Add command line option to ignore cache

[dgouldin]

I ❤️ Black!

It would be nice to have the option to ignore its local cache in order to force it to reevaluate all files. Perhaps a --no-cache flag?

Thanks!

[ambv]

What's your use case for this?

[JelleZijlstra]

This would have been useful for me during development. While I was testing .pyi mode, I repeatedly ran black /some/file/in/typeshed.pyi, but black would often refuse to do anything because the file hadn't been modified since the last time I ran it. There are other workarounds, like editing the file, but a --no-cache option would have made life easier for me.

[ambv]

There are a number of things we could do to make developmen a bit easier but that's not a use case I'd like to prioritize.

A need to ignore the cache suggests there is something wrong with it that needs to be fixed.

[tiran]

In the mean time and on Linux, you can set XDG_CACHE_HOME to an invalid path to suppress caching:

$ XDG_CACHE_HOME=/invalid python3 -m black black.py 

[ambv]

X Desktop Group to the rescue.

[dgouldin]

From the source, it looks like a cache hit is determined by a file's size and modified time. I can conceive of ways this could result in false positives, but admittedly it would be an unusual case.

I think it's reasonable to say you don't want to prioritize this feature. I hadn't seen it proposed yet, so I thought I would. Thankfully the cache storage path is well documented, so when I was uncertain about the cache hits I was seeing, I was able to rm -rf the cache itself.

[ambv]

If this is reassuring to you, what Black is caching on is exactly what git status is using to determine changes. It would take a deliberate attempt or a broken file system to fool it.

Or a bug, which is why I'd like those exposed and fixed rather than worked around.

[ambv]

I'll close this as this is not a planned feature.

[sanmai-NL]

What's your use case for this?

How about not caching stuff for ephemeral, CI-like runs?

[sanmai-NL]

There are a number of things we could do to make developmen a bit easier but that's not a use case I'd like to prioritize.

A need to ignore the cache suggests there is something wrong with it that needs to be fixed.

Suggests, but doesn't imply. Not all calls to disable a feature are due to (functional) correctness defects, it's also a matter of performance efficiency.