feat: make Table.fill_null() have consistent semantics with .select(), etc #11698
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
github-actions
bot
added
tests
NickCrews
force-pushed
the
consistent-fill_null
branch
from
56e7f4a to
20997c8
Compare
…, etc
This is breaking.
Before, doing `t.fill_null("a")` filled with the string literal "a".
It also was impossible to use column references such as
`t.fill_null(t.other_col)` or `t.fill_null(ibis._.other_col)`.
Now, it is consistent with how .select(), .mutate(), etc all work.
This also simplifies the internal representation so that the ops.FillNull always stores a `Mapping[str, <bound ops.Value>]`
This also changes the behavior of the polars backend.
Before, Table.fill_null() would fill in NaN.
Now it does not, consistent with other backends,
and the rest of ibis's Column.fill_null() semantics,
including how polars treats Column.fill_null()
NickCrews
force-pushed
the
consistent-fill_null
branch
from
20997c8 to
d5b6c2f
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
This is breaking.
Before, doing
t.fill_null("a")filled with the string literal "a". Now it tries to bind against a column "a".It also was impossible to use column references such as
t.fill_null(t.other_col)ort.fill_null(ibis._.other_col). Now, it is consistent with how .select(), .mutate(), etc all work, and those two examples work.This also changes the behavior of the polars backend. Before, Table.fill_null() would fill in NaN. Now it does not, consistent with other backends, and the rest of ibis's Column.fill_null() semantics, including how polars treats Column.fill_null()
This does mean there is now some inconsistency between Table.fill_null() and Value.fill_null(). Value.fill_null("a") interprets the "a" as a literal value, not a column reference. We could adjust that API as well, or at least add a very visible warning to both docstrings. But then what does that mean for the other Value APIs, like Value.isin(["a"])? Is "a" a literal or column reference? Maybe the difference is if the method is on a Table, all strings are treated as column references, but on Values they are treated as literals. To be even more breaking, but what I think would be the most consistent, would be if we made "a" ALWAYS be a literal, and you HAVE to use
_.ato reference a column. This would be super breaking, but we could probably provide a decent upgrade path by using a heuristic ofis "a" a column in the table?to give warnings for several ibis versions.As a non-public-facing refactor, this also simplifies the internal representation so that the ops.FillNull always stores a
Mapping[str, <bound ops.Value>]