Details
- Reviewers
- None
- Group Reviewers
Restricted Project
Diff Detail
- Repository
- rK kolab
- Branch
- dev/mollekopf
- Lint
Lint Skipped - Unit
No Test Coverage - Build Status
Buildable 47967 Build 18189: arc lint + arc unit
Event Timeline
This only implements migrating imap messages but highlights how the current framework doesn't cleanly facilitate not using the job per item strategy.
I think there's a cleaner approach to implementing the two strategies than what I did above, e.g. by moving the processFolder logic entirely to the exporter.
| src/app/DataMigrator/Kolab.php | ||
|---|---|---|
| 180 ↗ | (On Diff #13818) | I think we should probably do this in steps, e.g. 50 or 100 messages per job. E.g. get all UIDs in folder and invoke a job for every 50 of them. |
| 192 ↗ | (On Diff #13818) | A corner case, but Message-Id header may not exist. Maybe fallback to md5(Date.From.To.Subject). Also, I guess we'd need to handle \Seen state change (and probably all other flags). |
| 196 ↗ | (On Diff #13818) | There is handlePartBody() that can be used to get the full message source in a memory efficient way using streams. |
| src/include/rcube_imap_generic.php | ||
| 29 | This is not needed, but you probably have to run composer dump-autoload. | |
Good progress, but I was about to work on DAV, and I already have some conflicting code. Please, don't work on DAV at this moment.
Small imap fixup, let the exporter implement the migration strategy
So EWS can do per item jobs, and IMAP not.
I though we discussed on monday that I would finish with dav... Anyways, if you like you can just take this over. The basics seem to work for me, but on the dav side I only looked at events (for addressbooks/todos the serialization is missing).
From my side this is good enough for a merge though.
For production I think we'll have to introduce a secondary job processing queue for long-running jobs.
In D4845: IMAP data migrator and other improvements I took this differential of yours, but I re-implemented some things differently. I think it's cleaner. Still a lot of TODO lines there. So, we might improve while writing tests. I have also an idea to try to do mail streaming for best performance.