regenconf is super slow

[selfhoster1312]

I placed a "bug" label because it's unbearably slow even on a really fast machine.

There's a ton we can do to make it faster... i've been digging into the source code and so far i've found:

• the regenconf file is called many times because it's read in a loop for category, conf_files in _get_pending_conf(names).items(): conf_hashes = _get_conf_hashes(category)
• md5 is slow... do we actually need hashing? can't we use mtime timestamps instead? if we need hashing can we use a much faster hashing algorithm?
• rewrite it in rust for blazing fast correctness and fearless concurrency? 8)

I have yet to properly benchmark this on a raspberry pi, but on x86, a naive (not optimized) reimplementation in Rust gets:

Command	Python	Rust	Python (raspi)	Rust (raspi)
regen-conf --list-pending	0.265s	0.003s	11.555s	0.026s
hook list conf_regen	0.150s	0.002s	5.975s	0.025s

I haven't reimplemented the rest of regenconf (yet), but there's not a lot of big features missing (syntax coloring, custom log sharing..) which i don't think is as important as running fast (and could be reimplemented if really needed). Two orders of magnitude of slowness (x100) is quite a big deal! :(

[alexAubin]

md5 is slow...

Honestly I very much doubt that md5 hashing is the issue ... just look at the call to yunohost settings get etc from the bash hooks ...

[selfhoster1312]

Added benchmarks from an actual Raspberry pi running the two versions (Python/Rust).