State of the library & moving forward

[sheerlox]

Roadmap

Based on the bellow observations (see Context), here's what I think would be a good way going forward:

V2 features and bugfixes

• feat: import type definitions from @types/cron #669
• fix: replace loop timeout by max match date #686
• fix: fix range parsing when upper limit = 0 #687
• fix(deps): update dependency luxon to v3.3.0 & add @types/luxon #689
• fix: added fractional offset support #685

V3 release (breaking changes)

• feat!: UNIX standard alignments #667 (available on beta channel)
• fix!: return empty array from nextDates when called without argument #519 (available on beta channel)
• refactor!: migrate to TypeScript #694
• feat!: rework utcOffset parameter #699
• support both ESM & CJS (NodeJS/Deno/Bun) (see [DISCUSSION] Module resolution: ESM vs CJS #700)

post V3 release (ordered by priority)

• Error: Something went wrong. cron reached maximum iterations. #467
• [BUG] 2-0 week expression error #654
• README API documentation overhaul #441
• error handling #426
• option to wait for job completion before executing the next one #556
• option to wait for callbacks completion on job.stop() #713
• support extended cron syntax #435
• Some jobs will be skipped in some cases #315
• Cron Job executing twice on a single scheduled job #489
• cron fails on two servers, approximately same time, without error #423
• Delayed Jobs with suspend #671
• support using both day-of-month and day-of-week in a cron expression
• browser support (partial work done in feat: support browser environment #629)
• performance analysis & improvements

Misc/Rehabilitation

• automated release process & enforced conventional commits (ci: setup conventional commits & release automation #673)
• thorough inspection of every currently opened issue, push forward on fixing critical ones, and close outdated/irrelevant ones
• take over relevant staled opened PR and push them through, and close outdated/irrelevant ones
• add dependencies bot & unify dependencies version pinning (GitHub app needs to be installed, see ci(renovate): configure renovate #683)
• improve the GitHub "Community Standards" for this repository #696

Candidates for next major release

• swap C & OC generic type parameters on every type (QoL)
• add runOnce as a CronJob option
• revamp all class properties (make some private & potentially add getters/setters)

Less urgent

• re-enable CodeCov
• automate checking if the mandatory "Related Issue" field in PR is filled

Context

Click to expand

Motivation

In my opinion, the OpenSource spirit is more about trying to maintain/improve existing projects (especially when they are "industry leaders") than trying to create a new project overseeding the existing ones.

---

On February 12, this issue was created to replace cron with croner (by the creator of the later) in the @nestjs/schedule package. Work is being done by the same individual in this PR to meet that goal.

In this @nestjs/nest issue (one year ago), OP mentioned this issue from our repository, that is closed but according to some comments, it might still be occurring (see issue #467 for follow-up on this), although we cannot be sure users are using the latest version. We should implement test cases that somehow limit/overload cpi & memory resources to test for this kind of problem (this has been successfully implemented, see #467 (comment)).

Here are the arguments to migrate to croner according to its creator (from the @nestjs/schedule PR):

supports both ESM and CJS.

• migrating to Typescript will enable supporting both ESM and CJS

is ready for the future, supporting Node, Deno, Bun and Browser environments.

• I never tried to use this library in the browser, not sure what are the use cases but I guess we can make sure it is portable. This should be taken care of during the Typescript migration as well.
• Regarding Deno & Bun, I have no experience with those but we would need to look further into that (looks like Deno only supports ESM and that it's the only blocking point. Bun should already be working according to their docs)
• Browser-support related PR: feat: support browser environment #629
  

has more predictive behaviour and less bugs. (not affected by Scheduled jobs execute multiple times #454 as an example)

• this issue (Scheduled jobs execute multiple times nestjs/schedule#454) has been resolved for two years (not explicitly but if it was still the case we would have heard about it), but there might be a good point regarding predictability and bugs.

uses the same pattern as vixie cron (see EVERY_YEAR Cronexpression is invalid #1159)

• we could add the support for L for the last day and weekday of the month in cron strings?
• there is an issue opened about this: support extended cron syntax #435

has no dependencies (see Vulnerability found in dependency #1147)

• I personally don't think having luxon as a dependency is an issue, it's even a feature since it allows us to have support for the DateTime object. That is debatable though, and I'm welcoming any objecting point of view so we can find the solution that best fits everyone's use cases.
• We definitely need to add some dependencies scanning & automated version bumps PR tool (Renovate/Dependabot), as is standard in the Javascript OpenSource ecosystem.

As another useful reference, here's a link to the features comparison from croner.

Two features I think are particularly interesting:

• Over-run protection: an option to prevent the next occurrence from running if the first one is still processing (long-running jobs)
• Error handling: this is something I've seen multiple times in our issues' comments, if something goes wrong it seems to go unnoticed most of the time, maybe we should think about a better way of detecting and reporting potential errors to the user (EDIT: error handling in croner is actually providing a way to handle errors that might occur in the tick function, see Hexagon/croner/EXAMPLES.md#error-handling)

---

---

This issue was initially discussed on Discord.

[mryellow]

Given the state of #467 and the rate at which it keeps me entertained with report after report, I think this whole thing needs to be abandon. There are other options and they work.

p.s. TypeScript won't solve anything.

[sheerlox]

@mryellow if you want to move to another library, please feel free to do so :)
If, on the other hand, you'd like to start being constructive and helpful, then useful comments and PRs are welcome.