorithms and never invents detail that wasn’t in your GIF, so faces and text are not changed.

The resampling choice decides how the softness looks. Lanczos, bicubic or bilinear for photographic frames produces a smooth, natural result that reads like a larger video. Nearest neighbour for pixel art preserves sharp square pixels exactly. xBR/hqx sits between them, redrawing blocky edges as clean curves for emoji, stickers, sprites and other small low-resolution graphics.

The practical ceiling. ConvertICO’s own testing calls 2x the sweet spot for most GIFs, and notes that beyond 3x the result looks soft unless the GIF is pixel art. That’s a useful boundary. Past roughly 3x, blur becomes unavoidable for anything that isn’t pixel art, and the better move is to go back to the original video, project file or vector source rather than upscaling the GIF.

One trade worth taking. In ConvertICO’s tests a 2x animated WebP came in under one third of the size of the equivalent 2x GIF. If the platform accepts WebP, upscaling into WebP is usually the better outcome than upscaling into GIF — you get more detail at less weight.

And the rule that never changes. Upscaling never helps a file-size problem. A 2x upscale has 4x the pixels in every frame, so GIF files grow quickly. Never use upscaling as a weight fix.

Choosing a Resampling Method by Content Type

Content type	Method	What it does
Clips, reaction GIFs, photos, gradients	Lanczos / bicubic / bilinear	Smooth and natural, like a larger video
Emoji, stickers, sprites, small low-res GIFs	xBR / hqx	Blocky edges redrawn as clean curves
Real pixel art, retro game graphics	Nearest neighbour	Every pixel becomes a sharp square

For downscaling the same logic applies with a different risk profile: nearest neighbour on photographic footage produces aliasing and shimmer, while smooth interpolation on pixel art or UI screenshots produces blur where you wanted crisp edges.

Resized GIF Looks Wrong? A Diagnostic Checklist

Every ranking page asserts that timing is preserved. None of them say how to prove it or what breaks it. This checklist does.

Symptom → Cause → Fix

Banding or posterised gradients. Colour reduction went too far, or the palette was re-quantised on export. Fix: raise the colour count or re-export from the original source. Photographic and gradient-heavy GIFs show banding quickly, while flat UI screenshots and line art often survive 64 or even 32 colours with no visible difference.

Everything looks soft. You upscaled past the source size, or smooth interpolation was applied to pixel art or UI screenshots. Fix: go back to the original video or source file, or re-export with nearest neighbour.

Edges shimmer or look blocky. Nearest neighbour was used on photographic footage, or the tool rebuilt the animation from extracted screenshots instead of writing scaled frames. Fix: switch to Lanczos or bicubic and use a tool that resizes in place.

File size barely moved. The GIF was already optimised. Fix: the real levers now are colour count and frame rate, not dimensions. ShotEdit notes that if a GIF has already been through a compressor, the easy savings are gone, and pushing lossy higher shrinks it further only at the cost of visible dithering artifacts.

Timing or loop count changed. The pipeline extracted and reassembled frames rather than resizing in place. Fix: compare frame delay values and loop count before and after, then switch tools. This is the specific failure documented in the 2006–2009 forum workflow, where frame delays had to be manually tuned up or down until the new GIF matched the original’s playback speed.

Transparency turned black or white. Either the destination format has no alpha channel — MP4 cannot store transparency, so those areas become white — or the transparent colour index was dropped during processing. Fix: re-export as GIF or animated WebP. The historical version of this failure is documented in the same forum thread, where transparency had to be re-selected manually on every frame during both extraction and saving because the tool converted it to a background colour by default.

The animation is static after uploading. Some platforms render animated avatars and emoji as still images without the right account tier. Wondershare notes that on Discord, animated server emoji and animated profile pictures both require Nitro; without it they appear static. The file itself may be fine.

Conclusion

Resizing a GIF changes pixels and nothing else — which is exactly why it’s the safest first move, and also why it’s capped in how much weight it can remove. Start from your largest source, resize with the aspect-ratio lock on, then check the export against the specific platform cap in the table above — and re-verify that cap against the platform’s own help centre, since tier limits shift and published figures for the same platform already disagree. If you still miss the target, move down the ladder: lossy strength, colour count, frame count. Switch to animated WebP — or MP4 when transparency isn’t needed — before you keep degrading a GIF that has nothing left to give.

FAQ

Does resizing an animated GIF reduce its file size?

Yes, but only through pixel area. Halving width and height removes about 75% of per-frame pixels, so weight drops roughly in proportion at first. It’s not linear forever — heavily optimised GIFs and flat UI recordings barely move. If weight must fall further, reduce colours or frame count rather than shrinking again.

Will resizing a GIF break the playback speed or the loop?

A true resize only rewrites the pixel canvas; the frame delay array and loop metadata are copied unchanged. Speed changes only when a tool extracts and reassembles frames, or stores delays in different units. Verify by comparing frame delay values and loop count before and after export.

Can I enlarge a small GIF without making it blurry?

No — upscaling interpolates; it cannot invent detail that was never captured. Use nearest neighbour for pixel art and UI, smooth interpolation such as Lanczos or bicubic for footage. Around 2x is the practical ceiling for most GIFs; beyond 3x, go back to the original video or source file instead.

How do I resize a GIF to 128×128 for a Discord emoji?

Crop to a square first so the aspect-ratio lock doesn’t distort the subject. Resize to 128×128 px, then check the weight against the 256 KB emoji ceiling. If it’s still over, reduce colours or drop frames — do not shrink below 128×128, since that is a hard requirement.

What’s the difference between resizing, cropping, trimming and compressing a GIF?

Resize changes canvas dimensions; crop changes the visible area; trim cuts frames and duration; compress reduces the colour or lossy budget. Only resize and trim reliably cut file size — cropping often leaves weight almost untouched. Order of use: resize, compress, reduce colours, drop frames.

Are my GIFs uploaded to a server when I use an online resizer?

It depends on the tool’s architecture. Server-side tools upload and return the file; client-side tools run everything in the browser. Client-side tools use WebAssembly builds such as gifsicle, so the file never leaves your device. Test by disconnecting the network mid-resize — local processing still completes.

2026年10月9日

HTML to JPG (html-to-jpg): Complete Conversion Guide

To convert HTML to JPG (html-to-jpg), upload your file or enter a URL into a free online converter such as CoolUtils or Aspose. The tool renders the page and saves it as a JPG snapshot. For batch or automated work, desktop tools or command-line APIs are the better path.

How to Convert HTML to JPG: Step-by-Step

The fastest way is to use a browser-based converter. You upload an HTML file or paste a URL, pick JPG as the output format, and download the rendered image. Most free tools work this way, though file size limits, quality controls, and deletion policies vary.

Using an Online Converter (CoolUtils or Aspose)

CoolUtils and Aspose both offer free online HTML to JPG converters that run right in the browser. No installation is needed.

CoolUtils workflow

Open the CoolUtils online HTML to JPG page.
Upload an HTML, HTM, XHTML, MHTML, or MHT file. The converter accepts files up to 50 MB.
Choose JPG or JPEG in the output format list.
Add optional header or footer text if you need it.
Click Download Converted File to save the JPG snapshot.

According to CoolUtils, the service processes uploads over HTTPS/TLS and automatically deletes both input and output files after conversion. The same page reports that CoolUtils has converted 157,328,305 files totaling 4,801 TB.

Aspose workflow

Open the free Aspose HTML to JPG converter.
Click the file drop area to upload or drag and drop an HTML file. You can also enter a URL.
Click Convert to render the page as a JPG.
Download the result from the link that appears.

Aspose’s free tier is more limited. According to Aspose, free users can process 5 files per day, with a 5 MB total size limit and 2 operations daily. Files are automatically deleted from Aspose servers after 24 hours.

Converting Multiple Files with a Desktop Tool

Online converters get frustrating once you have many HTML files. Desktop software like Total HTML Converter can process entire folders in a single run and keeps everything on your local machine.

The basic desktop workflow is:

Add a folder of HTML, HTM, or MHT files.
Select JPG as the output format.
Set JPEG quality and output dimensions if needed.
Run the batch conversion, then grab the JPG files from the output folder.

CoolUtils says Total HTML Converter can convert hundreds of files in three clicks, works offline, and includes batch conversion, JavaScript toggles, CSS rendering, and JPEG quality control.

HTML to JPG Conversion Methods Compared

There are five practical ways to convert HTML to JPG: online converters, desktop software, command-line tools, APIs, and headless browsers. They differ in privacy, batch capability, file size limits, and how well they fit automation. The right choice depends on how many files you handle, where the files live, and whether the workflow must run unattended.

Online Converters: Free, Fast, and Limited

Online HTML to JPG converters work well for one-off conversions and quick previews. You don’t need to install anything, and they work from any device with a browser.

The main trade-offs are file size limits and privacy. CoolUtils accepts files up to 50 MB and automatically deletes input and output files after conversion, according to CoolUtils. Aspose is stricter: free accounts can handle 5 files, 5 MB total size, and 2 operations per day, with files deleted after 24 hours, according to Aspose.

Online tools also tend to process one file at a time. CoolUtils notes that its online MHTML to JPG converter handles a single file at a time, while batch conversion requires the desktop version. If you only need occasional snapshots or want to test a workflow, an online converter is the easiest place to start.

Desktop Software: Batch Processing and Total HTML Converter

Desktop software removes the file size, privacy, and volume limits that come with online tools. CoolUtils sells Total HTML Converter as the desktop option, with a one-time price of $59.90 for a lifetime license and commercial use.

Key capabilities include:

Batch conversion of HTML, HTM, MHT, and XHTML files
Full CSS rendering with controls to enable or disable JavaScript
Fit-to-page layout for consistent output
Adjustable JPEG quality to balance file size and image fidelity
Output to JPG, PNG, TIFF, PDF, DOC, XLS, and other formats
Watermark, page counter, and date options for output files
Command-line execution and an SDK for integration

Desktop conversion keeps files on your own machine, which matters for sensitive documents. It also removes the 50 MB online file limit: if an HTML file or web archive exceeds 50 MB, the desktop tool can still process it.

Command-Line Tools and APIs for Automation

Automated workflows need a path that runs without manual uploads or clicking through a GUI. The two main options are local command-line converters and cloud APIs.

Total HTML Converter X is CoolUtils’ server-oriented command-line tool. It converts HTML files and live URLs to JPG from cmd.exe or PowerShell using HTMLConverter.exe. A basic batch command looks like this:

HTMLConverter.exe C:\Pages\*.html C:\Output\ -c JPEG -ImgQuality 90 -ImgWidth 1024

This renders every HTML file in the source folder to a JPG with the same base name. Flags such as -ImgQuality, -ImgWidth, and -combine control JPEG quality, output dimensions, and full-page rendering. The tool can run from a .bat file, Windows Task Scheduler, or an ActiveX/COM interface from .NET, PHP, Python, and other environments.

For cloud-based integration, HTML/CSS to Image offers an API that returns PNG, JPG, WebP, or PDF output from HTML/CSS, URLs, or reusable templates. The documentation says the service has generated more than 100 million images and has been in production since 2018. Companies including Uber, HubSpot, Airtable, monday.com, Yamaha, Wayfair, PlanetScale, Expensify, and Coldwell Banker use it.

Headless browsers like Puppeteer or Playwright can also render HTML to JPG, but they bring maintenance and resource overhead. Marek Dvorak, Senior Backend Engineer, puts it this way: “We replaced a Puppeteer-based screenshot pipeline with Total HTML Converter X about a year ago. The Puppeteer setup needed constant babysitting after Chrome updates and consumed ~400 MB of RAM per worker. The CoolUtils command line takes a fraction of the resources, never breaks on updates, and produces consistent JPG output. We render about 40,000 product pages a night without a single failure.”

For compliance snapshots, Janelle Foster, Compliance Manager, describes a scripted approach: “I wrote a small .bat script that pulls a list of URLs from a CSV and runs HTMLConverter.exe against each one, naming the output JPGs with the date. It runs on a Windows Server 2019 box, uses a few percent of CPU, and produces clean, identical-looking screenshots regardless of when the script runs.”

JPEG Quality Settings and Output Control

JPG is a lossy format, so the quality setting controls how much compression gets applied. That directly affects file size and visual fidelity. Lower values create smaller files but can introduce visible artifacts; higher values preserve more detail but produce larger files.

CoolUtils recommends values between 85 and 92 as the sweet spot for web use. In Total HTML Converter X, the -ImgQuality flag accepts any value from 1 to 100. The -ImgWidth and -ImgHeight flags set output dimensions in pixels, and -combine renders the whole document as one tall JPG instead of splitting it into pages.

Desktop Total HTML Converter also exposes JPEG quality as a user-facing option. CoolUtils notes that when converting HTML to JPEG, users can “set the desired quality of the JPEG image to compromise the file size.” Online converters usually offer fixed or limited quality settings, so if you need precise control over output size and sharpness, a desktop or command-line tool is the more reliable route.

Method selection at a glance

Use case	Recommended path
One-off conversion or quick preview	Free online converter such as CoolUtils or Aspose
Batch conversion of local files	Desktop Total HTML Converter
Sensitive or legal documents	Desktop or command-line tool that keeps files local
Automated server pipeline without a browser	Total HTML Converter X command line
Cloud-based integration or template-driven generation	HTML/CSS to Image API
JavaScript-heavy single-page apps	Headless browser or Chrome-rendering API

HTML, HTM, MHTML, MHT: What’s the Difference?

HTML to JPG conversion often involves several related file extensions. The input type affects which tool can handle the file and how the rendered output looks.

HTML vs HTM: Same Language, Different Extension

HTML and HTM are the same markup language. The .htm extension is the older three-character form that came from the DOS 8.3 filename limit, which allowed only three characters after the dot. Modern systems treat .htm and .html files the same way, and browsers open both natively.

CoolUtils describes HTM as “functionally identical to HTML” and notes that the .htm extension is often used for compatibility with legacy systems and older operating systems. When you convert to JPG, choosing HTML or HTM makes no difference to the process or the result.

MHT vs MHTML: One Archive Format, Two Extensions

MHT and MHTML are also the same format, but they are not the same thing as a plain HTML page. MHTML, short for MIME HTML, is defined by RFC 2557. It bundles a complete web page into a single file using MIME multipart encoding, combining HTML markup, CSS, JavaScript, and embedded images.

The only difference between MHT and MHTML is the extension. .mht is the older three-character form used by Internet Explorer, while .mhtml is the longer, modern form used by Edge and some browser extensions. CoolUtils states that both extensions implement RFC 2557 MIME encapsulation and that its converter processes them identically.

Browser Support and Security Implications

MHTML support has shrunk in modern browsers. According to CoolUtils, Chrome removed native MHTML support in 2020 and Firefox removed it in 2021, both citing security vulnerabilities. The worry is that MHTML files can embed active scripts that execute in privileged contexts, creating an attack vector for malicious .mht or .mhtml attachments.

Edge still supports MHT natively, but on many non-Windows systems the files cannot be opened reliably. Converting MHTML or MHT to JPG sidesteps the compatibility issue entirely: you get a static image that requires no browser or archive viewer.

When choosing an input type, use HTML or HTM if you have raw markup or a saved page in plain markup form. Use MHTML or MHT if the file was saved as a web archive from Internet Explorer, Edge, or a browser extension, because those files include embedded resources that an ordinary HTML file may reference externally.

Full-Page vs Viewport Capture: What Gets Captured?

Capture behavior varies by tool and sometimes by input type. Some converters capture only the visible viewport, while others render the full page height from top to bottom.

The difference matters for long pages. A viewport-only capture shows only the part visible without scrolling, so content below the fold can get cut off. A full-page capture stitches the entire document height into a single JPG, which works better for archived articles, complete receipts, and legal notices.

CoolUtils illustrates both behaviors on its own pages. For MHTML and MHT online conversion, CoolUtils says the converter renders the complete document from top to bottom and captures the entire page height as a single JPG. For HTM online conversion, though, CoolUtils says the online converter captures the initial viewport area at a standard desktop viewport width, while the desktop Total HTML Converter offers a full-page rendering mode.

So don’t assume every “HTML to JPG” tool behaves the same way. Before converting a long page, check for a “full-page,” “entire page,” or “combine” option. If the online converter only captures the viewport, use a desktop or command-line tool instead.

Output resolution also depends on the rendering engine and viewport width. CoolUtils notes that its online HTM converter renders at a 1280-pixel viewport width at screen resolution, while the desktop version lets you set custom viewport width and DPI for higher-resolution output.

JavaScript Rendering and Dynamic Web Pages

Many lightweight HTML to JPG converters do not execute JavaScript. If a page loads content dynamically through scripts, that content may be missing from the resulting JPG. CoolUtils states plainly for its online MHTML converter that “dynamic JavaScript is not executed; the page is captured in its loaded state.”

You’ll notice this limitation most with single-page applications and other JavaScript-heavy sites. A renderer that stops after loading the initial HTML will capture placeholders, empty containers, or loading spinners instead of the final rendered content.

For such pages, you have three options:

Use a headless browser like Puppeteer or Playwright, which executes JavaScript and renders the page the way a real browser would.
Use an API that renders in a browser engine. The HTML/CSS to Image API documentation notes that its service renders HTML/CSS in Google Chrome, which means JavaScript can run before capture.
Use a converter with a JavaScript enable/disable option, such as CoolUtils Total HTML Converter, which lets users enable or disable JavaScript and stop slow scripts.

The trade-off is resource and maintenance cost. Headless browsers give the most faithful rendering but require careful sandboxing and can break after browser updates. As Marek Dvorak, Senior Backend Engineer, put it, a Puppeteer-based pipeline “needed constant babysitting after Chrome updates and consumed ~400 MB of RAM per worker,” which led his team to replace it with Total HTML Converter X for 40,000 nightly renders.

Which Tools Execute JavaScript?

Not all converters execute JavaScript, and this isn’t always obvious from a product page. In general:

Lightweight online converters often render the page in its loaded state without running dynamic scripts.
Desktop CoolUtils Total HTML Converter offers an option to enable or disable JavaScript and can stop slow scripts.
The HTML/CSS to Image API renders in Google Chrome, so JavaScript can be part of the capture.
Headless browser automation such as Puppeteer or Playwright executes JavaScript by default.

If your page needs JavaScript to display visible content, check the tool’s rendering engine before relying on the output.

Privacy and Security: What Happens to Your Files?

Online converters require uploading files to a third-party server, which raises privacy considerations no matter what security claims the provider makes.

CoolUtils describes its online conversion as encrypted in transit with HTTPS/TLS and processed in data centers with multi-factor access control. It also says both input and output files are deleted automatically from its servers after conversion, and that files are never read, shared, or repurposed. Aspose states that uploaded files are automatically deleted after 24 hours and that users can delete them immediately.

Even with these policies, online conversion isn’t right for every document. For sensitive legal, compliance, or internal business content, local processing is safer. Desktop software like Total HTML Converter and command-line tools like Total HTML Converter X keep files on the local machine. CoolUtils emphasizes this point in its desktop comparison: files “never leave your machine.”

File size limits also overlap with privacy and workability. CoolUtils online supports files up to 50 MB, while Aspose free users are limited to 5 MB total and 2 operations per day. Files that exceed online limits must be converted locally, which is often the better privacy choice anyway.

Use Cases: Who Needs HTML to JPG Conversion?

HTML to JPG conversion serves several different groups, each with its own needs for volume, privacy, and automation.

Casual users need quick, one-off conversions for sharing or embedding. A free online converter like CoolUtils or Aspose is usually enough, with no installation or signup required.

Compliance and legal teams need archived snapshots with timestamps and consistent rendering. Janelle Foster, Compliance Manager, uses a .bat script to pull URLs from a CSV and run HTMLConverter.exe against each one, naming the output JPGs with the date. The script runs on a local Windows server and produces “clean, identical-looking screenshots regardless of when the script runs.”

E-commerce and development teams often need batch product page rendering or CMS thumbnail generation. The same command-line approach can run nightly over a catalog. Marek Dvorak’s team renders about 40,000 product pages a night with Total HTML Converter X after moving away from Puppeteer.

Students, researchers, and content creators usually want to save web content as images for presentations or documentation. CoolUtils user reviews describe converting market research pages into visual summaries, archiving competitor product pages as a visual timeline for PowerPoint reports, and generating JPG screenshots for legal documents without needing technical viewers.

Across these use cases, the pattern stays the same: online converters handle occasional needs, while batch, legal, and production workflows benefit from desktop or command-line tools that keep files local and run without manual interaction.

Conclusion

HTML to JPG conversion can be as simple as uploading a file to a free online converter, but the right method depends on volume, privacy needs, and automation requirements. For one-off tasks, start with CoolUtils or Aspose and download the rendered snapshot in seconds. For batch jobs, sensitive files, or unattended workflows, choose a desktop tool like Total HTML Converter or a command-line solution like Total HTML Converter X. Matching the tool to the use case helps you avoid file-size limits, privacy exposure, and manual overhead.

FAQ

What is the difference between MHT and MHTML files?

MHT and MHTML are the same web archive format. The only difference is the file extension: .mht is the older three-character form from DOS 8.3 filename limits, while .mhtml is the modern four-character extension. Both implement RFC 2557 MIME encapsulation of HTML, and converters process them identically.

Why can’t Chrome open MHTML or MHT files?

Chrome removed native MHTML support in 2020–2021 because of security vulnerabilities tied to embedded active scripts. Firefox also removed support in 2021. Edge still opens MHT natively. To view the content without a compatible browser, convert the MHTML or MHT file to HTML or JPG.

Will the JPG capture the full page or just the visible viewport area?

It depends on the tool’s settings. Some converters capture only the visible viewport, while others render the full page height. Before converting a long page, check for a “full-page,” “entire page,” or “combine” option, or use a desktop tool with full-page rendering.

Can I batch convert multiple HTML files to JPG in one run?

Yes. Desktop tools like CoolUtils Total HTML Converter support batch processing and can convert entire folders in a single run. Online converters typically handle one file at a time or impose strict daily limits, so they aren’t practical for large batch jobs.

Is online HTML to JPG conversion safe for sensitive or legal documents?

Online tools upload files to third-party servers, which creates privacy risk even when deletion policies exist. For sensitive documents, use a desktop tool or command-line script that keeps files local. CoolUtils and Aspose both publish deletion policies, but sensitive work is safer offline.

What resolution and JPEG quality should I expect from the output?

Output resolution depends on the rendering engine and the page’s dimensions. JPEG quality is often adjustable from 1 to 100; 85–92 is the typical sweet spot for web use. Higher values preserve more detail but create larger files.

2026年10月9日

Invoice Numbers Explained: What They Are & How They Work

Invoice numbers explained: an invoice number is a unique ID assigned to every invoice you issue. It tracks payments, reconciles records, and keeps you audit-ready. Never reuse, skip, or delete one. The US sets no federally mandated format — but IRS recordkeeping rules still apply.

What Is an Invoice Number? (And What It Is Not)

An invoice number is the unique identifier assigned to one specific invoice you issue. It exists so that you, your customer, your bookkeeper, and a tax authority can locate the same transaction without relying on names, dates, or amounts that repeat constantly across a set of books.

In practice, the invoice number is the primary key of your accounts receivable ledger. When a customer’s payment lands with a memo line, the characters in that memo are the only thing that reliably ties the cash to the invoice it settles. Statrys describes the number as the reference that lets you locate and verify any specific transaction.

Every invoice number comes down to three components: a prefix, a sequential number, and optionally a date segment, plus zero-padding to keep the digits consistent. The format itself is yours to choose. Uniqueness is not. As Lev Tretyakov, CEO of Fortador USA, puts it: “We always include an invoice number for all sales we make. The invoice number makes it easy to pull up the transaction and provide accurate assistance to our clients.”

It also helps to know what an invoice number is not. It is not a purchase order number, a quote number, a receipt number, or a credit note number. Each of those is issued by a different party, at a different point in the transaction, for a different purpose.

The short version: numbering is an internal control, not an aesthetic decision.

The Three Parts of Every Invoice Number

A prefix identifies the document type or issuing entity, such as INV for a sales invoice or a branch code in a multi-location business. A date segment places the invoice in the right tax period, which speeds up retrieval during a review. The sequential number is the part that carries the control — it must increase logically and never repeat. Statrys recommends starting at 0001 rather than 1 so the sequence accommodates growth without a mid-year format change.

Where the Concept Came From: From Clay Tablets to Internal Controls

Numbered commercial records are not a modern invention. Around 5,000 years ago, merchants in ancient Mesopotamia used cuneiform script on clay tablets to document trade transactions, including quantities, prices, and the parties involved — and roughly 85% of the cuneiform tablets uncovered at Uruk (3000 BCE) represent economic records, according to Wikipedia’s overview of invoicing. Standardized printed invoice forms emerged with the printing press in the 15th century, and printed invoices with company names and logos were widespread by the 19th century. Digital and automated numbering systems followed in the 20th and 21st centuries.

The One Rule of Invoice Numbering: Never Reuse, Skip, or Delete

If you remember one rule from this guide, make it this one: one number maps to one invoice, once and only once. Every serious source on the topic converges on the same principle, and almost every numbering problem traces back to violating it.

A gap in the sequence is not automatically wrong, but an unexplained gap is a problem. During an audit or an internal review, a missing number raises a specific question — was there an invoice here, and if so, where did it go? Alaan notes that gaps are acceptable only when voided invoices are clearly documented; otherwise the sequence looks incomplete.

Reusing a number creates a different failure. If two transactions carry the same identifier, a payment that arrives against that number cannot be matched to a specific invoice, and the ambiguity persists in the ledger indefinitely.

This is why numbering matters beyond bookkeeping mechanics. A continuous, explainable, traceable sequence is itself evidence that your reported revenue is complete. Statrys points out that gaps or duplicate numbers raise questions about whether all income was reported, which matters more if your books are ever reviewed.

The operational rule that follows is simple: when an invoice is wrong, void it rather than delete it. Keep the number, mark the document void, and note the reason. The number stays in the sequence permanently, even though it will never produce a payment.

There is a second baseline rule that keeps your numbering aligned with reality. Lilach Bullock uses a fixed rule of invoicing within 24 hours of work being completed or a milestone being hit, with no exceptions and a calendar reminder attached. That discipline keeps the sequence order identical to the order in which business actually happened.

Why a Missing Invoice Number Triggers Questions

A missing number forces anyone reviewing your records to search by customer name, amount, or date instead of by a key. That search is slower and less precise, and it can end in a dead stop. Statrys frames the practical cost this way: when a client references a past transaction, the invoice number is what matches it to a record, and without one you rely on memory or manual searching. Multiply that across several clients and the gap becomes genuinely difficult to resolve.

Does the IRS Require Invoice Numbers? What US Rules Actually Say

Here is the answer most US readers are looking for: there is no federal requirement governing invoice number format, prefix, starting value, or continuity. Statrys states plainly that there is no legal requirement in the US to use a specific invoice number format, and no single federal requirement that invoice numbers be strictly sequential — though sequential numbering is standard practice and the default in most accounting software.

But “no mandated format” is not the same as “no recordkeeping requirement.” IRS Publication 334 (Tax Guide for Small Business) sets out the general record-keeping standards that apply to small business income documentation. An invoice number is what lets those records tie to each other — it is the thread that connects a payment to a sale to a reported figure.

Retention is the next layer. Ramp notes that the IRS and most tax authorities require keeping invoices for at least three to seven years depending on jurisdiction and business type, and recommends erring longer since digital storage is inexpensive. The practical implication for numbering is that your sequence needs to remain readable and searchable across that entire window, which is an argument against formats you will not be able to reconstruct in five years.

Why does this matter in a review? A complete sequence requires almost no explanation. Statrys describes the alternative: gaps or duplicate numbers raise questions about whether all income was reported, and this matters more if you are subject to an IRS audit or need to produce invoices on request.

For a US small business that wants a defensible default with no deliberation required, use INV-YYYY-NNNN, start at 0001, and decide explicitly whether the sequence resets annually or runs continuously. The European Commission’s work on the Late Payment Directive has estimated that non-payment and late payment of invoices cause 25% of corporate bankruptcies — a non-US figure included here only to illustrate why collection discipline carries real economic weight, not to draw conclusions about US requirements.

One caveat before moving on: this is general information, not tax advice. Confirm your specific recordkeeping and retention obligations with a tax professional.

Federal vs. State: Where Invoice Number Rules Actually Live

The federal layer governs income recordkeeping. The state layer is where invoice content requirements are more likely to appear, particularly around sales tax. Because sales tax is administered state by state, the required fields on a sales invoice can differ depending on where you operate and where your customer is. If you sell into multiple states, check the requirements for each jurisdiction you have nexus in rather than assuming one template covers everything — and confirm the specifics with your accountant.

How Other Jurisdictions Differ (UAE, UK, India, Philippines)

Other countries make numbering far more prescriptive than the US does, which is useful context for anyone comparing systems or reading international guides. In the UAE, VAT law requires a unique, sequential invoice number on every tax invoice, Alaan explains, and missing or duplicate numbers can trigger audit queries or rejected claims; UAE VAT records must generally be retained for at least five years. In the UK, GOV.UK sets out what an invoice must include for VAT purposes. India’s GST e-invoicing system uses a jurisdiction-specific Invoice Reference Number (IRN), a registration identifier generated for an e-invoice rather than the generic “reference” field on an ordinary bill. In the Philippines, invoicing and tax requirements are administered by the BIR. If you operate across borders, treat numbering as a per-jurisdiction compliance question.

Invoice Number vs. PO Number vs. Reference Number: A Numbering Identity Table

The fastest way to eliminate numbering confusion is to separate each identifier by what it points to, who issues it, and when it comes into existence. The table below maps the identifiers that most often get conflated on a single invoice.

Identifier	What it identifies	Who issues it	When it is generated	Where it usually appears	Can it be the invoice number?
Invoice number	The invoice itself	Seller’s finance or billing system	When the invoice is issued	Invoice header and accounting record	—
PO number	The buyer’s authorized purchase order	Customer’s procurement team	Before supply or billing	Customer PO, cross-referenced on the invoice	No
Reference number	A customer, project, contract, job, or internal record	Varies by party	Varies	Invoice header, labeled separately	No (it is a different object)
Credit note / credit memo number	An adjustment document	Seller’s finance or billing system	When the credit is approved	Credit note, linked to the original invoice	No — use a separate series
Receipt number	Collection of payment	Seller’s finance or billing team	When payment is received	Receipt and collection record	No
Pro forma invoice number	A preliminary estimate or quote	Seller	Before the sale is confirmed	Pro forma document	No — use a separate series
Sales order number	The seller’s accepted order	Sales or order-management team	After order acceptance	Sales order, fulfillment record, invoice	No
Delivery receipt number	Proof or tracking of delivery	Warehouse or logistics team	At dispatch or delivery	Delivery receipt, fulfillment record	No
Payment reference	A remittance or bank transaction	Customer or bank	When payment is sent	Remittance advice, bank record	No — it identifies the payment event

The core distinctions are worth stating plainly. A PO number identifies the buyer’s purchase order, and the buyer — not you — supplies it. An invoice reference number connects the invoice to something else useful, such as a customer account, project, job, contract, or order, and does not necessarily identify the invoice itself. A receipt number confirms collection of payment, not the sale. InvoiceAtlas summarizes the relationship well: an invoice can contain both an invoice number and a PO number because each one connects the transaction to a different document, and they are not duplicates.

If your customer requires a PO number, quote the one they gave you — never substitute an internal reference for it. InvoiceAtlas lists “creating a PO number instead of using the one supplied by the customer” among the most common reference mistakes. If the required PO number is missing, ask the customer’s contact before you send the invoice.

The Same Document, Two Perspectives: AR and AP

One party’s invoice is the other party’s bill. From your side of the transaction, the invoice number identifies a receivable you are owed. From your customer’s side, the same number identifies a payable they must clear. Invoice Data Extraction emphasizes that the supplier-issued invoice number is the durable link back to the source document, and that AP teams should record the full value as printed — including prefixes, suffixes, separators, and leading zeros. Changing INV-00417 to 417 makes duplicate detection and supplier queries less reliable because the recorded value no longer matches the source.

Why Separate Number Series Prevent Downstream Matching Errors

Credit notes, pro forma invoices, and any other adjustment documents should run in their own series with their own prefixes, such as CN- or PF-. Alaan recommends distinct prefixes for invoices and credit notes to prevent sequence contamination and protect audit clarity. Statrys makes the same case for pro forma invoices: because they are not final billing documents, they should not run in the same sequence as your official invoices. Mixing them creates documents that look like duplicates of legitimate invoices, which is a self-inflicted reconciliation problem.

Invoice Numbering Formats Explained: Sequential, Date-Based, Client-Coded, Project-Based

Take a typical format apart once and the rest are variations on the same theme. INV-2026-000127 carries a prefix (INV) that identifies the document type, a year segment (2026) that makes retrieval by tax period straightforward, and a four-digit zero-padded sequence (0127) that supports more than ten thousand invoices per year without changing the number of digits.

Four approaches cover the overwhelming majority of US small businesses. Sequential numbering increments by one and is the default in most accounting software. Date-based numbering embeds the period in the number. Client-coded numbering attaches a customer or account prefix. Project-based numbering attaches an engagement code. None of these is more compliant than another in the US — the choice depends on how you actually search your records.

A comparison from Tofu lays out the trade-offs fairly: sequential numbering produces no gaps and audits easily but carries the risk of overlap during system updates; chronological numbering organizes by date but can produce long numbers; alphanumeric formats encode useful detail but require strict format documentation; and fully custom formats adapt to unusual needs at the cost of complexity.

Approach	Example	Strength	Limitation	Best fit
Sequential	INV-0001, INV-0002	Short, system-friendly, easy to audit	No period or entity context	Single-entity small businesses
Date-based	2026-04-001	Organizes by tax period	Longer references; reset rules must be documented	Businesses that reconcile monthly or by tax year
Client-coded	CLIENT01-001	Supports sorting by account	Requires a controlled client code list	Agencies and consultants with recurring clients
Project-based	PROJ-MKTG-01	Groups invoices by engagement	Codes can become lengthy; watch field-length limits	Project-based and milestone-billed businesses

Sequential Numbering: The Default That Audits Easiest

Sequential numbering is the lowest-friction option and the one most software defaults to. Start at 0001 and increment by one for each invoice. Tofu describes it as well suited to small businesses issuing regular invoices without complex categorization needs, and Statrys recommends starting at 0001 rather than 1 to accommodate growth and present more professionally on client-facing documents. The one weakness to plan around is overlap during system changes or migrations, which is why a documented freeze point matters more than the format itself.

Date-Based Numbering and the Annual Reset Question

Date-based numbering embeds the period directly into the identifier, using formats such as 2026-04-001 or YYYYMMDD-XXX. Tofu notes that if you issue multiple invoices on the same day, a numerical suffix distinguishes them. This approach naturalizes annual resets because the year segment already separates one period from another — but it does not remove the need for a written policy. Statrys cautions that switching reset behavior mid-year without documentation creates confusion, and that either a continuous or an annual-reset sequence is acceptable as long as you pick one and stay with it.

Client-Coded and Project-Based Numbering for Agencies and Multi-Entity Businesses

Client-coded formats attach a customer prefix, so CLIENT01-001 sorts invoices by account without opening individual files. Project-based formats attach an engagement code so all invoices tied to one piece of work can be pulled together. Tofu illustrates the pattern with formats like CL-001 for a client code or PROJ-A-001 for a project identifier. The governance cost is real: prefix lists need controls so two branches or teams cannot claim the same code, and codes must stay short enough to survive portal, spreadsheet, and API field limits. Set the code list centrally before anyone starts issuing.

How to Choose and Set Up Your Invoice Numbering System: A 6-Step Framework

This is a decision path, not a format menu. Answer three questions first: is your business a single entity, multi-branch, multi-client, or project-based? Do you need to retrieve invoices by tax year or accounting period? Who actually assigns numbers, and in what system? Only then pick a format.

Step 1 — Choose your prefix. Use prefixes to distinguish document types: INV for sales invoices, CN for credit notes, PF for pro forma documents. Alaan recommends separate prefixes for invoices and credit notes specifically to avoid sequence contamination. If you operate multiple entities or branches, add a controlled entity code and make sure no two entities share it.

Step 2 — Set the sequence length and zero-padding. Work backward from your maximum expected invoice volume over the next three to five years, then pad to that width. Four digits (0001) covers up to 9,999 invoices per year; five digits covers 99,999. Choosing the width up front avoids the renumbering problem that appears when you outgrow a three-digit sequence in month eight.

Step 3 — Write down your reset policy. Decide whether the sequence resets at the start of the year or runs continuously forever. If it resets, decide how the reset avoids collisions with historical numbers — the common solution is to embed the year in the identifier rather than returning the running number to 0001. Record the policy in an accounting policy document rather than leaving it as tribal knowledge.

Step 4 — Document who assigns numbers and when. The assignment rule that matters most is timing: numbers should be generated when the invoice is issued, not when a draft is created. Alaan explains that assigning numbers at the draft stage leaves gaps behind when drafts are abandoned or edited. Lilach Bullock applies the same principle at the workflow level with her rule of invoicing within 24 hours of completion or milestone, which keeps the sequence order consistent with the order work was actually delivered.

Step 5 — Fix the template. Place the invoice number at the top of the document, next to the issue date and the customer’s details, and keep that position identical across every template version. Statrys notes that most invoicing software and standard templates already place it there by default, which makes it easy for a client to find when paying or asking a question.

Step 6 — Set a check point. Run a monthly reconciliation of your numbering sequence against your issued invoices. Look for duplicates and unexplained gaps while the month is still open. Alaan recommends periodic sequence reviews to confirm there are no unexplained gaps or duplicates, because early detection avoids painful corrections during an audit.

Where the Invoice Number Should Appear on the Document

The invoice number belongs near the top of the invoice, in the header block, close to the issue date and customer information. That placement lets a customer retype it accurately into a payment memo and lets an accounts team find it without scanning the whole document. Whatever template you use, the field label and position should stay consistent — InvoiceAtlas recommends keeping important identifiers together near the main invoice details, with the invoice number shown first, followed by the PO number or other reference, then the invoice date and due date.

How to Write an Annual Reset Policy (and When Not To Reset)

A workable reset policy answers four questions in writing: does the sequence reset, on what date, how does the new series avoid colliding with historical numbers, and where is the policy recorded? Alaan confirms that invoice numbers can restart at the beginning of a new financial year provided the reset is clearly documented and the sequence stays continuous within that year. The safer pattern for most US businesses is to embed the year in the identifier — INV-2026-0001 followed by INV-2027-0001 — rather than returning a bare running number to 0001, which risks two open invoices sharing a short reference. If you have open receivables spanning a year end, resetting is usually more trouble than it is worth; run a continuous sequence instead.

Changing Formats Mid-Year: Freeze Points and Migration Rules

Changing formats mid-year is one of the most reliable ways to break continuity. Alaan warns that format changes mid-period complicate reconciliation and reporting, and Tofu lists inconsistent formats as a direct cause of sorting errors when invoices stop following the established system.

If you need to change, set a freeze point — typically the start of a fiscal year or a new quarter — reconcile the old sequence completely before the switch, and never return to the old format. Historical invoices are not renumbered. HashMicro’s numbering guidance describes the migration sequence in practical terms: freeze the cutover point, reconcile open receivables, communicate the new format, and obtain approval for any exceptions before the first invoice leaves the new system.

Why Invoice Number Discipline Is a Cash Flow Lever, Not Just a Compliance Task

Numbering and invoice timing are two halves of the same cash flow mechanism. The invoice date starts the clock on your payment terms. The number is what lets the incoming payment be matched to the right invoice once it arrives. Together they determine how long your cash sits outside your account.

That second half is where the friction shows up. Payment memos are frequently incomplete or mistyped, and a unique, continuous invoice number is the one key that can still reconcile an imperfect payment. Statrys describes the consequence of not having one: tracking outstanding invoices and reconciling payments relies on memory or manual searching, and managing multiple invoices across several clients makes any gap in the sequence significantly harder to resolve.

The scale of the problem is well documented on the receivables side. According to the 2025 Intuit QuickBooks Small Business Late Payments Report, 56% of small businesses have late payments owed to them. On the more severe end, the European Commission’s Late Payment Directive work has estimated that non-payment and late payment of invoices cause 25% of corporate bankruptcies — a figure from a non-US regulatory context, included here as background on why collection timing has outsized consequences.

There is a first-person illustration of how much timing alone can cost. Lilach Bullock describes sending invoices a week or more after delivering work while on standard 30-day terms, which stretched what looked like a 30-day cycle to roughly 45 days from delivery to payment. Across five or six clients on rolling retainers, that produced a persistent cash gap of a few thousand pounds every month. The fix was a single rule — invoice within 24 hours of completion or milestone, no exceptions — which pulled almost two weeks back into the cash cycle without changing a single client’s payment terms.

The same discipline shows up on the service side. As Lev Tretyakov of Fortador USA notes, the invoice number “makes it easy to pull up the transaction and provide accurate assistance to our clients.” Both benefits come from the same underlying habit: a numbering system that stays in step with real business events.

Common Invoice Numbering Mistakes and How to Repair a Broken Sequence

Most numbering failures fall into a short list: duplicate numbers, skipped numbers, deleted or edited numbers, voided invoices with no explanation, format changes mid-year, and multiple systems each generating their own numbers. Each has a documented repair path.

For context on how common invoicing errors are in general, a second-hand compilation cited by Tofu states that 39% of invoices contain errors. That figure is republished without a marked statistical year and should be treated as indicative only — it is useful for illustrating that invoicing errors are not rare, not for drawing conclusions about numbering accuracy.

I Found a Duplicate Invoice Number — Now What?

Do not delete either document. First confirm which of the two is the original issued invoice. Keep the original number as it stands. Void the duplicate and reissue the correct invoice under a new number. Note the original number and the reason for the void directly on the record, and retain the correction in your files. Statrys describes the equivalent workflow for a numbering mistake: mark the incorrect invoice as void and note the reason, then issue a corrected invoice with a new number, which keeps records complete and prevents gaps that could raise questions during an accounting review. Never reassign the duplicated number to a new document.

How to Document a Skipped Number Without Reusing It

A skipped number needs a written explanation, not a backfill. Do not issue a new invoice into the empty slot, and do not reuse the number for anything else. Instead, record the exception in a log with five fields: the number, the date, the reason (voided, never issued, or system error), the action taken, and the reviewer. HashMicro’s control framework treats missing exception logs as a distinct failure mode, because without one, finance cannot distinguish a legitimate exception from an error. Alaan makes the same point about voided invoices: leave voided numbers in the sequence with a documented reason so reviewers can see the run is complete.

Migrating from Excel to Invoicing Software Without Breaking the Sequence

Migration is where duplicate numbers most often appear, because a legacy spreadsheet and a new system can both be generating numbers for a short window. Fix it in order. Freeze the cutover point and stop issuing from the old system. Reconcile the old sequence completely, including any open receivables. Configure the new system’s starting number and format. Preserve legacy references in a separate field rather than overwriting them. Then test duplicate prevention before going live. HashMicro notes that legacy references should be preserved in a separate field when the new system key differs, because legacy and new sequences can collide after cutover.

Two related traps deserve specific mention. Credit notes and pro forma invoices must run in their own series; if they share the invoice sequence, they will look like duplicates of legitimate invoices. And when multiple systems generate numbers, designate one system of record for number generation and let the others sync rather than create.

Invoice Automation: How Software Removes Human Numbering Errors

Automation addresses numbering at the level where the errors actually originate. A system that generates the number at issuance enforces uniqueness by construction, blocks a duplicate before it posts, and holds the sequence intact when several people issue invoices concurrently. Alaan describes automated systems as ensuring numbers are assigned once and only once, sequences remain intact across users and departments, and the same number flows into accounting tools and reports.

Three error classes largely disappear with automation: manual duplicates, format drift between team members, and skipped or never-issued numbers caused by draft-stage assignment. What remains is a different set of exceptions — data migration leftovers, multiple entities sharing a sequence they should not share, and manual overrides of system-generated numbers. Those are governance problems, not software problems, and they need policy rather than a feature.

Buyer-side cost is worth understanding too, because it explains why large customers care how you number. HashMicro cites Ardent Partners’ 2025 global benchmark, which reported a 22% average invoice exception rate — a figure that shows how often invoices fail to process cleanly on the receiving end. Inconsistent or ambiguous references contribute to that failure rate, which is why enterprise AP teams often require a valid PO number and a clear invoice number on every document they receive.

What to Look For in Invoicing Software

Evaluate numbering capability with specific questions rather than a feature tour. Does the system support custom prefixes and zero-padded sequences without manual formatting? Can it share a sequence across entities, or isolate one per entity where that is required? Does it detect duplicates at entry rather than at month-end? Does it keep an audit history showing who issued, voided, or changed a number, and does it prevent edits to issued references? HashMicro recommends restricting manual edits, deletion, backdating, and sequence changes through role-based access, and testing the full workflow — including a duplicate attempt the system must reject and a cancelled document whose reference never returns — before committing to a vendor.

Invoice Numbers in E-Invoicing Standards

If you ever exchange invoices in structured electronic formats, the invoice number becomes a data field with defined transmission rules rather than a visible label. Wikipedia’s overview of invoicing notes that the United Nations EDIFACT “INVOIC” standard includes standard codes for transmitting header information common to the entire invoice as well as line-item detail, and can also carry credit and debit memos. XML-based standards include the UN/CEFACT cross industry invoice and OASIS Universal Business Language (UBL), which is used in public-sector implementations including Denmark, with further implementations in Italy, Spain, and the Netherlands, and the PEPPOL framework for European public procurement. ISDOC was developed in the Czech Republic as a universal electronic invoice format.

There is also a process model that bypasses invoices altogether for some transactions. Evaluated Receipt Settlement (ERS) pays from the packing slip rather than a separate invoice: the supplier ships against an advance shipping notice, and the purchaser confirms the corresponding purchase order or contract, verifies the goods, and pays. If your customers use ERS, your internal numbering still governs your own records even though no invoice number travels with the payment. Because public documentation on e-invoicing standards ages quickly, verify current requirements against the standards body or your customer’s specification rather than a secondary summary.

Conclusion

Invoice numbering is an internal control, not a formatting preference. A unique, continuous, explainable sequence makes payment matching faster and audit questions rarer — and in the US, the real constraint is not the format but whether your records can be traced and reconciled. Three actions to take today: write down your format and reset policy, including prefixes and zero-padding; pin the invoice number to the top of every template and commit to invoicing within 24 hours of completion or milestone; then run one continuity reconciliation and log every existing duplicate and gap in an exception file.

FAQ

Does the IRS require invoice numbers on invoices?

No. There is no federal requirement for invoice number format or continuity. However, IRS Publication 334 requires businesses to keep records supporting income and expenses, and invoice numbers are what let those records cross-reference each other. Some state sales tax rules impose additional invoice field requirements, so check the rules for states where you operate.

Can I reset my invoice numbers at the start of a new tax year?

Yes, but there is no single mandatory US rule — what matters is writing the policy down and applying it consistently. Alaan notes that a reset is acceptable when clearly documented and the sequence stays continuous within the year. A safer pattern for most businesses is embedding the year in the identifier, such as INV-2026-0001, rather than returning the running number to 0001 and risking collisions with open historical invoices.

What should I do if I skip or duplicate an invoice number?

For a duplicate, confirm the original, void the duplicate, and reissue under a new number while recording the original number and reason. For a skip, do not backfill or reuse the number — document it in an exception log with the date, reason, and reviewer. Both cases need a written trail so the sequence stays explainable during any review.

Is an invoice number the same as a receipt number?

No. An invoice number identifies a request for payment; a receipt number identifies confirmation that payment was received. Statrys describes them as different document types serving different purposes that both belong in your recordkeeping system. Many systems generate the receipt number separately, so store them in distinct fields.

Can two different suppliers use the same invoice number?

Yes. Uniqueness is only required within a single issuing entity — Alaan confirms invoice numbers only need to be unique per supplier, not across different businesses. That is why buyer systems typically identify an invoice using the combination of supplier plus invoice number, and why enterprise customers require your business identifier and invoice number to appear together.

Where should the invoice number appear on an invoice?

Place it in the header, near the invoice date and the customer’s details. Statrys notes that most invoicing software and standard templates position it there by default, which makes it easy for a client to locate when paying. Keep the field label and position identical across every template version you use.

2026年10月9日

Discord File Size Limit 2026: 20MB Cap and What to Do

The Discord File Size Limit 2026 is 20MB per file for free accounts, up from 10MB in August 2026. Nitro Basic raises that to 50MB, Nitro goes up to 500MB, and a server Boost lifts the cap for the whole server. The 8MB, 10MB, and 25MB numbers still floating around online are out of date.

Discord File Size Limit in 2026: 20MB Free, 500MB With Nitro

The current limit for a free, non-Nitro account is 20MB per file, effective August 2026 and up from the old 10MB ceiling. As the Discord File Attachments FAQ puts it, “for non-Nitro subscribers, the maximum file upload is 20MB.” The same page also says Discord is “currently experimenting with various file upload size limits,” which is why a small number of accounts may see a different number in the app.

Paid tiers raise the ceiling. Nitro Basic takes the per-file limit to 50MB, and full Nitro supports uploads up to 500MB, according to the same Discord support docs. Server Boost works separately: it changes the limit for the entire server, not just for you. Level 2 is 50MB and Level 3 is 100MB — Section 3 covers that in detail.

Discord’s developer docs back up the consumer-facing numbers. The Discord API reference says “the file upload size limit applies to each file in a request. The default limit is 20 MiB for all users, but may be higher for users depending on their Nitro status or by the server’s Boost Tier.” When two independently maintained sources land on the same figure, that’s the strongest signal 20MB is the real baseline in 2026.

All current Discord upload limits at a glance

Access type	Per-file upload limit	Scope
Free account	20MB	Account-wide (servers + DMs)
Nitro Basic	50MB	Account-wide
Nitro	Up to 500MB	Account-wide
Server Boost Level 2	50MB	Server-wide (all members)
Server Boost Level 3	100MB	Server-wide (all members)

The 20MB baseline is the number to compare against when a file fails. Anything below it should upload without a paid plan, as long as the channel and server also allow the file.

Per file, not per message — and why that matters

The limit applies per file, not per message. A free user can attach several files in one message, and each one is measured on its own. Three 15MB files in a single message add up to 45MB, but they pass fine because no single file crosses the 20MB line.

That distinction changes what “too big” means. When an upload is rejected, it’s the single attachment that’s over the cap, not the whole message. The same rule applies at every tier: Nitro and Server Boost raise the per-file ceiling, not the number of files you can send.

Why Do 8MB, 10MB, and 25MB Still Appear? A Discord Upload Limit Timeline

Figure	Effective period	Status in 2026
~8MB	Legacy era	Out of date
10MB	Until August 2026	Replaced by 20MB
20MB	From August 2026	Current free limit
25MB	Undated	Conflicts with Discord’s documentation

The confusion is mostly a search-decay problem. Older articles that were never updated still rank for upload-limit searches, and some of them quote a stale number in the headline while linking to Discord’s current Help Center page in the body. The mismatch looks authoritative even though the two halves of the page disagree.

Discord’s own content isn’t perfectly consistent either. The GitHub issue tracking the Cove video downloader bot points out that “Support caps article still documents server Level 0/1 as 10MB in places; the live free-user/default behaviour is 20MB as of August 2026.” When the documentation and the live app disagree, treat the in-app number as the tiebreaker — that’s what the server actually enforces.

The 10MB legacy limit (pre-August 2026) and why it is still everywhere

Until August 2026, 10MB was the correct free upload limit, and a huge number of tutorials, forum posts, and bot configs were built around it. Discord’s August 2026 update confirmation says the free upload limit was raised from 10MB to 20MB in a platform-wide change, not a limited test. According to TechJournal, Discord confirmed the increase in its September 8, 2026 patch notes and said the larger limit applies across all platforms, including desktop and mobile.

That same release batch shipped client performance updates, which helps date the change window. Discord reported its mobile apps moved to React Native 0.86 with roughly 5% lower average CPU use and about 5% faster average launch times, and its desktop app moved to Electron 42 with roughly 17% better p50 CPU usage alongside an approximately 4% p50 memory regression. These are company-reported averages, not independent benchmarks.

Where the 8MB and 25MB figures came from

The 8MB figure lives on in GIF-specific content. A 2026 reference page from BulkPicTools still lists 8MB as the free limit and even claims 10MB is the outdated number. That page is wrong about the current free chat limit — Discord’s Help Center says 20MB — but the 8MB number does describe something else: it’s the profile avatar and banner asset limit for Nitro users, not the per-file chat upload limit.

The 25MB figure shows up in third-party guides such as transientfiles.com, which states “on a free account that’s 25 MB.” That number doesn’t match Discord’s Help Center or its API reference and shouldn’t be treated as fact. Put it plainly: any page citing 8MB or 10MB as the current free chat limit in 2026 is wrong, and 25MB should be treated as an unverified claim.

Which Limit Applies to You? Account Tier × Server Boost × Channel

Four layers decide the real ceiling: your account tier, the server’s Boost level, the surface you’re posting on (server channel vs. DM), and where Discord checks the size. The effective limit is usually the higher of your personal tier and the server’s Boost tier, but only inside that server.

Server Boost lifts the cap for everyone in that server. At Level 2, the server-wide limit is 50MB; at Level 3, it’s 100MB. A free account posting inside a Level 3 server can upload a 100MB file that would fail in an unboosted server. A Nitro or Nitro Basic subscriber keeps their personal ceiling in every server and in DMs — Boost doesn’t lower a paid tier.

DMs follow the personal account limit only. Server Boost never applies there, so a Level 3 server’s 100MB ceiling disappears the moment you open a direct message. A Nitro subscriber in an unboosted server still keeps their 50MB or 500MB personal limit in that server and in DMs.

Below the Boost tiers, the documentation is genuinely inconsistent. Some official passages still list 10MB at Boost Level 0/1 while the live free default is 20MB. Bot developers working against these tiers often leave about 1MB of headroom to absorb rounding and encoding overhead — a pattern visible in the Cove bot repository, which uses 49.0MB for Level 2 and 99.0MB for Level 3, and proposed moving the Level 0/1 margin from 9.5MB to 19.5MB after the August 2026 change.

To confirm your own current limit rather than trust any published table — including this one — check the original file size in your OS, compare it against your account tier and the server’s Boost level, and try the upload. If it fails, Discord’s error message tells you whether size is the blocker. The Help Center’s core point is simple: the number shown in the app is the one that counts.

Scenario: free account inside a Level 3 boosted server

You have no Nitro. The server you’re posting in has reached Boost Level 3. In that server’s text channels, you can upload files up to 100MB — the same ceiling a paying subscriber would get from Boost alone. Your personal free limit of 20MB doesn’t restrict you here, because the server-wide Boost value is higher.

The moment you leave that server, you’re back to 20MB. Post the same 100MB file in a DM or in an unboosted server and it will be rejected. The boosted ceiling belongs to the server, not your account.

Scenario: Nitro subscriber in a server with no Boost (and in DMs)

You pay for Nitro. You’re posting in a server with no Boost at all. Your personal ceiling of up to 500MB applies, and the server’s lack of Boost doesn’t reduce it. The same goes for DMs, where only your account tier matters.

If that unboosted server also has channel-level restrictions — a missing “Attach Files” permission, for example — the size limit is irrelevant; the file won’t go through at any size. Size and permission are separate gates.

How to confirm your own current limit in the app

Discord’s experiment with upload limits means a small number of accounts don’t match the published table. The practical check is to look at the file size before you send. On Windows, right-click the file and open Properties; on macOS, right-click and choose Get Info; on mobile, open the file manager and tap the file for details. Compare that number against your account tier and the server’s Boost level, then send.

If a file under your expected limit still fails, the fastest diagnostic is to try the same file on desktop, where the pre-compression check has always applied. Discord’s Help Center notes mobile was moved to the same pre-compression check desktop uses, so any device-specific gap should now be narrow.

How to Send Files Larger Than 20MB on Discord

Start with a three-branch triage before touching any tool. Branch A: the original file is at or under 20MB — just send it, and if it still fails, jump to Section 5. Branch B: the file is bigger and quality matters — share a link. Branch C: the file must live inside Discord as an attachment — compress to roughly 19MB.

Check the ORIGINAL file size first. Since August 2026, Discord compares against the pre-compression size on both desktop and mobile, so guessing wastes attempts. The starting threshold for a free account is the 20MB per-file baseline documented in the Discord File Attachments FAQ.

Branch 1: Compress to about 19MB (FFmpeg, HandBrake, 7-Zip)

A cross-platform comparison checked on September 9, 2026 by SqueezeVid lists Discord’s practical target as 19MB at 720p — about 1MB of headroom under the documented free limit. That margin absorbs container overhead and rounding so the file doesn’t land on the wrong side of the cap.

The most reproducible route is FFmpeg. A test published by DisCoRibute on September 13, 2026, run with FFmpeg 7.0.2-static on Ubuntu 24.04.4 LTS, took a generated 20-second 1920 × 1080 clip and produced this result:

ffmpeg -i "input.mp4" -vf "scale=1280:-2" -c:v libx264 -b:v 1800k -c:a aac -b:a 96k -movflags +faststart "discord-video.mp4"

The source file was 85,414,421 bytes. The exported discord-video.mp4 was 4,801,249 bytes, about 4.8MB, and it uploaded to Discord successfully. The parameters are a starting point, not a guarantee — a longer or more complex clip will land higher, and if the output is still too large the same source recommends lowering the bitrate and exporting again with a new filename. For the framework behind those settings — codec choice, CRF, bitrate budgeting — see how to compress a video without losing quality.

HandBrake and 7-Zip cover the rest of the toolchain: HandBrake for GUI-driven video re-encoding — iPhone and QuickTime footage has its own quirks, covered in how to compress a MOV file — and 7-Zip or WinRAR for archive compression when the file is a folder of documents or assets rather than media.

When quality matters, a link beats compression because the file never passes through Discord’s encoders — the core of bypassing Discord’s file limit without Nitro. Choose a host by criteria, not brand. According to TinyGrab, the questions that separate the options are:

Does the recipient need an account? WeTransfer allows downloads without an account for basic transfers; Google Drive and Dropbox often require a Google or Dropbox account depending on the sender’s settings.
Does the link expire? WeTransfer keeps files available for a limited time on the free tier, while Google Drive, Dropbox, Mega and MediaFire generally keep them available as long as the account is maintained.
Does the host re-encode? A cloud drive link preserves the original file; a video streaming preview does not. If the recipient needs the original, permission to download the file itself must be granted.
Can you set permissions or an expiry? Public links work for anyone with the URL; private links require authentication and are the better default for sensitive files.
Does it work on mobile? Recipients on phones should be able to open the link in a browser without installing a client.

Mega adds end-to-end encryption and generous free storage, which makes it the fit when the file is sensitive. WeTransfer is the fit when the transfer is one-off and simplicity matters more than permanence.

Branch 3: Split archives, and when not to bother

Splitting into volumes with 7-Zip or WinRAR works, but only if each part independently stays under the applicable limit. Every part has to be downloaded before the archive can be rejoined, and on Windows joining RAR parts requires WinRAR — so the recipient needs the right tool. Treat this as a last resort for document bundles and folders, not for a single large video. For simple bundling without the split-and-rejoin dance, zipping a file on Mac and Windows is the lighter option.

One path worth knowing that skips all three branches: the Tenor and GIPHY GIF picker inside Discord shares a link rather than uploading a file. Per BulkPicTools, using the built-in picker means “the file never touches your upload limit.” Pasting a Tenor or GIPHY URL directly into chat behaves the same way.

Why Does Discord Reject a File That’s Already Under 20MB?

The root cause is timing. Since August 2026, Discord checks file size before compression on both desktop and mobile. The Help Center explains it directly: previously on mobile, Discord would try to compress the file first and only tell you afterward if it was still too large, so the limit you saw didn’t always match what applied. Discord then “moved mobile to the same pre-compression check that desktop has always used, so the number you see is the number that counts.”

That single change explains the most common false alarm. A video that shows as 20MB in your file manager, and that would have compressed well below the limit, is measured at its original size — and if the original is 20.4MB, it fails.

Order the checks so you don’t waste time. First, confirm the original file size in your OS. Second, note the platform — Android versus desktop or iOS. Third, confirm whether you’re looking at a pre-compression or post-compression number. Fourth, check format restrictions. Fifth, consider content scanning. Sixth, check account and server permissions.

Two non-size blockers are worth naming because they look like size failures. Only .jpg, .jpeg, .png, .webp and .gif render as inline embedded media per the Discord API reference; other formats still upload but appear as downloadable files, which users often read as a failure. Separately, Discord’s NSFW content scanning can block an upload if the system flags the image, and a missing “Attach Files” permission or role restriction in a server channel blocks it regardless of size.

The pre-compression check, explained

The pre-compression check evaluates the source file, not the compressed output. Discord’s stated reason is consistency: evaluating the original prevents the same file from appearing eligible on one device and failing on another after processing. The practical consequence is that you can’t rely on Discord’s own compression to rescue an oversized upload.

This is why “compress it in the app” isn’t a workaround. If the original is over your limit, the upload is rejected before any processing happens. The fix is to reduce the source file on your device — export a lower-bitrate copy, resize the image, or split the archive — and then upload the smaller file.

Android: why lowering your recording quality does not fix it

Discord’s own documentation warns that Android users may hit the limit more often because source files tend to be larger. The Android free upload limit is not different from desktop or iOS; the difference is that an Android device can produce or retain a larger original file before Discord ever sees it.

Lowering in-app camera or recording quality settings doesn’t change a file that already exists on the device. If a clip was recorded at 4K and saved at 180MB, adjusting the recording setting forward has no effect on that file. The correct fix is to re-export or compress the existing file, or to check the original size in the file manager before trying again.

The fastest diagnostic is to try the same file on desktop. If it uploads there and fails on Android, the issue is the source file on the phone, not the account. Contact Discord Support only after the file succeeds on one platform and fails on another with the same account.

Video and GIF Limits: Re-Encoding, 10-Minute Splits, and Auto-Play

Discord re-encodes uploaded video to its own preferred settings. That’s why gameplay footage or a screen recording can look softer after it’s sent — the file you uploaded is not the file the recipient downloads or streams. The transientfiles.com guide describes this as re-encoding “to Discord’s preferred settings — which is why gameplay footage looks softer after you send it.”

Long clips are also split. The same source notes videos longer than about ten minutes are split into ten-minute clips by the app itself, commonly around the 10-minute mark. If continuous playback matters, uploading a link to a host that preserves the original is the better route.

GIFs are subject to the same per-file upload limit as any other file. The much-quoted ~256KB figure is a playback threshold, not an upload rule. Per BulkPicTools, GIFs above a certain size will not animate automatically and sit as static images until someone clicks them, and “the number that comes up most often is 256KB, though Discord hasn’t made this an official published spec.”

Server asset limits are a different dimension entirely. Animated emoji cap at 256KB, animated stickers at 512KB and must be exactly 320×320px, and both limits are stricter than the chat upload limit and do not change with Nitro status. The 8MB profile avatar and banner figure is a Nitro profile asset limit, not a chat upload limit, and animated avatars are a Nitro-only feature regardless of file size.

What’s official and what’s community-tested

Label every number by provenance. The 20MB free limit, the 50MB Nitro Basic ceiling, the up-to-500MB Nitro ceiling, the 20 MiB API default, and the Boost Level 2 (50MB) and Level 3 (100MB) values are documented by Discord. The ~256KB GIF auto-play threshold is community-tested and explicitly not an official published spec. The 19MB / 720p practical target is a cross-platform recommendation, not a Discord rule. The 8MB legacy figure and the 25MB claim are neither official nor current for chat uploads — 8MB matches the Nitro profile avatar and banner asset limit, and 25MB conflicts with Discord’s own documentation.

Running a Server or a Bot? How Boost Tiers Change the Limit for Everyone

Boost level is a server-wide setting. Raising it raises the ceiling for every member of that server, including free accounts and accounts that belong to other servers — the higher value applies inside that server regardless of who is posting. A free member in a Level 3 server uses the 100MB ceiling. A Nitro subscriber in a Level 2 server uses whichever of their personal limit and 50MB is higher.

Bot developers should read the permitted attachment size from the API rather than hardcoding a number. The Discord API reference states the default limit is 20 MiB per file and notes it “may be higher for users depending on their Nitro status or by the server’s Boost Tier,” and the attachment_size_limit value provided in interaction payloads is calculated as the maximum of these values. Reading that field keeps a bot in sync with the account and server context it’s acting in.

Common practice is to leave about 1MB of headroom under each tier cap so rounding and encoding overhead don’t cause rejections. The Cove bot repository uses 49.0MB for Level 2 and 99.0MB for Level 3 for exactly this reason, and after the August 2026 change it proposed moving Level 0/1 from 9.5MB to 19.5MB. Carrying forward 8MB or 10MB constants causes unnecessary re-compression and visible quality loss for media that would now upload fine.

The bot’s limit and the client’s displayed limit must agree. If a bot enforces 10MB while the client allows 20MB, users get contradictory error messages and lose quality to compression that was never needed.

Conclusion

In 2026, the number is 20MB per file for free accounts — not 8MB, not 25MB — and the cap you actually get depends on your account tier, the server’s Boost level, and where you’re posting. Check the real size of the original file first. If it’s under 20MB and still rejected, test it on desktop. If it’s over, pick your branch deliberately: compress to about 19MB with the FFmpeg recipe, or share a link when quality matters. Confirm your own ceiling in the app rather than trusting any published table, including this one.

FAQ

Is Discord’s free upload limit 8MB, 10MB, 20MB or 25MB in 2026?

20MB per file is current, in force since August 2026. 10MB was correct until that change, and 8MB is older still. The 25MB figure conflicts with Discord’s own Help Center and API reference and should not be treated as fact.

Is the Discord upload limit per file or per message?

Per file — you can attach several files in one message. Three 15MB files are fine on a free account because no single file exceeds 20MB. Server Boost and Nitro raise the per-file ceiling, not the number of files you can send.

Do I need Nitro to send a file larger than 20MB?

Not necessarily. Nitro Basic gives 50MB and Nitro up to 500MB, but a Boosted server raises the limit for free members too — Level 2 is 50MB and Level 3 is 100MB. Compressing to around 19MB or sharing a cloud link also works without paying.

Does Discord compress files I send, and does it check size before or after compression?

Since August 2026, both desktop and mobile check the pre-compression size, so the original file size decides whether the upload is accepted. Discord may still re-encode video after it is accepted, which is why quality can drop even when the upload succeeds.

Why does Discord reject my upload on Android when the file is under 20MB?

The pre-compression check means Android’s larger exported source files are what get measured. Lowering in-app recording quality does not change the size of a file already on the device. Test the same file on desktop to confirm it is a source-file issue rather than an account limit.

Can a server Boost level raise the upload limit for everyone in that server?

Yes — Boost is server-wide: Level 2 lifts the ceiling to 50MB and Level 3 to 100MB. The higher of your account tier and the server’s Boost level applies inside that server. Boost does nothing in DMs.

Can I upload a larger file if I’m in a boosted server but my own account is free?

Yes, inside that server you can use the boosted ceiling, up to 100MB at Level 3. The moment you post in a DM or in an unboosted server, your free 20MB limit applies again. Nitro subscribers keep their own higher limit everywhere.

Do GIFs have a different size limit than other files on Discord?

Uploads use the same per-file limit as any other attachment — 20MB on a free account. The ~256KB figure people quote is a community-tested auto-play threshold, not an official upload rule. Sending a GIF through the Tenor or GIPHY picker shares a link and skips the upload limit entirely.}

2026年10月9日

Days Until Christmas 2026: Exact Count and How to Check Today’s Number

As of September 25, 2026, there are 91 days until Christmas 2026. Christmas Day falls on Friday, December 25, 2026 — that’s 13 weeks, 13 weekends, or 91 sleeps. A live countdown timer shows 90 full days plus about 12 hours. And because every countdown number on this page drops by one each day, use this table to read the count for the date you’re actually on:

If today is…	Days until Christmas 2026
October 9, 2026	77
November 1, 2026	54
November 26 (Thanksgiving)	29
November 29 (Advent Sunday)	26
December 1, 2026	24
December 15, 2026	10
December 24 (Christmas Eve)	1
December 25 (Christmas Day)	0

The convention throughout: today is not counted, Christmas Day is. Every row follows the same rule the article unpacks below — and the spreadsheet formula near the end gives you a number that updates itself daily.

How Many Days Until Christmas 2026? The Direct Answer

There are 91 days until Christmas Day 2026, which falls on Friday, December 25, 2026. Days.To breaks the same wait down as 13 weeks, 13 weekends, 64 weekdays, 2,184 hours, and 91 sleeps.

Quick answer: 91 days, 13 weeks, 91 sleeps

How Many Days Until Christmas reports 91 calendar days from September 25, 2026 in New York, or exactly 13 weeks and 0 days. Inch Calculator also lands on 91 days, or 13 weeks and 2,184 hours, and notes that Christmas is a U.S. federal holiday.

As-of date and timezone: America/New_York

Every number on this page is tied to an as-of date of September 25, 2026 and to the America/New_York timezone. That phrase — “as of September 25, 2026 in New York, there are 91 calendar days” — matters because the count drops by one each day. Inch Calculator shows 95 days on September 21, 94 on September 22, 93 on September 23, 92 on September 24, and 90 on September 26. If you read this on a later date, expect a smaller number.

Why Your Live Christmas Countdown Timer Shows 90 Days, Not 91

The 90-versus-91 gap isn’t an error. It comes from two different ways of measuring the same stretch of time: calendar date labels and complete 24-hour periods.

A calendar-day count leaves out today, September 25, and includes December 25. That gives you 91 days. A live timer measures full 24-hour blocks left until midnight starts Christmas Day, so it shows 90 full days plus a partial day. Days.To displays both on one page — a timer reading 90d:12h:10m:26s next to a written answer of “91 days until Christmas Day.” That side-by-side display is the most common source of confusion. How Many Days Until Christmas explains the same split in its FAQ: one display counts New York date changes, the other counts complete 24-hour cycles, and a partial day can push the full-day figure down by one.

Calendar days vs. 24-hour periods: the rule to remember

Use calendar days when you want a date-to-date answer: September 25 to December 25 is 91 days. Use the live timer when you want the exact time remaining: roughly 90 days and 12 hours from the moment the timer was read. Neither is wrong. They answer slightly different questions.

Which number should you quote?

Quote 91 days for conversations, planning, and headlines tied to a date. Quote “90 days plus about 12 hours” only when someone wants precision down to the current hour. If a friend insists the real answer is 90, they are reading a live timer, not a calendar.

91 Days in Other Units: 13 Weeks, 13 Weekends, and 91 Sleeps

The same 91 days can be broken into a few useful units. Each one suits a different planning purpose.

Weeks, weekends, and sleeps

How Many Days Until Christmas gives 13 weeks and 0 days, 13 weekends, and 91 sleeps. Days.To agrees on 13 weeks, 13 weekends, and 91 sleeps, and adds 2,184 hours. Your Christmas Countdown explains the sleep convention plainly: add 1 to the number of days shown, and that is how many sleeps remain. Sleeps are the version most families and classrooms use, because each night marks one visible step toward Christmas morning.

Hours and minutes (if you really want precision)

Two sources report slightly different hour totals — 2,184 from Days.To and Inch Calculator, versus 2,177 from How Many Days Until Christmas — because each page measures from the moment it was loaded. Treat any hour or minute figure as a snapshot, not a fixed fact. Re-check it if you need the number for a specific moment.

Shopping days and business days

How Many Days Until Christmas counts 90 potential shopping dates after September 25 through Christmas Eve, December 24, 2026 — noting that this does not check store hours, holidays, stock, or delivery cut-offs. One older convention is worth retiring: skipping Sundays as shopping days predates online carts and Sunday trading, as FailFlow points out. Workday counts get their own section below, because sources disagree.

Christmas Day 2026 Falls on Friday: What It Means for Christmas Eve and Boxing Day

Christmas Day 2026 is Friday, December 25, 2026. That puts Christmas Eve on Thursday, December 24, and Boxing Day on Saturday, December 26.

The 2026 Christmas week at a glance

How Many Days Until Christmas lists the surrounding U.S.-relevant dates for 2026: Thanksgiving on Thursday, November 26; the First Sunday of Advent on November 29; Christmas Eve on Thursday, December 24; Christmas Day on Friday, December 25; Boxing Day on Saturday, December 26; New Year’s Eve on Thursday, December 31; and New Year’s Day on Friday, January 1, 2027.

Boxing Day is not a U.S. federal holiday. WatchIsUp notes that December 26 is a holiday in the UK, Ireland, Canada, Australia, and New Zealand, but in the United States it is an ordinary working day, so many businesses stay open. Christmas Day itself is a U.S. federal holiday, according to Inch Calculator.

Is Christmas 2026 a long weekend?

Whether a Friday Christmas creates a long weekend depends on your employer, your state, and applicable holiday rules. How Many Days Until Christmas states it directly: Thursday and Monday are not automatically days off, so check school calendars, opening hours, and travel availability before booking anything.

One pattern to remember: the same source’s Christmas Day-of-Week History table shows 2024 on Wednesday, 2025 on Thursday, 2026 on Friday, 2027 on Saturday, 2028 on Monday, 2029 on Tuesday, and 2030 on Wednesday.

Weekdays and Business Days Until Christmas 2026

Workday counts for the same 91-day window differ by one depending on the source and the counting convention.

Days.To reports 64 weekdays (Monday to Friday) until Christmas Day, not counting the day itself. Inch Calculator reports 65 weekdays and warns that its figure does not account for holidays falling on weekdays, such as Thanksgiving. FailFlow also lands on 65 business days counting weekdays only, with public holidays not subtracted.

How weekdays are counted

The typical method counts Monday-through-Friday dates between today and the target date. September 25, 2026 falls on a Friday, so the count includes every remaining weekday in September, all of October and November, and the December weekdays before Christmas — except for the holiday adjustments each source chooses to apply or ignore.

Why 64 vs 65 matters

Christmas Day is a U.S. federal holiday, so a business-day calculation can legitimately exclude December 25 as a non-working day. That is why the totals sit one apart. For office planning, commutes, and deadline-setting, use 64 weekdays as the conservative figure and subtract any additional company holidays you observe.

Advent 2026 and US Thanksgiving: Your Next Countdown Anchors

Ninety-one days is too long to hold in your head. Two late-November anchors — Thanksgiving and Advent — break the wait into manageable phases.

Thanksgiving 2026 countdown

US Thanksgiving is Thursday, November 26, 2026. How Many Days Until Christmas places it in the 2026 key-dates list as the unofficial start of the American Christmas season. Days.To listed 62 days to Thanksgiving when its Christmas page was captured on September 25, 2026. That gives you roughly two months from the as-of date to plan menus, travel, and early decorating.

Advent 2026: November 29 vs December 1

In Western Christian calendars, Advent 2026 begins on Sunday, November 29, 2026, according to How Many Days Until Christmas, which also records Advent at 65 days out from September 25. WatchIsUp adds the seasonal range: Advent counts the four Sundays before Christmas, so it starts between November 27 and December 3 each year. A chocolate Advent calendar is a different convention entirely — those always begin on December 1. Keep the two separate when you plan, or you will be four days off.

From Countdown to Action: A US Christmas Planning Timeline

With 91 days and roughly 90 potential shopping dates through December 24, the practical move is to split the countdown at U.S. holiday anchors rather than treat it as one long block.

November: Thanksgiving and Advent kickoff

Thanksgiving on November 26 and Advent on November 29 sit four days apart in 2026, so late November is your natural planning checkpoint. Use the days before them to agree on budget, recipients, and likely guests, and to verify long-lead items — custom orders, travel bookings, and shared family dates — because those decisions need the most lead time. How Many Days Until Christmas frames this as a flexible sequence: set limits, check long-lead plans, choose details, confirm final arrangements, and leave room for changes.

December: shopping, shipping, and Christmas Eve

December holds the compressed work. Days.To counts 64 weekdays for planning, and Christmas Eve lands on Thursday, December 24 — one day before the holiday. FailFlow notes that Christmas Eve comes 90 days after September 25, 2026, and that all 91 days count as shopping days under modern retail conditions. Work backward from December 24 for gifts that must ship, and treat the last week as buffer rather than primary shopping time.

What to verify instead of copying dated shipping deadlines

Carrier deadlines are the least reusable content in Christmas planning. Countdown Christmas publishes Royal Mail and Parcelforce posting dates tied to UK addresses and a specific year — a useful illustration of why any deadline table you find may be both out of date and in the wrong country. Instead, check the official carrier page for the year and destination you are shipping to, and confirm processing times, stock, delivery estimates, and return terms before paying. Finally, remember that countdown answers expire daily: always match a number to its as-of statement.

How to Calculate Days Until Christmas 2026 Yourself (Google Sheets / Excel)

If you want a number you control, a spreadsheet handles it in one cell — and the same calendar-day conventions (exclude today, include the target) are exactly the rules covered in how many days between two dates, applied to a fixed destination.

Basic days formula

Type =DATE(2026,12,25)-TODAY() into any cell and press Enter. Inch Calculator documents this formula and a variant for a fixed start date in another cell: =DATE(2026,12,25)-A1. The result updates automatically as TODAY() moves forward.

Weeks and business days formulas

For whole weeks, use =INT((DATE(2026,12,25)-TODAY())/7), the method Inch Calculator lists for weeks until Christmas. For workdays, the standard =NETWORKDAYS(TODAY(), DATE(2026,12,25)) counts Monday through Friday but does not subtract holidays, so add a holiday range or edit the result manually — Thanksgiving and Christmas Day both fall on weekdays in 2026. Set the spreadsheet timezone to America/New_York so your count matches the countdown pages referenced here.

Classroom Countdown Ideas: 91 Sleeps and 64 Weekdays

Countdown numbers work well in classrooms because they are visual, finite, and easy to update each morning.

Paper chain and daily countdown

How Many Days Until Christmas describes the classic approach: make one paper chain loop for every remaining sleep, then remove one after waking each day. With 91 sleeps as of September 25, 2026, a full chain is long enough to fill a wall and visibly shrink. Teachers tracking working days only can use the 64-weekday figure instead, which fits a classroom calendar better than calendar days.

Advent calendar vs. Advent Sunday in the classroom

Advent 2026 begins Sunday, November 29, and a numbered December calendar starts December 1 — four days apart. Label which one your class is following, or the count will look wrong to students. Optional add-ons mentioned by countdown sites include watching one Christmas film a week from mid-November and using an Elf on the Shelf routine, which Your Christmas Countdown links to as a family and classroom activity.

2026 Christmas News and Entertainment Updates

Updated September 2026. The items below are the entertainment stories countdown trackers were covering in late September 2026. They are listed with their reporting dates rather than treated as confirmed schedules.

Your Christmas Countdown reported that Mariah Carey is bringing her Christmas show to the UK in 2026, that Tyler Perry’s The Gospel of Christmas is coming to Netflix in 2026, and that its Christmas Movies 2026 roundup of confirmed films was updated in September 2026. Treat announcements as announcements: verify release dates on the platform or ticketing page before planning around them, and expect titles without a firm date to shift.

Conclusion

As of September 25, 2026, there are 91 days until Christmas 2026, and Christmas Day lands on Friday, December 25, 2026. Pick one anchor rather than tracking all 91 days: US Thanksgiving on November 26, Advent on November 29, or a December shopping milestone that forces early decisions. Use the live countdown for daily motivation, but check any number against its as-of date and timezone, because every countdown answer on this page drops by one tomorrow.

FAQ

Does the Christmas countdown include today?

No. The standard calendar-day count leaves out today and includes Christmas Day. That is why the answer as of September 25, 2026 is 91 days. How Many Days Until Christmas shows the same convention in action, displaying 1 on Christmas Eve and 0 throughout Christmas Day. Including today would give 92, but most countdowns do not count the current day.

Why does the live timer show 90 days when the calendar says 91 days?

The calendar answer counts date labels: September 25 to December 25 spans 91 days. A live timer measures full 24-hour periods, so it shows 90 full days plus about 12 hours. Days.To displays both a 90d:12h timer and the written answer “91 days” on the same page, which is exactly where the confusion starts.

How many sleeps until Christmas 2026 for kids?

There are 91 sleeps until Christmas 2026 as of September 25, 2026, according to both Days.To and How Many Days Until Christmas. Sleeps are a family-friendly count because each night marks one step closer to Christmas morning. Naps do not reduce the total, and the number drops by one each morning after waking.

When does Advent start in 2026?

In Western Christian calendars, Advent 2026 begins on Sunday, November 29, 2026, per How Many Days Until Christmas. An Advent calendar with numbered doors usually starts December 1, a different convention. WatchIsUp confirms the range: Advent begins between November 27 and December 3 depending on the year.

What day of the week is Christmas 2026, and does that create a long weekend?

Christmas Day 2026 is Friday, December 25, 2026, with Christmas Eve on Thursday and Boxing Day on Saturday. Whether that becomes a long weekend depends on your employer, state, and federal holiday rules. How Many Days Until Christmas warns that Thursday and Monday are not automatically days off.

How many shopping days or business days are left until Christmas 2026?

How Many Days Until Christmas counts about 90 potential shopping dates through December 24, 2026. Business-day totals vary: Days.To reports 64 weekdays, while Inch Calculator reports 65, a difference driven by whether Christmas Day is excluded as a federal holiday.

Is Christmas always on December 25?

For most Western Christian traditions, Christmas is observed on December 25, and 2026 is no exception. The date is not universal: How Many Days Until Christmas notes that churches using the Julian calendar currently observe Christmas on January 7 in the Gregorian calendar. Other traditions follow their own calendar practices for gift-giving and observance.

When is Christmas 2027?

Christmas Day 2027 falls on Saturday, December 25, 2027, one day later in the week than 2026. Days.To listed it as 456 days away when its Christmas page was captured on September 25, 2026. That 456 is simply the 91 days to Christmas 2026 plus the 365 days of 2027 — and because 365 is 52 weeks plus 1 day, the weekday slides one slot each year.

2026年10月9日

Unicode Escape Sequence: \uXXXX Syntax, Surrogates & Fixes

A Unicode escape sequence is an ASCII-only way to write a Unicode character — usually \uXXXX, where XXXX is the character’s hexadecimal code point. Java and JSON require four hex digits, JavaScript adds \u{...} and \p{...}, HTML uses &#xhhhh;, and URLs percent-encode UTF-8 bytes as %XX.

What Is a Unicode Escape Sequence?

A Unicode escape sequence is a notation that writes any Unicode character using only ASCII characters, so the text survives tools and protocols that never learned about anything beyond ASCII. The most common form is \uXXXX, where XXXX is four hexadecimal digits giving either a code point or, in UTF-16-based languages, the 16-bit code unit at that value.

Unicode itself identifies characters with U+ notation: U+0000 is NUL, U+001B is the escape character, U+FFFF is the last code point in the Basic Multilingual Plane, and U+1F600 is 😀. Specifications, documentation, and error messages all use the U+XXXX form to name a character; \uXXXX is how that same character gets written inside source code, string literals, and wire formats.

The scale behind that notation is large. According to Wikipedia’s list of Unicode characters, Unicode 18.0 assigns 310,351 characters with code points across 175 modern and historical scripts, plus multiple symbol sets. No single page can enumerate them, which is exactly why a compact escape notation exists.

Unicode Escape vs. Property Escape vs. HTML Character Reference vs. Percent-Encoding

Four notations look alike but solve different problems. \uXXXX — and JavaScript’s \u{...} — represent a code point or UTF-16 code unit inside a string or source file. \p{...} and \P{...} are Unicode property escapes, which per MDN match a set of characters by Unicode property and are only supported in Unicode-aware mode; outside that mode, \p is merely an identity escape for the letter p.

HTML and XML numeric character references get resolved by the markup parser, not by a string parser. Wikipedia describes the format as &#nnnn; for decimal code points, &#xhhhh; for hexadecimal, and &name; for a predefined entity name. URL percent-encoding is a fourth system again: %XX encodes bytes, not characters.

Here’s the one-line rule: read the prefix. \u means a code-point escape, \p means a property class, &#x means a markup character reference, and % means percent-encoded bytes.

Why Escapes Exist: ASCII-Only Representation

The purpose is stated directly in the Java Language Specification (Java SE 7, §3.3): Unicode escapes let a program include any Unicode character “using only ASCII characters.” ASCII-only is the lowest common denominator that compilers, JSON parsers, log pipelines, configuration files, and older protocols all handle.

With 310,351 assigned characters in Unicode 18.0, no toolchain can be expected to carry every character natively. A toolchain can reasonably be expected to carry backslash, u, and hexadecimal digits — so those characters become the transport layer for everything else.

How to Write \uXXXX in Java, JavaScript, JSON, HTML and URLs

No reference page on the search results lays the same character side by side across all five contexts. The table below does that, using é (U+00E9, one code unit) and 😀 (U+1F600, a surrogate pair) as the test characters.

Context	Form	é (U+00E9)	😀 (U+1F600)
Java source / string literal	`\uXXXX`, exactly four hex digits	`"\u00e9"`	`"\uD83D\uDE00"`
JavaScript string	`\uXXXX` for BMP, `\u{...}` for any code point	`"\u00e9"` or `"\u{e9}"`	`"\u{1F600}"`
JavaScript regex	`\p{...}` / `\P{...}` — a property class, not a character	`\p{Letter}`	not applicable
JSON string	`\uXXXX` only	`"\u00e9"`	`"\uD83D\uDE00"`
HTML / XML	`&#xhhhh;`, `&#nnnn;`, `&name;`	`é` or `é`	`😀`
URL	`%XX` octets of UTF-8	one or more octets	four octets (surrogate pair)

Java Source: Escapes Are Processed Before Tokenization

JLS §3.3 defines three lexical translation steps applied in order: Unicode escapes are translated first, line terminators are recognized second, and input elements and tokens are formed third. Because escape processing happens before tokenization, the character an escape produces does not participate in further escapes.

The canonical example: the raw input \u005cu005a yields the six characters \ u 0 0 5 a, not the character Z. The escaped backslash is not reinterpreted as the start of a second escape. Two related rules matter in practice: the u marker may be repeated (\uuuu0041 is legal), and a backslash can only begin an escape when the number of contiguous backslashes immediately before it is even.

JavaScript: \uXXXX, \u{…} and \p{…}

\uXXXX in JavaScript covers the Basic Multilingual Plane; \u{...} is used for code points above U+FFFF. \p{...} and \P{...} are a different construct — Unicode property escapes for regular expressions, with \P building a complement class instead of a single character.

According to MDN, property escapes have been Baseline widely available across browsers since July 2020, and with the v flag \p can match finite-length strings, which is useful for emoji sequences made of multiple code points.

JSON: What JSON.stringify Actually Emits

JSON string escaping uses \uXXXX. The failure mode is not in the format but in the serializer: JSON.stringify emits \u0000 and lone surrogates verbatim and performs no validation of whether a downstream store will accept them.

That is the step where a bad value escapes memory and reaches the database. The serializer’s job ends at producing valid JSON text; whether the JSON text is storable in a particular column is a separate question that the serializer never asks.

HTML &#xhhhh; and URL %XX Are Not \uXXXX

An HTML numeric character reference is read by the markup parser. So é and é are the same character to a browser, but mean nothing to a JavaScript string parser. If you need to go the other way and resolve a page of &#xhhhh; and &name; references back to text, an HTML entity decoder does that at the markup layer. Named entity references cover only a limited, predefined set of names.

URLs take a different route entirely. MDN’s encodeURI() reference explains that the function replaces characters with one, two, three, or four escape sequences representing the UTF-8 encoding of the character, and that four sequences appear only for characters composed of two surrogate units. Because percent-encoding works on bytes rather than code points, one character can become several %XX groups.

Why Do Emoji Need Two \uXXXX Escapes? Surrogate Pairs Explained

A character inside the Basic Multilingual Plane corresponds to one escape. A character above U+FFFF is represented in UTF-16 as a pair of 16-bit code units, so it takes two consecutive escapes to write one character.

JLS §3.3 states this as a rule rather than an edge case: representing supplementary characters requires two consecutive Unicode escapes. The string literal rules in the same specification make the split explicit — one escape sequence for characters in the range U+0000 to U+FFFF, two escape sequences for the UTF-16 surrogate code units of characters in the range U+010000 to U+10FFFF.

The worked example is 😀. Its code point is U+1F600, and in UTF-16 it is written as \uD83D\uDE00 — a high surrogate from U+D800–U+DBFF followed by a low surrogate from U+DC00–U+DFFF. That’s two escapes for one character, which is why length counts, substring operations, and truncation behave in ways that surprise anyone who assumes one escape equals one character.

BMP vs. Supplementary Characters: One Escape or Two

The dividing line is U+FFFF. Characters from U+0000 to U+FFFF need one \uXXXX escape, because four hexadecimal digits cannot express anything larger. Characters from U+010000 to U+10FFFF need two, because each \uXXXX escape writes exactly one 16-bit code unit.

This also explains why Java character literals cannot hold supplementary characters at all — the specification limits a char literal to values from \u0000 to \uffff, so a supplementary character must be written as a surrogate pair in a char sequence or as an integer, depending on the API.

What a Lone Surrogate Actually Is

A lone, or unpaired, surrogate is one half of an emoji with the other half missing. It is not a character. It shows up when a surrogate pair is truncated, or when arbitrary bytes are decoded as UTF-16.

Lone surrogates fail loudly in several places. MDN notes that encodeURI() throws a URIError when it encounters a surrogate that is not part of a high-low pair, and that String.prototype.toWellFormed() can replace lone surrogates with the Unicode replacement character U+FFFD, while String.prototype.isWellFormed() can check for them. PostgreSQL jsonb rejects them outright, as the next section shows.

One consequence deserves emphasis: paired surrogates must be preserved exactly. A sanitizer that treats every high or low surrogate code unit as an illegal character will delete legitimate emoji, which is a hidden bug in a lot of cleanup logic.

How to Fix “unsupported Unicode escape sequence” (SQLSTATE 22P05)

This error is the reason many people arrive at this topic at all. Roughly half the results surfaced for this query are issue reports opened in the September 11–24, 2026 window that exist because of this specific database error, yet none of the reference pages explain it. If you are holding a stack trace, the definition of an escape sequence is not the useful part — the fix is.

What PostgreSQL Rejects and What It Accepts

The boundary is narrow, and it is worth proving with a runnable statement. In standard-conforming SQL strings, \\ is a literal backslash, so the JSON text below contains a real \uXXXX escape:

select '{"t":"a\\u0000b"}'::jsonb;        -- ERROR: unsupported Unicode escape sequence (22P05)
select '{"t":"a\\ud800b"}'::jsonb;        -- ERROR: unsupported Unicode escape sequence (22P05)
select '{"t":"\\ud83d\\ude00"}'::jsonb;   -- OK
select '{"t":"\\u0001"}'::jsonb;          -- OK

According to GitHub issue #1493 in THU-MAIC/OpenMAIC, PostgreSQL rejects exactly two escape families that serializers emit verbatim: \u0000 (a NUL byte) and lone UTF-16 surrogates such as \ud800. It accepts \u0001, \u007f, and other control characters, and it accepts paired surrogates such as \ud83d\ude00. The distinction is NUL and unpaired surrogates — not control characters in general, and not escaping as such.

Sanitizing Before the Write, Not After Serialization

The fix that issue #1493 converged on is to sanitize string values while serializing, using the replacer argument of JSON.stringify, so the check runs against the in-memory string rather than against already-serialized text. The replacer walks code units: drop 0x0000; when a high surrogate is followed by a low surrogate, keep the pair and skip past it; drop any remaining lone high or lone low surrogate.

The issue records a second, harder-won lesson: do not run a regular expression over the serialized JSON. A regex cannot distinguish a real lone-surrogate escape from the literal characters \ and ud800 appearing in content, and stripping those produces invalid JSON — the issue’s author reports hitting SQLSTATE 22P02 that way, which is a worse failure than the original error.

Two variants of the same fix are noted in the issue. \u0000 can be replaced with U+FFFD instead of dropped if you’d rather keep a placeholder. And the sanitizer must not touch paired surrogate code units, since stripping them destroys valid emoji.

Binary content needs separate handling. According to GitHub issue #403 in LabsConnected/litlabs-website, an .ico or .jpeg asset read as UTF-8 produced lone surrogates, and the whole deployment failed at the file-write step with unsupported Unicode escape sequence. The issue’s suggested direction is to base64-encode binary assets before the write, reject or sanitize non-text-safe content with a per-file error naming the offending path, and add a regression test. For URL paths, MDN’s recommendation is to pass the string through String.prototype.toWellFormed() — or check isWellFormed() first — before calling encodeURI().

The impact justifies the guard rails. Issue #1493 reports that a single NUL character or lone surrogate aborts an entire session run, destroying 3–4 already-generated lessons per occurrence during batch course generation on PostgreSQL 16.11 with Node v24.18.1. Issue #403 reports that one bad binary file fails an entire deployment because there is no per-file guard or fallback. Per-record and per-file checks with a fallback path limit the blast radius.

When a Failed Record Cannot Even Be Logged as Failed

There is a second-order failure mode that few write-ups mention: the system cannot even record that it failed. According to GitHub issue #901 in readur/readur, the null-byte guardrail added in PR #308 / v2.6.0 does not cover every database write path. On the current latest image, PDFs whose OCR text contains \u0000 still fail, and the follow-up attempt to insert into failed_documents was itself rejected — this time by a check_failure_reason constraint.

The result is a document that is permanently stuck. It cannot be ingested, it cannot be logged as failed, and it therefore never appears in the “Failed OCR” UI, so the retry function cannot pick it up. A prior database inspection in the same issue showed 105 rows stuck in pending in ocr_queue, and manually resetting statuses did not help because ingestion fails before a queue entry is created.

The issue’s expected behavior is a reasonable template for any pipeline that stores escape-bearing text: sanitize before any database write, including the failure-logging path; fall back to a generic failure reason instead of raising a constraint violation that masks the original error; and always persist failed records so they remain visible and retryable.

Reading Escapes You Didn’t Write: Literal \u00e9 Text vs. Mojibake

Not everyone arriving at this topic is writing escapes. Some are staring at a literal \u00e9 in a dialog box, or at cafÃ© in a string, and trying to work out whether the escape was never interpreted or interpreted one time too many. Pasting the escape into a Unicode escape decoder confirms what the four hex digits name; what follows is about why the surrounding text went wrong.

Symptom	Likely cause	One-line check
`\u00e9` appears as literal text to a human	The layer rendering it does not interpret `\u`, or the text was escaped for one context and injected into another, such as JavaScript-escaped text placed into an HTML attribute	Does the layer that displays the string know how to interpret a `\u` prefix?
`cafÃ©` instead of `café`	A string that was already decoded got decoded again with the wrong encoding — typically UTF-8 bytes read as latin-1	Compare the character count against the expected value and look for `Ã` or `Â` where accented letters belong
The escape resolves correctly but the non-ASCII text beside it is destroyed	A codec that expects bytes was applied to a `str`	Check what input type the codec expects before calling it

Two recent reports map cleanly onto the first and third rows. According to GitHub issue #17357 in mautic/mautic, a confirmation cancel button label stored through a data attribute renders as literal \uXXXX text in the Russian locale instead of the localized word; the proposed fix is to render the label with HTML-attribute escaping rather than JavaScript escaping, and to read it in JavaScript instead of passing HTML content through a data attribute.

According to GitHub issue #53 in colinta/SublimeStringEncode, a Unicode Escape command built on codecs.decode(text, 'unicode-escape') corrupts every non-ASCII character in the selection, because the codec takes bytes and the plugin’s str is first encoded to UTF-8 and then decoded as latin-1. Selecting café produces cafÃ©, and selecting \u00e9 é produces é Ã© — the escape decodes correctly while the literal é next to it is destroyed.

The reusable rule is to identify what the current layer expects before changing anything. Confirm whether that layer consumes escape text or already-decoded characters, then fix that layer, rather than re-encoding repeatedly at the wrong level.

Which Edition and Version Should You Cite?

Version context across the current search results is thin and in places self-contradictory. The only Java source surfaced for this topic is the Java SE 7 specification rather than the current edition. Wikipedia’s character list states that Unicode 18.0 assigns 310,351 characters, yet elsewhere on the same page cites Unicode 17.0 for the number of characters classified as Latin script — two different versions in one article. Only the MDN pages display current revision dates, both last modified on September 2, 2026.

The practical rule is to name the version on every citation rather than treating any edition as “the current rule.” Where a rule has been stable across editions — such as the processing order of Unicode escapes and the non-recursive expansion rule, both of which come from JLS §3.3 and are stated in the same terms in the Java SE 7 edition cited here — say so explicitly and attribute the source. That way a reader is not misled by an older edition’s wording.

Character counts must be version-bound for the same reason. The figure of 310,351 assigned characters across 175 scripts is a Unicode 18.0 figure and changes with each version of the standard, so it should not be mixed with counts drawn from an earlier release.

Conclusion

A Unicode escape sequence is just a notation for writing any Unicode character in ASCII — but the notation takes different forms in Java, JavaScript, JSON, HTML, and URLs, and the edge cases are where it breaks. Surrogate pairs and U+0000 are the two boundaries that cause most real incidents. Start by identifying which of the four escape families you are holding, using the cross-context table above. Before any database write, sanitize at the in-memory string level with a JSON.stringify replacer: drop \u0000, keep paired surrogates intact, and base64-encode binary assets. When you see literal \uXXXX text or cafÃ©, determine whether the string was never interpreted or interpreted twice, then fix that specific layer.

FAQ

What is the difference between a Unicode escape sequence, an HTML numeric character reference (`&#xhhhh;`), and URL percent-encoding (`%XX`)?

They encode different things. \uXXXX encodes a code point, or a UTF-16 code unit in languages that use UTF-16. &#xhhhh; is a markup character reference resolved by an HTML or XML parser. %XX encodes UTF-8 bytes, so one non-ASCII character can become several %XX groups. Reading the prefix — \u, \p, &#x, or % — tells you which system must interpret the text.

Why do emoji and other supplementary characters need two `\uXXXX` escapes instead of one?

One \uXXXX escape writes a single UTF-16 code unit, and its four hexadecimal digits cannot express anything above U+FFFF. Characters above that boundary are stored in UTF-16 as a high surrogate plus a low surrogate, so they need two consecutive escapes. U+1F600 is written \uD83D\uDE00: two escapes, one character. Delete either half and you have a lone surrogate.

Why does PostgreSQL reject my data with “unsupported Unicode escape sequence”?

The error carries SQLSTATE 22P05, and it triggers on a payload containing \u0000 (NUL) or a lone surrogate such as \ud800. It does not reject all control characters — \u0001 and \u007f write successfully, and a paired surrogate like \ud83d\ude00 is accepted. The usual source is JSON.stringify emitting those escapes verbatim, so sanitize the in-memory string before the write. To see which escapes a failing payload actually contains, run it through an online Unicode decoder first.

How many hexadecimal digits does `\uXXXX` require, and can extra `u` characters be added?

The standard form is \u followed by exactly four hexadecimal digits. Java permits a repeated u marker, so \uuuu0041 is legal, but the digit count is unchanged. Expansion is not recursive: \u005cu005a produces the six literal characters \ u 0 0 5 a, not Z. The \u{...} form belongs to JavaScript’s supplementary-character syntax, not to an extended \uXXXX.

What is the escape character U+001B, and why is this notation called “escape”?

U+001B is ASCII 27 in decimal, Unicode U+001B, and the Ctrl+[ combination — the character the Esc key produces. Wikipedia’s article on the Esc key attributes the idea of putting that functionality into a character-encoding convention, and calling it escape, to Bob Bemer’s contributions to ASCII around 1960. Terminals used it to begin control sequences; \u001b today is that control character itself.

2026年10月5日

Blog

Cron Expression Syntax: Fields, Operators, and Dialects

What Is Cron Expression Syntax?

Anatomy of 0 9 * * 1-5: every token, decoded

The Five Fields: Minute, Hour, Day of Month, Month, Day of Week

Value ranges, name aliases, and the values each field rejects

Field-order mistakes that silently shift your schedule

The Four Operators: * , – / and How They Combine

Steps are not intervals: why */7 is not “every 7 minutes”

Which Cron Syntax Are You Actually Writing? A Dialect Matrix

Is Sunday 0, 1, or 7 in your dialect?

The six-field, seconds-first family: Quartz, Spring, Azure NCRONTAB

What changed recently: OCPS 1.0 (2025) and 2026 vendor documentation

Common Cron Schedules You Can Copy — and What They Actually Run

Shorthand macros: @daily, @hourly, @weekly, @reboot — and where they don’t work

The Day-of-Month OR Day-of-Week Rule (The Most Expensive Misreading)

Schedules Five Fields Cannot Express: First Monday, Last Day, Friday the 13th

L, W, # and ? on Quartz, AWS and Cloudflare

The Linux workaround: escaped % and a date test inside the command

Does Cron Use UTC or Local Time? Time Zones and DST

The skipped hour and the repeated hour

Mitigations: shift to 03:00, guard with a lock, or pin to UTC

crontab -e: Where the Expression Actually Lives

System crontab’s extra user column

Before You Deploy: A Sanity-Check Checklist

Verify per platform: runnable commands

Stop overlapping runs and spread load: flock, Jenkins H, RandomizedDelaySec

Kubernetes CronJobs and GitHub Actions: Same Syntax, Different Rules?

Kubernetes CronJob: UTC default, concurrencyPolicy, startingDeadlineSeconds

GitHub Actions: best-effort timing, UTC only, disabled after 60 days

Managed Schedulers Beyond Linux: Vercel, AWS, and systemd Timers

Vercel Cron Jobs: no name aliases, one day field, always UTC

AWS EventBridge and Systems Manager: six fields, ?, Sunday = 1

systemd timers: OnCalendar, Persistent, RandomizedDelaySec

Conclusion

FAQ

Is Sunday 0 or 7 in a cron expression?

Does cron use UTC or the local time zone?

Why does my cron job run more often than I expected?

Can cron run a job every 30 seconds?

What does */5 * * * * mean, and what does 5 mean instead?

How do I schedule “the first Monday of the month” or “the last day of the month”?

Rep Counter: How Automatic Counting Works (2026 Guide)

What Is a Rep Counter—and What Is It Actually Counting?

“Rep Counter” vs. Three Other Meanings of “Rep”

How Does Automatic Rep Counting Actually Work? Camera, Sensor, and Manual Compared

Camera-Based Pose Tracking: Landmarks, Joint Angles, and a State Machine

Sensor and IMU Counting: Why It Works on Barbells and Fails on Cables

Manual Counting: Still the Only 100% Reliable Option (and Why That’s Fine)

What Counts as a Full Rep? Depth, Lockout, and the Partial Rep Problem

Hard Rules First, Classifier Second: A Layered Form Gate

Why You Can’t Judge Depth From a Front-Facing 2D Pose

Which Exercises Can a Rep Counter Handle—and Which Ones Break It?

Exercise × Equipment Decision Table

Why Cable and Machine Sets Are the Hardest Case

Does a Rep Counter Record or Upload Your Workout Video?

Sets and Reps: How a Rep Counter Fits Into Your Training Log

Rep Counter Apps Compared: RepCount, Push Up Counter, and PushForm

How to Choose: Four Questions That Decide It for You

Turning Rep Counts Into Progress: e1RM and RPE Baselines

How to Build Your Own Rep Counter (Developer Path)

Conclusion

FAQ

Do I need a wearable or smart mat to use a rep counter?

Does a rep counter count partial reps, and what happens to them?

Is my workout video recorded or uploaded to a server?

Can a rep counter track cable, machine or leg-press sets?

Is a rep counter app free, or do I need a subscription?

What does “rep counter” mean when it refers to Congress?

Luau Formatter: What It Changes and How to Use It

What a Luau Formatter Actually Changes (and What It Never Touches)

The two whitespace categories behind every Luau formatting rule

FormattingAnnotator: marking the whitespace a formatter is allowed to edit

How to Run a Luau Formatter on Save in VS Code and the Web Editor

What a safe Luau formatting diff should look like

Confirming you are running the Sparkdown document formatter

Luau Formatter Spacing Rules: The Before-and-After Reference

One-line Luau tables and table types: inner brace spacing

Tight suffixes: generics, optional ?, and interpolation braces

Operators that keep their spaces: :: and ..=

Anatomy of `0 9 * * 1-5`: every token, decoded

Steps are not intervals: why `*/7` is not “every 7 minutes”

The Linux workaround: escaped `%` and a date test inside the command

AWS EventBridge and Systems Manager: six fields, `?`, Sunday = 1

What does `/5 * * *` mean, and what does `5` mean instead?

Tight suffixes: generics, optional `?`, and interpolation braces

Operators that keep their spaces: `::` and `..=`

Why did formatting turn `return {1}` into `return{1}`?