Recurring/Scheduled/"Cron-like" Tasks
Opened this issue · 4 comments
The README states that only the at-like functionality is implemented, and the cron-like functionality is planned for a future release. Upon investigation of how resque-scheduler works, it seems to rely on an external library to actually act as the cron portion, which then merely schedules tasks using the at-like functionality actually built in.
So my question - and this issue is more of a discussion starter than a bug report - is whether we want to implement this functionality in this manner, or if we wish to take a different approach that leverages the fact we're using Redis already anyway. I have an idea that I'd like to flesh out a bit before I propose it more formally, but I was curious if there were any other opinions already out there.
While waiting for this to be implemented, I created a quick class that will allow you to create a recurring job: https://gist.github.com/jesperborgstrup/93d943da96c653645b7d
That's more or less what I had in mind. I personally don't want to limit recurring jobs by class alone, so the final implementation would need to be adjusted some to allow developers to override the identifier with something else, but otherwise, my suggestion would be to adopt this approach, and adjust it so it isn't in the job class, but in the library itself.
Though it would also be good to support a more cron-like spec for intervals, probably in addition to a regular number-of-seconds approach.
As something like this isn't yet implemented in PHP Resque Scheduler I am using your class @jesperborgstrup, thanks for that.
However, from what I've read from your main notes:
- If a job is scheduled, and another schedule already exists for the same class, the old schedule
- will be overwritten.
So let's say I use those jobs to grab some information about a user somewhere on the Internet. I cannot schedule jobs for each user at the same time, because all users use the same class? Is that it?
If so, why did you put that restriction and how can I bypass it?
I guess you could remove the restriction, but do you really want to schedule a recurring job for each user?
Wouldn't it make more sense to have a single job to grab information about all users (or let the job decide which users to grab information for, and then do it)?