OPT: Delay Dataset.repo access until necessary #5076
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Results in substantial speedup for recursive processing, as datasets without any submodules are rejected prior any Repo instantion. For a dataset with 42k subdatasets this means going from ``` datalad subdatasets -r 270.45s user 615.64s system 104% cpu 14:07.22 total ``` to ``` datalad subdatasets -r 5.82s user 0.69s system 101% cpu 6.422 total ``` Fixes dataladgh-5075
|
Awesome! Thank you @mih! Now the problem is only that I need to figure out what to do with all that free time ;-) |
|
FWIW, even our tiny benchmarks show the gain which is great, so we don't miss some pessimisation if any comes |
Codecov Report
@@ Coverage Diff @@
## master #5076 +/- ##
==========================================
+ Coverage 89.78% 89.82% +0.04%
==========================================
Files 293 293
Lines 41254 41255 +1
==========================================
+ Hits 37039 37057 +18
+ Misses 4215 4198 -17
Continue to review full report at Codecov.
|
yarikoptic
approved these changes
Oct 23, 2020
yarikoptic
added a commit
to yarikoptic/datalad
that referenced
this pull request
Nov 20, 2020
While looking at py-spy stack from long running status I noticed that majority of time is spent on git config call, triggered by checking for fake dates config setting. AFAIK for read-only operations, which cannot produce a commit, it might benefit to skip that check. And _call_git already has that kwarg, but it is not proxied by "higher-level" helpers. To avoid breeding workarounds (like the one RFed within this commit) I decided that we might benefit from exposing that option in higher-level helpers as well. Then "read-only" invocations could set it to False and we might avoid needless config read in some cases (I have not done any timing on any prototypical use case which might benefit from this, but I would assume that status or diff might). We had an original discussion on either to expose check_fake_dates in higher level *Repo interfaces before: datalad#3791 (comment) and decided to not do it at that point. As for 'subdatasets' call, performance issue was now addressed with datalad#5076 , so this change is not strictly necessary to optimize "subdatasets", but would still be generally benefitial for possible other invocations of *Repo methods across many instances without unnecessarily triggering loading of the config, typically in the case of read-only git operations (git-annex might need to merge git-annex branch, so typically annex calls should not disable fake dates).
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Results in substantial speedup for recursive processing, as datasets
without any submodules are rejected prior any Repo instantiation.
For a dataset with 42k subdatasets this means going from
to
Fixes gh-5075