PropagateDirectory: Investigate if _subJobs should be pre-allocated (reserve()) or squeeze()

[guruz]

--- Want to back this issue? **[Post a bounty on it!](https://www.bountysource.com/issues/40930590-propagatedirectory-investigate-if-_subjobs-should-be-pre-allocated-reserve-or-squeeze?utm_campaign=plugin&utm_content=tracker%2F216457&utm_medium=issues&utm_source=github)** We accept bounties via [Bountysource](https://www.bountysource.com/?utm_campaign=plugin&utm_content=tracker%2F216457&utm_medium=issues&utm_source=github).

[ckamm]

@ogoffart I'd say we can squeeze() _subJobs in directories when we're done with them in OwncloudPropagator::start, do you agree?

[jturcotte]

I think that this is solved by #5400 as the whole PropagateDirectory job is now destroyed by its parent when it's finished.

[guruz]

@jturcotte I mean the pre-running peak. If you squeeze() or reserve() it won't take so much space in memory BEFORE running. (directly after constructing).

@jturcotte @ckamm Agree to reopen?

See @ckamm 's comment above..

[jturcotte]

I interpret "when we're done with them" as "when the job if finished".

reserve() is to prevent reallocating the vector as you fill it if you know the size already, but we don't know it since we iterate over the vector of SyncFileItem until we hit a different parent directory. It doesn't reduce the used memory.

squeeze() is the other way around, when you're removing objects from the vector, it won't reallocate immediately to avoid memcpying to a new location. What I meant is that it won't bring us anything to do this at the end of the PropagateDirectory job, since the whole job (and its vector) will be destroyed immediately after. We could do squeeze after each subjob is done, but this means that we'll reallocate the whole vector each time, only to save a bit of extra memory for the duration of the PropagateDirectory and not reducing the peak, so I'm not sure if it's worth it.

[jturcotte]

Ahh you guys mean to remove the extra capacity that append might reserve, yes that might help save a few KBytes.

[guruz]

Exactly!

[jturcotte]

You can only do this at the end of OwncloudPropagator::start after the foreach, so you'll already have created all the subjobs and reached the peak at that point, but could keep things tidy with #5451

[ckamm]

This has been resolved through #5451. During setup we only build a QVector<SyncFileItemPtr> now. We could squeeze that but it'd be a marginal gain.