Try to figure out where/how to insert the cache here #740
Conversation
|
@ebolyen is this along the lines of what we want? It seems to be doing what we wanted. This isn't the part of the framework I'm most familiar with. |
|
Almost, I would lift this logic up to Is the process pool the best place? It seems kind of reasonable, although it's looking less like a pool every day. The automated cleanup is a nice feature of it. We should also test this with your parsl situation with the provenance path not showing up. My hunch is it will still work since there's a flufl lock, but worth confirming before we break everything. |
|
Yeah while we were messing with that I was wondering if this functionality would resolve that |
|
Ok so we do want to put this in |
|
Something like how it is now and stop passing in prefix? I had figured this would bork things pretty bad, but maybe not. |
|
Yep, in theory that's all we should need to do. |
qiime2/core/path.py
Outdated
|
|
||
| cache = get_cache() | ||
| kwargs['prefix'] = os.path.join(cache.process_pool.path, | ||
| 'q2-%s-' % cls.__name__) |
There was a problem hiding this comment.
Let's put this under a tmp/ directory in the process pool (creating if needed). Maybe there should be a helper on the cache itself to return the temp directory?
|
This does not work how I thought it did at all. The actual path is one of the args, and the prefix doesn't seem to do much if anything when passed up to pathlib.Path |
|
@ebolyen I think putting the cache usage in |
| @@ -92,14 +92,20 @@ def _destruct(cls, path): | |||
| else: | |||
| os.unlink(path) | |||
|
|
|||
| def __new__(cls, dir=False, **kwargs): | |||
There was a problem hiding this comment.
Where there any callers of this? I swear we set prefix in a lot of places for stuff like q2-provenance... although come to think of it, cache has probably eroded out most of those callsites, but there may still be a few left.
There was a problem hiding this comment.
Only existing use is
Line 23 in c95b12b
Closes #730