Paweł ŚwiątkowskiYou learn interesting things when testing super-fast #hanami apps.
For example: that Faker is very slow to load and you should probably avoid it.
Or that database cleaning in really slow, so you should only clean when you really use the database.
Tim Riley@katafrakt woah, I’d have never considered this!
Peter Solnica@timriley @katafrakt this makes sense, faker comes with a loooot of files and we know that require is dead slow. Would be nice to do something about it.
Simon Telahug@stoivo @timriley @katafrakt yes but it doesn't change the fact that gems could either use zeitwerk or at least be cherry-pickable so that you don't have to load tons of files when you only need a small portion of an entire library. Also, bootsnap may cause issues sometimes, or at least that was my experience last time I used it.
@solnic @timriley @katafrakt right, you could save some time by only loading what you need with zeitwerk or alike
@solnic @timriley @katafrakt I did a some testing yesterday. I suspect the lowest part is reading and merging yaml files https://github.com/faker-ruby/faker/issues/2719
@solnic @timriley @katafrakt No, it loads all yaml files when you need the first translation. Most fakers use a translation as a set it samepls from. Faker::Crypto::Sha1 is one of the few which don't
@katafrakt @solnic @timriley whats the best way of benchmarking require, is it possible to do with ips?
Adam Lassek@solnic @timriley @katafrakt I tracked this down to `I18n.available_locales`, because Faker busts the cache when it loads.
Unfortunately this means zeitwerk/bootsnap are unlikely to make a difference. You can call this during your test suite startup so the time doesn't get randomly added to a test run.