BUG: df.iloc[:, :1] with EA column #32959
Conversation
jbrockmendel
added
Bug
Indexing
|
im now thinking the test doesnt belong in extension tests; if it were to fail (e.g. in the status quo), that wouldnt indicate that the EA author needed to fix somthing |
jbrockmendel
force-pushed
the
bug-iloc-ea
branch
from
d0a68fb
to
628e242
Compare
|
i think comments have been addressed @jorisvandenbossche, can you confirm? |
pandas/core/internals/blocks.py
Outdated
| raise AssertionError( | ||
| "invalid slicing for a 1-ndim ExtensionArray", first | ||
| ) | ||
| elif not (first == slice(None) or first == slice(1)): |
There was a problem hiding this comment.
I think the check for slice(1) is too strict. Eg, you can slice with out of bounds ends, like slice(2), and that should work as well
| # return same dims as we currently have | ||
| if not isinstance(slicer, tuple) and self.ndim == 2: | ||
| # reached via getitem_block via _slice_take_blocks_ax0 |
There was a problem hiding this comment.
Could also put this check in getitem_block or _slice_take_blocks_ax0 to ensure that _slice always gets the correct tuple of slices?
There was a problem hiding this comment.
Or, directly check the value? As here you convert it to a len-2 tuple, and then in a next step, you convert the tuple back to a slice?
There was a problem hiding this comment.
Hmm one way to avoid this PITA would be to have 2D EAs...
|
@jorisvandenbossche have your concerns/comments been addressed here? |
|
gentle ping @jorisvandenbossche |
| result = df.iloc[:, :1] | ||
| self.assert_frame_equal(result, df) | ||
|
|
||
| result = df.iloc[:, :2] |
There was a problem hiding this comment.
can you add some df.iloc[:, -1:] as well
|
needs a rebase as well |
|
thanks @jbrockmendel |
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff