Implement "in memory" backend#39
Conversation
References python-cachier#4 and references python-cachier#6
Now, the default is calculated dynamically to maintain previous behavior to default to pickle unless the ``mongetter`` argument is given. This makes it have the same behavior as before, so users aren't forced to change old code to use the ``backend`` argument
Closes python-cachier#6 Still needs unit tests
@shaypal5 these are copied from the pickle tests, the only one failing is test_memory_being_calculated. I could use your help with the implementation of the relevant function becuase I don't think I understand the threading part
|
|
||
| def wait_on_entry_calc(self, key): | ||
| entry = self.cache[key] | ||
| # I don't think waiting is necessary for this one |
There was a problem hiding this comment.
it turns out this assumption isn't at all correct 😆
|
Yes, you need locking for thread safety. :) I'm pulling this into a branch here and I'll add the locking functionality myself + README additions. |
| backend : str, optional | ||
| The name of the backend to use. If None, defaults to 'mongo' when | ||
| the ``mongetter`` argument is passed, otherwise defaults to 'pickle'. | ||
| Valid options currently include 'pickle' and 'mongo'. |
There was a problem hiding this comment.
Please add 'memory' as a valid option.
|
@shaypal5 the change of merge target was a bit unexpected, I think it's much better to just push to my branch for consistency. Anyway I hope that my commits don't get squashed and lost in the history once you finish it :) |
|
I just didn't want to have to wait for you to approve PRs from me to your branch, to then just have these as part of this PR. That sounds like a vicious cycle of waiting. I just wanted to get this done. :) |
|
And I AM going to squash all your commits into a single one, just for a more readable history. I've started to use dynamic rebasing a lot recently, and can finally appreciate how important it can be to maintain clean commit history. :) |
|
Nice! I think GitHub allows the repo owner to make commits to branches that are in PRs back to the upstream. I've done this before too |
|
Released in Congrats and great work, @cthoyt ! First major outside contribution to this project! |
2 participants
Closes #6
This PR implements an "in memory" backend. It mostly excludes all of the locking and waiting functionalities that were in the pickle backend, since the in-memory should have negligible write and read times (in comparison to pickle). I may have misunderstood the purpose of that functionality, so if it's needed, I think it will be easy to put back in.
This PR also includes the changes from #38 because those were necessary to enable switching between multiple backends.
Caveat the tests aren't written in a very reusable way, so I'm not sure the best way to test this without duplicating a lot of code. I'd appreciate some advice/help on what to do here