REF: reindex_indexer use np.void to avoid JoinUnit #43376
Merged
Conversation
This file contains hidden or 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
jreback
reviewed
Sep 4, 2021
jreback
added
Internals
jreback
reviewed
Sep 6, 2021
pandas/core/internals/concat.py
Outdated
| @@ -245,6 +247,38 @@ def concatenate_managers( | |||
| return BlockManager(tuple(blocks), axes) | |||
|
|
|||
|
|
|||
| def _reindex_columns_void( | |||
There was a problem hiding this comment.
i would rename to _maybe_reindex_columns_na_proxy, e.g. don't use the term void as we don't use it elsewhere and instead use na_proxy (I know the dtype is V which is fine); alt make this clear in the doc-string here.
jreback
approved these changes
Sep 6, 2021
|
greenish, just test_to_csv_compression_encoding_gcs |
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Outdated suggestions cannot be applied.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.
ATM we do a kludgy thing where we sometimes reindex before calling concat (e.g. in DataFrame.append) and that leads to all-NA columns that we choose to ignore when choosing the dtype for the concatenated column. This leads to a) funky behavior when we encounter an "real" all-NA column and b) a big performance hit in checking for all-NA columns.
Instead of doing that, this PR implements the keyword use_na_proxy (chosen to match ArrayManager's keyword) that ends up creating ndarrays with np.void dtype instead of all-NA float64 ndarrays.
This will allow us to simplify a bunch of code in internals.concat (e.g. get rid of case where
unit.block is None) and improve the funky behavior mentioned above. Both of these are left to follow-ups to keep the diff manageable.