ENH: Add lazy copy to replace #50746
Conversation
| value_getitem(val_locs), BlockPlacement(blk.mgr_locs[blk_locs]) | ||
| ) | ||
| if self.refs is not None: | ||
| self.refs.extend([self.refs[blkno_l]] * len(nbs_tup)) |
There was a problem hiding this comment.
Do we need to clear the refs at self.refs[blkno_l] now, since first_nb should not have a reference block?
There was a problem hiding this comment.
Thx for catching, must have removed the line accidentially
There was a problem hiding this comment.
Adjusted one of the tests to cover this
| ([0], np.array([-1, -2, -3], dtype=np.intp)), | ||
| ([1], np.array([-1, -2, -3], dtype=np.intp)), | ||
| ([5], np.array([-1, -2, -3], dtype=np.intp)), | ||
| ([0, 1], np.array([-1, -2, -3], dtype=np.intp)), |
There was a problem hiding this comment.
cc @lithomas1 I don't think that this is valid or can be reached in this way. I think the dimension of your values has to be the same as your indexer. This raises if blk.should_store_value is False, because it does not get broadcast.
There was a problem hiding this comment.
You're right. Even for the non CoW, the iset operation doesn't fail(strangely), but printing the DataFrame afterwards does.
There was a problem hiding this comment.
Yep you get a block that has dimension 2 according to block placement but the underlying array has only one column, doesn’t really error when creating but as soon as you access it youll get into trouble :)
Co-authored-by: Thomas Li <[email protected]>
|
thanks @phofl. |
|
Thx to you as well, you did most of the work :) |
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.cc @lithomas1 as discussed on slack