NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #303 most downloaded on PyPI
croniter provides iteration for datetime object with cron like format
Latest stable
7.0.0
Last release 13 days ago
09 Nov 2026
Release timing varies
gaps range from 8 days to 1.2 years
Nearly every release is documented
notes for 58 of the last 60 stable releases
Nothing withdrawn
no release was ever pulled
17 years old
107 releases · first in 2010
Drag across the ruler to compare, or tap a release to open it.
Prepare for 6.2.5 development by @potiuk in #242
Full Changelog: 6.2.4...7.0.0
Your coding agent can read these notes before it upgrades. Set up the MCP server →
This is a major release because several bug fixes change what existing, valid expressions mean. Read Upgrading to 7.0 before you upgrade. Every change is listed in full under Bugfixes.
Changes 1 and 2 are silent: nothing fails, schedules just fire at different times. Changes 3 and 4 raise an exception.
1. A step that starts at the field maximum now means one value, not the whole field (#246). Each field's maximum is 59 for seconds and minutes, 23 for hours, 31 for day-of-month, 12/DEC for month, 6/SAT (or 7) for day-of-week and 2099 for years.
Who is affected: any expression in which a field starts with its maximum followed by /, for example 59/15, 23/6, 31/5, DEC/3, SAT/2, 6/1 or 2099/1.
What to do: find these expressions in your code, configuration and stored data. A field value such as 59/... or 6/... is the pattern to look for. Decide what each one was meant to do:
It meant "every": replace it with * or */{step}. * * * * 6/1 becomes * * * * *, and 59/15 * * * * becomes */15 * * * *.
It meant "only the maximum": no change is needed, because 7.0 now does that.
It is in a year field (2099/1) or on a day that can be missing (31/5 with month 2): the expression may now have no future match and raise CroniterBadDateError (see 4 below). Rewrite it with *.
2. ``expand_from_start_time=True`` keeps the bounds of ranges (#255, #256, #246). If you never pass expand_from_start_time, skip this item: default-mode schedules are unchanged.
Who is affected: expressions with an explicit range, an equal range such as 1-1/3, a wrapping range such as 6-0, or a stepped start such as 5/15. In 6.x such a range could fire outside its own bounds, so * 8-17 * * * fired at hours 0-17. The phase was also read in UTC instead of the start time's own time zone.
What to do: for each schedule that uses the flag, compare the old and new expanded values. Install 7.0 in a separate environment, or run the snippet below under both versions with your own expressions, start times and time zones:
from datetime import datetime, timezone
from croniter import croniter
start = datetime(2026, 1, 1, 7, 7, tzinfo=timezone.utc) # use your real start
for expr in ["* 8-17 * * *", "5/15 * * * *"]: # your expressions
print(expr, croniter(expr, start, expand_from_start_time=True).expanded)
Where the output differs, 7.0 matches the expression as written. If you relied on the old output, write that schedule out explicitly, for example as a list of values, instead of relying on the flag. A schedule can now fire less often (45/15 fires once an hour, not four times), or more often in the case of equal and wrapping ranges (1-1/15 in the minute field fires four times an hour). See the warning under #255 for the full breakdown.
3. Some expressions that used to be accepted now raise ``CroniterBadCronError``.
Who is affected: expressions containing any of the following:
non-ASCII digits or non-ASCII whitespace (#284, #288);
a H(a-b) or R(a-b) range outside its field (#289);
day-of-month 0 in the 6-field seconds form (#280);
an out-of-range # such as 1#0 or 1#7 (#278).
What to do: where expressions come from users or storage, check them with croniter.is_valid() before you upgrade, and fix the ones that fail. Each rejected form either matched something the text did not say or depended on the hash id, so there is no behaviour to keep.
4. Some schedules now raise ``CroniterBadDateError`` or ``TypeError``/``ValueError``.
An expression changed by item 1 can have no remaining match. get_next then raises CroniterBadDateError, so rewrite the expression as described in item 1.
max_years_between_matches must now be an int >= 1 or None (#272). Replace True with 1, 0 or a negative number with 1, and pass int(...) instead of a float or string.
Other changes in behaviour. These are corrections of wrong results, and no action is expected:
get_next, get_prev, match and croniter_range no longer report times invented inside a DST gap (#266).
pytz zones with negative DST offsets, such as Europe/Dublin, no longer crash (#267).
Day-of-week 7 is accepted in the 6- and 7-field forms (#263).
A few hashed divisions wider than their range (H/60, H/100) produce different values (#253).
If you cannot audit your expressions yet, pin croniter<7.
Reject hashed and random ranges whose bounds lie outside their field, such as H(0-99) in the minute field. The bounds were only checked through the value drawn from them, so such an expression was valid for some hash ids and not others, and is_valid("R(0-99) * * * *") returned True or False depending on the draw. They now raise CroniterBadCronError for every hash id and draw. [#289, @Raghuvardhan-Reddy78]
max_years_between_matches must be an int greater than or equal to 1, or None. bool is rejected (True used to become 1 via int(True)). 0, negatives, floats, and strings are rejected too: 0 and negatives used to be raised to 1, floats were truncated, and numeric strings were parsed by int(). A bad type raises TypeError. An int below 1 raises ValueError. [#272]
Reject non-ASCII whitespace between cron fields. str.split() splits on any Unicode whitespace, so 0 0 * * * written with a no-break space or an ideographic space between the fields passed is_valid and ran as 0 0 * * *, while cron only separates fields on spaces and tabs. These now raise CroniterBadCronError. The ASCII separators \x1c-\x1f are no longer treated as field separators either. Follow-up to #284. [#288, @Raghuvardhan-Reddy78]
Fix day-of-month 0 being silently rewritten to 1 and accepted in the 6-field seconds form (e.g. 0 0 0 * * *): it is now rejected with CroniterBadCronError, matching the 5-field form fixed for #9 and the 7-field years form. [#280]
Fix get_next/get_prev raising AssertionError for pytz timezones whose tzdb encoding uses a negative DST offset (for example Europe/Dublin). At an ambiguous local time the resolution assumed the is_dst=True interpretation is always the later UTC instant; such zones invert that ordering, so the internal assertion tripped and evaluation crashed instead of returning a fire time. Both interpretations are now ordered by their absolute instants and the first one satisfying the successor requirement in the requested direction is returned. [#267, @abhijeet117]
Fix day-of-week 7 (Sunday) being rejected in the 6-field (seconds) and 7-field (year) forms. 7 is now accepted as an alias for 0 in every form, as it already was in the classic 5-field form. [#263, @nikolauspschuetz]
Reject non-ASCII digits in cron expressions. The parsing regexes used Unicode \d and \w, and int() also accepts any Unicode decimal digit, so ٥ * * * * (Arabic-Indic five) or 0 0 * * * (fullwidth zeros) passed is_valid and ran as 5 * * * * and 0 0 * * *. These now raise CroniterBadCronError. [#284, @Raghuvardhan-Reddy78]
Fix get_next, get_prev, match and croniter_range reporting a DST gap shift as a fire time. When a cron slot fell into a spring-forward gap and resolving the alternative across the offset change walked forward to the next existing wall clock, that manufactured wall clock was returned although its fields did not match the expression: for 0 * * * * in Australia/Lord_Howe, whose transition is only 30 minutes, get_prev from 03:00 landed on 02:30, get_next from 01:03 returned 02:30, match(02:30) was true, and croniter_range listed 02:30. A shifted alternative is now discarded like the regular candidate already was when hours are wildcarded, so the four APIs agree again. [#266, @abhijeet117]
Fix get_next/get_prev raising CroniterBadDateError when day-of-month and day-of-week are both restricted and the day-of-month can never occur in the selected month (e.g. 0 0 31 2 0). Under the Vixie OR semantics the unsatisfiable side now contributes no dates instead of aborting the whole expression, and match() and croniter_range(), which swallow the error, no longer return silently wrong results for these expressions. [b245ab6, #243, @semx, @MildlyMeticulous]
Fix hashed divisions (H/{divisor}, H({begin}-{end})/{divisor}) when the divisor is as wide as, or wider than, the range it is drawn from. Three symptoms, one cause: the offset was drawn from the first period without regard for where the range ends, and the result was emitted as {offset}-{end}/{divisor} even when the step could not reach a second value. H/60 lost minute 59 and fired at minute 0 twice as often as any other minute (59 distinct schedules over 20,000 hash ids, not 60); H(50-59)/10 fired six times an hour instead of once for the ids hashing to 59; H(50-52)/10 fired outside its own range for 7 hash buckets in 10, and H/100 raised CroniterBadCronError for roughly a third of hash ids. The offset is now clamped to the declared range and emitted as a single value when the step cannot reach a second one. Divisors that fit their range keep their existing hashes, so no schedule that was already correct moves. [#253, @potiuk]
Fix expand_from_start_time taking the phase of a cycle from the UTC form of a timezone-aware start_time rather than from the wall clock the caller wrote. _get_low_from_current_date_number read the start time back with tz=UTC, so 0 */5 * * * from 2025-01-01 05:00+13:00 phased on UTC hour 16 and expanded to hours 1,6,11,16,21 instead of 0,5,10,15,20. The hour, day-of-month, month and day-of-week fields were all affected, and at a date boundary the day and month could be a whole day or month out. The start time's own tzinfo is now carried into the expansion. A naive start_time is converted to a timestamp as if it were UTC, so UTC round-trips it to the same wall clock and naive callers see no change at all — verified over 45,312 expansions with naive start times, none of which differ. [#256, @potiuk]
Fix expand_from_start_time raising ValueError: Can't get current date number for index larger than 4 for any range or step in the optional seconds or years field. croniter("* * * * * */15", start, expand_from_start_time=True) aborted, and because the exception came from outside the CroniterError hierarchy, code catching CroniterError never saw it. Both fields now re-base like the five that already worked: seconds take their phase from the start time's second, and years — which do not start at zero — take theirs from the field minimum, as day-of-month and month already do. [#254, @potiuk]
Reject out-of-range nth day-of-week values (1#0, 1#7) with an ordinary check instead of an assert. Under python -O the assert was stripped, so 1#0 passed is_valid and fired on every Monday and 1#7 passed and never fired. [#278, @Raghuvardhan-Reddy78]
Fix {start}/{step} expanding to the whole field when start is the field maximum. 59/15 * * * * fired at :00/:15/:30/:45 instead of :59, and the same held in every field at its own maximum: 23/6 hours, 31/5 day-of-month, DEC/3 months, SAT/2 day-of-week, and — not in the original report — 59/15 seconds and 2099/5 years. {start}/{step} normalizes to {start}-{max}/{step}; when start is the maximum the two bounds collide, making the token indistinguishable from an explicitly written equal range such as Jan-Jan, which croniter deliberately expands to the whole cycle. The two forms are now distinguished. An explicitly written equal range still means the whole cycle, and every start below the field maximum is unchanged. The fix applies in expand_from_start_time mode as well as by default. (Under that flag the whole cycle an equal range names now takes its phase from start_time, as every other cycle does — see #255 below.) [#246, @earfman]
Warning
This changes existing schedules, silently and substantially. Any {max}/{step} expression that was firing across the whole field now fires once per cycle. The {max}/1 forms are the sharpest case, because they read as "every": * * * * 6/1 meant every day and now means Saturdays only; 59/1 * * * * meant every minute and now means minute 59. The same applies to 23/1 hours, 31/1 day-of-month, 12/1 months, and to the optional seconds and years fields (59/1, 2099/1). Use a wildcard (* or */1) for "every". Before upgrading, audit for any field whose value begins with that field's maximum followed by a slash.
One consequence is worth stating separately: a {max}/{step} expression can now be unsatisfiable rather than merely different, because it no longer covers the whole field. 0 0 1 1 * * 2099/1 previously expanded the year field to every year and now means 2099 alone, so iterating from a later start time raises CroniterBadDateError instead of returning a date. The same happens without a year field wherever the remaining value cannot occur: 0 0 31/5 2 * is now day 31 of February.
Fix expand_from_start_time discarding the lower bound of every range. Re-basing overwrote low outright instead of supplying the phase of the cycle, so any range whose start was above the field minimum fired below its own start: * 8-17 * * * expanded to hours 0-17, * * * * MON-FRI included Sunday, and 10-50/15 expanded to minutes 7,22,37 from a start at minute 7. Over a sweep of 101,088 expand_from_start_time expansions — all seven fields, every legal lower bound, eight token shapes, seven steps, four start times — 73,898 contained at least one value outside the range the expression declares; now none do. Re-basing still supplies the phase — */15 is unchanged, and so is 5/15 whenever the start time's phase is at or above 5 (from minute 2 it now gives 17,32,47, not 2,17,32,47) — but the phase is advanced to the first value inside the declared range, and where no value of the phase fits, the expression is honoured as written, so an expansion can never come out empty. Hashed and random divisors (H/15, R/15, H(10-50)/7) take their phase from the start time inside their declared range, exactly like */15 and 10-50/7: the hash offset is a phase, so it is not read as a lower bound. [#255, @potiuk]
Two side effects of overwriting the bound are fixed with it: colliding low onto high made an ordinary range read as an explicit equal range and expand to the whole cycle (* * 10-11/2 * * became every second day of the month), and re-basing a wrapping range unwrapped it (* * * * 6-0, Saturday and Sunday, became every day of the week).
An explicitly written equal range means the whole cycle, and that cycle now takes its phase from start_time like any other, so the two spellings of one schedule agree: 1-1/3 and */3 in the month field both expand to February, May, August and November from a start in February. Under the old behaviour they agreed by accident — the discarded bound happened to land on the phase — and pinning the bound without this would have made 1-1/3 ignore the flag outright, contradicting what the README documents for it.
Also fixed, on the same line: a start_time of exactly the Unix epoch skipped re-basing altogether, because the start time reaches the expansion as a timestamp and 0.0 is falsy. 0 0 * * 2-6/3 from 1970-01-01 expanded to days 2,5 where the same expression one second later expanded to day 4.
Warning
This changes most ``expand_from_start_time`` schedules. 93,456 of the 101,088 expansions in the sweep differ. Any explicitly written range, equal range, wrapping range or stepped start is affected; the untouched shapes are the plain wildcard, the bare value, and */step except where the step is wide enough to collide the field's own bounds (*/60 in the minute field re-based to minute 0 rather than to its phase). Default-mode expansions do not move at all, no expansion became empty, and no error status changed.
A schedule can fire less often than it did. 45/15 goes from four fires an hour to one, which is what that expression already means without the flag. More generally, re-phasing a step inside a fixed window can cost one period even when nothing was wrong with the schedule: 10-20/7 is 10,17 by default but 14 alone from a start at minute 7. Measured against the default expansion of the same expression, no re-based expansion in the sweep fires more often, 13,539 fire less, and the drop is never more than one period and never to zero.
A schedule can also fire more often than it did, in 17,038 of the swept expansions. Every one of them is an explicit equal range or a wrapping range — the two shapes the old behaviour re-read as something else. 1-1/15 in the minute field expanded to [0] from a start at minute 0 and is now [0, 15, 30, 45], the whole cycle the equal range asks for; the wrapping range 2-1, which covers the whole minute field, expanded to minutes 0,1 and is now every minute. These are meanings being restored, not values being invented: no expansion holds a value the expression does not declare.
Document what expand_from_start_time actually means — that it moves the phase of a cycle and not the values a field may take, that every field re-bases including the optional seconds and years, and that the phase is read in the start time's own timezone. Also document that a range with two equal bounds (Jan-Jan, Sun-Sun) means the whole cycle rather than the value it names, which was undocumented and is the rule behind several of the bugs fixed in this release. [#257, @potiuk]
Document the bounds half of expand_from_start_time: that a field declaring its own range keeps both bounds and only the phase moves, that moving a phase inside a fixed window can cost one period, that an expression whose phase does not fit is honoured as written, and that a wrapping range is left alone. [#255, @potiuk]
Bump pinned GitHub Actions via dependabot. [6e69553, #247; 3c6ce9b, #251; d2e8a13, #260; d9f40f0, #269; 4be99c3, #274; 9fe300d, #285; 2725ef5, #290]
Bump pinned build dependencies (hatchling, packaging) via dependabot. [70564f2, #245; 3dd4d14, #262; e3ba628, #265; 309703f, #283; b4c7661, #291]
Remove the unused workflow that built the 32-bit Docker test images. [1b0c241, #292, @potiuk]
Fix expand_from_start_time day-of-week low bound so Sunday wraps correctly for stepped day-of-week ranges ( isoweekday() % 7 instead of weekday() + 1
Fix expand_from_start_time day-of-week low bound so Sunday wraps correctly for stepped day-of-week ranges (isoweekday() % 7 instead of weekday() + 1). [44e3090, #239, @alhudz]
Bump pinned GitHub Actions via dependabot. [fd13d2e, #240; 5cd2e46, #238]
Features and Improvements ~~~~~~~~~~~~~~~~~~~~~~~~
Fix quadratic expansion of comma-separated range lists for a large speed-up on expressions with many ranges. [1d79fc6, #231, @alhudz]
Reject a zero step (e.g. 5-5/0) in equal and reversed cron ranges instead of silently accepting it. [ab27767, #232, @alhudz]
Fix expand_from_start_time month low-bound off-by-one so stepped month ranges start on the correct month. [f329964, #235, @alhudz]
Fix zizmor-reported security findings in GitHub Actions workflows. [ed19f94, #230]
Bump pinned build and CI dependencies via dependabot.
Upgrade locked development and build dependencies (uv lock --upgrade).
Switch build backend from flit_core to hatchling with pinned dependencies. [4cce160]
Switch build backend from flit_core to hatchling with pinned dependencies. [4cce160]
Add PEP 639 license expression (license = "MIT") to project metadata. [4cce160]
Pin all GitHub Actions to SHA hashes across CI workflows. [4cce160]
Add dependabot for automated dependency updates. [4cce160]
Add zizmor workflow for CI security scanning. [4cce160]
Fix get_prev skipping Feb 29 on leap years for day-of-month expressions. [441a0b5, #203, @Souls-R]
Fix get_prev skipping Feb 29 on leap years for day-of-month expressions. [441a0b5, #203, @Souls-R]
Add License :: OSI Approved :: MIT License classifier. [270cbbc, @cdheiser]
Features and Improvements ~~~~~~~~~~~~~~~~~~~~~~~~
Add support for W (nearest weekday) in the day-of-month field. [c79d88a]
Add precision_in_seconds parameter to match() and match_range() to allow overriding the default matching precision. [7415d1e, #58]
Add strict parameter to is_valid() and expand() for cross-field validation of impossible day/month combinations. [0352d48, #199]
Fix memory leak by removing EXPRESSIONS global dict cache. [6ebfe56, @Souls-R]
Document day_or and implement_cron_bug parameters for day-of-month/day-of-week interaction. [289bf73, #75]
Document precision_in_seconds parameter and default precision behavior for match(). [7415d1e, #58]
Features and Improvements ~~~~~~~~~~~~~~~~~~~~~~~~
Support for zoneinfo timezones. [b4e7295, Benjamin Drung (@bdrung)]
Add type hints to croniter.__init__, timestamp_to_datetime, and property initializations. [b171ea9, 257741e, c4f5e44, Benjamin Drung (@bdrung)]
Simplify code for max_years_between_matches. [e065efa, Benjamin Drung (@bdrung)]
Extend type check for hash_id to empty str/bytes. [6fe1d43, Benjamin Drung (@bdrung)]
Drop unused _get_next_nearest and _get_prev_nearest. [d7dab3a, Benjamin Drung (@bdrung)]
Reduce line length to 99 and unfold long Python code lines. [ecea402, f38d5f4, Benjamin Drung (@bdrung)]
Fix memory leak by removing TIMESTAMP_TO_DT_CACHE global dict cache. [1a9d3c0, Rafał Safin (@rafsaf)]
Fix memory leak by removing EXPRESSIONS global dict cache. [1883e15, @Souls-R]
Fix default value of second_at_beginning to a boolean. [dc63ed2, Benjamin Drung (@bdrung)]
Fix always missing the timestamp to datetime cache. [18eb299, Benjamin Drung (@bdrung)]
Fix skipping first of March. [de21a9d, Benjamin Drung (@bdrung)]
Fix DST handling by rewriting the DST logic. [f48387b, Benjamin Drung (@bdrung)]
Fix all flake8 complaints. [08c1f83, Benjamin Drung (@bdrung)]
Test: fix date in DST test. [871e391, Benjamin Drung (@bdrung)]
Test: document time jumps in DST test cases. [2d72258, Benjamin Drung (@bdrung)]
Test: use isoformat() to compare dates with timezone information. [1dd3562, Benjamin Drung (@bdrung)]
Announce back that croniter is maintained now as part of pallets-eco. [Jarek Potiuk (@jarekpotiuk)]
Nothing published for this version
Announce for now that croniter dev is ended (CRA).
Announce for now that croniter dev is ended (CRA).
Rework timestamp_to_datetime to use whatever timezone [kiorky]
Make datetime_to_timestamp & timestamp_to_datetime public [kiorky]
Fix EPOCH calculation in case of non UTC & 32 bits based systems [kiorky]
Apply isort formatter [kiorky]
Reintegrate test_speed [kiorky]
Apply black formatter [evanpurkhiser, kiorky]
Remove unused _get_caller_globals_and_locals [evanpurkhiser]
Remove single-use bad_length [evanpurkhiser]
Remove unused days in proc_month [evanpurkhiser]
Use field_index over i for readability [evanpurkhiser]
Always use """ for docstrings [evanpurkhiser]
Make helper instance methods that do not use self static [evanpurkhiser]
Remove unusd call to sys.exc_info [evanpurkhiser]
Remove unused ALPHAS [evanpurkhiser]
Improve croniter.expand documentation [evanpurkhiser]
Community wanted: Reintroduce 7 as DayOfWeek in deviation from standard cron (#90). [kiorky]
Community wanted: Reintroduce 7 as DayOfWeek in deviation from standard cron (#90). [kiorky]
Remove DayOfWeek alias 7 to DayOfWeek 0 to stick to standard cron (#90). [kiorky]
Remove DayOfWeek alias 7 to DayOfWeek 0 to stick to standard cron (#90). [kiorky]
Fix DOW ranges calculations when lastday is a Sunday. [kiorky]
Fix overflow on 32bits systems (#87) [kiorky]
Fix overflow on 32bits systems (#87) [kiorky]
Fix python2 testing (related to #93) [kiorky]
Modernize packaging. Special thanks to Aarni Koskela (akx) for all the inputs. [kiorky, akx]
- fix lint [kiorky]
fix lint [kiorky]
Fix start_time not respected in get_next/get_prev/all_next/all_prev (#86) [hesstobi, kiorky]
Fix start_time not respected in get_next/get_prev/all_next/all_prev (#86) [hesstobi, kiorky]
Add an update_current argument to get_next/get_prev/all_next/all_prev to facilitate writing of some downstream code, see #83. [kiorky]
Add an update_current argument to get_next/get_prev/all_next/all_prev to facilitate writing of some downstream code, see #83. [kiorky]
Support for year field [zhouyizhen, kiorky]
Support for year field [zhouyizhen, kiorky]
Better support for 6 fields (second), and 7 fields crons [zhouyizhen, kiorky]
Better fix hashed expressions omitting some entries (#82, #42, #30) fix is retained over #42 initial fix [zhouyizhen, kiorky]
Ensure match return false when not time available (#81) [zhouyizhen, kiorky]
- fix doc
fix doc
Implement second_at_beginning [zhouyizhen, kiorky]
Implement second_at_beginning [zhouyizhen, kiorky]
Support question mark as wildcard [zhouyizhen, kiorky]
Support to start a cron from a reference start time [mghextreme, kiorky]
- No changes, fix lint [kiorky]
No changes, fix lint [kiorky]
Support hashid strings in is_valid [george-kuanli-peng, kiorky]
Support hashid strings in is_valid [george-kuanli-peng, kiorky]
Avoid over-optimization in crontab expansions [Cherie0125, liqirui <liqirui@baidu.com>, kiorky]
Add match_range function [salitaba]
Add match_range function [salitaba]
fix leap year (29 days in February) [zed2015]
fix leap year (29 days in February) [zed2015]
- Fix release issue [kiorky]
Fix release issue [kiorky]
Add Python 3.12 support [rafsaf]
Add Python 3.12 support [rafsaf]
Make major release instructions [kiorky]
Make a retrocompatible version of 1.4.0 change about supporting VIXIECRON bug. (fix #47) [kiorky]
Make a retrocompatible version of 1.4.0 change about supporting VIXIECRON bug. (fix #47) [kiorky]
Added "implement_cron_bug" flag to make the cron parser compatible with a bug in Vixie/ISC Cron [kiorky, David White ] *WARNING*: EXPAND METHOD CHANGE
Added "implement_cron_bug" flag to make the cron parser compatible with a bug in Vixie/ISC Cron [kiorky, David White <dwhite2@cisco.com>] WARNING: EXPAND METHOD CHANGES RETURN VALUE
Fix hashed expressions omitting some entries [@waltervos/Walter Vos ]
Fix hashed expressions omitting some entries [@waltervos/Walter Vos <walter.vos@ns.nl>]
Enhance .match() precision for 6 position expressions [@szpol/szymon <szymon.polinkiewicz@gmail.com>]
- Lint
Lint
- Add check for range begin/end
Add check for range begin/end
- restore py2 compat
restore py2 compat
Do not expose i into global namespace
Do not expose i into global namespace
better error handling on py3 [kiorky]
Fix DOW hash parsing [kiorky]
better error handling on py3 [kiorky]
Add Python 3.11 support and move docs files to main folder [rafsaf]
Add Python 3.11 support and move docs files to main folder [rafsaf]
Fix croniter_range infinite loop [Shachar Snapiri ]
fix tests
Fix croniter_range infinite loop [Shachar Snapiri <ssnapiri@paloaltonetworks.com>]
Nothing published for this version
Add Python 3.10 support [eelkevdbos]
Add Python 3.10 support [eelkevdbos]
Really fix compat for tests under py27 [kiorky]
Really fix compat for tests under py27 [kiorky]
Fix compat for tests under py27 [kiorky]
Fix compat for tests under py27 [kiorky]
Fix #12: regressions with set_current [kiorky, agateblue]
Fix #12: regressions with set_current [kiorky, agateblue]
Restore compat with python2 [kiorky]
Restore compat with python2 [kiorky]
Add a way to make next() easier to use. This fixes #11 [kiorky]
Add a way to make next() easier to use. This fixes #11 [kiorky]
Enforce validation for day=1. Before this release we used to support day=0 and it was silently glided to day=1 to support having both day in day in 4t
Enforce validation for day=1. Before this release we used to support day=0 and it was silently glided to day=1 to support having both day in day in 4th field when it came to have 6fields cron forms (second repeat). It will now raises a CroniterBadDateError. See https://github.com/kiorky/croniter/issues/6 [kiorky]
Enforce validation for month=1. Before this release we used to support month=0 and it was silently glided to month=1 to support having both day in mon
Enforce validation for month=1. Before this release we used to support month=0 and it was silently glided to month=1 to support having both day in month in 4th field when it came to have 6fields cron forms (second repeat). It will now raises a CroniterBadDateError. See https://github.com/kiorky/croniter/issues/6 [kiorky]
- restore py2 [kiorky]
restore py2 [kiorky]
- better type checks [kiorky]
better type checks [kiorky]
Fix ZeroDivisionError with * * R/0 * * [cuu508]
Fix ZeroDivisionError with * * R/0 * * [cuu508]
Nothing published for this version
Add support for hashed/random/keyword expressions Ryan Finnie (rfinnie)
Add support for hashed/random/keyword expressions Ryan Finnie (rfinnie)
Review support support for hashed/random/keyword expression and add expanders reactor [ kiorky ]
fix bug: bad case:0 6 30 3 * [zed2015(zhangchi)]
fix bug: bad case:0 6 30 3 * [zed2015(zhangchi)]
Add support for L in the day_of_week component. This enable expressions like * * * * L4, which means last Thursday of the month. This resolves #159. [Kintyre]
Create CroniterUnsupportedSyntaxError exception for situations where CRON syntax may be valid but some combinations of features is not supported. Currently, this is used when the day_of_week component has a combination of literal values and nth/last syntax at the same time. For example, 0 0 * * 1,L6 or 0 0 * * 15,sat#1 will both raise this exception because of mixing literal days of the week with nth-weekday or last-weekday syntax. This may impact existing cron expressions in prior releases, because 0 0 * * 15,sat#1 was previously allowed but incorrectly handled. [Kintyre]
Update croniter_range() to allow an alternate croniter class to be used. Helpful when using a custom class derived from croniter. [Kintyre]
Remove external library natsort. Sorting of cron expression components now handled with sorted() with a custom key function. [Kintyre]
Remove external library natsort. Sorting of cron expression components now handled with sorted() with a custom key function. [Kintyre]
Remove futures dependency [kiorky]
Remove futures dependency [kiorky]
Update _expand to lowercase each component of the expression. This is in relation to #157. With this change, croniter accepts and correctly handles *
Update _expand to lowercase each component of the expression. This is in relation to #157. With this change, croniter accepts and correctly handles * * 10-L * *. [cuu508]
Fix _expand to reject int literals with underscores [cuu508]
Fix _expand to reject int literals with underscores [cuu508]
Remove a debug statement to make flake8 happy [cuu508]
Fix combination of star and invalid expression bugs [kiorky]
Fix combination of star and invalid expression bugs [kiorky]
Security fix: fix overflow when using cron ranges [kiorky]
Security fix: fix overflow when using cron ranges [kiorky]
- Spelling fix release
Spelling fix release
Fix #155: raise CroniterBadCronError when error syntax [kiorky]
Fix #155: raise CroniterBadCronError when error syntax [kiorky]
Fix match when datetime has microseconds [kiorky]
Fix match when datetime has microseconds [kiorky]
no changes, just to make sense with new semver2 (making croniter on a stable state) [kiorky]
no changes, just to make sense with new semver2 (making croniter on a stable state) [kiorky]
Added Python 3.8 and 3.9 support [eumiro]
Added Python 3.8 and 3.9 support [eumiro]
One column per quarter.