BUG: fix Panel.fillna() ignoring axis parameter (re-submission) #11445
Conversation
jreback
added
Bug
Missing-data
| @@ -1764,6 +1764,35 @@ def interpolate_2d(values, method='pad', axis=0, limit=None, fill_value=None, dt | |||
| return values | |||
|
|
|||
|
|
|||
| def interpolate_nd(values, method='pad', axis=0, limit=None, fill_value=None, dtype=None): | |||
There was a problem hiding this comment.
if you are going to do this, ok, but I think a reorg of all of these little functions is in order. If before this PR (e.g. send another), put all of the interp methods that are in core/common.py into core/missing.py would be helpful
|
I merged the previous Looking through the code, I see a number of places where the terms "pad", "fill", and "interpolate" are used semi-interchangably and with different meanings. I think it would clear things up if, at least internally, these terms had these fixed definitions: fill: fill in missing values with a static value I'm not suggesting any changes to the public API, but I would be willing to work on a new PR to try and makes things internally consistent. I don't want to confuse this PR by trying to roll it in here. In any event, this fix seems to work OK. Not sure if it needs more documentation other than mention of bugs fixed in whatsnew. |
| if ndim == 3: | ||
| if axis == 0: | ||
| for n in range(shape[1]): | ||
| values[:,n] = func(values[:,n], axis=1) |
There was a problem hiding this comment.
this looks quite duplicative.
this type of if/then is so error prone. just compute the axis to use.
|
updated |
| limit=limit, | ||
| fill_value=fill_value, | ||
| dtype=self.dtype) | ||
| values = mis.pad(values, |
There was a problem hiding this comment.
make this import as missing and not mis
|
Implemented most of the suggestions - just need clarification on what is meant by about the ndims being the same. |
|
|
||
| if ndim == 3: | ||
| axis = 0 if (axis > 1) else 1 | ||
| else: |
There was a problem hiding this comment.
you are changing the meaning of axis here. it is exactly [slice(None)]*ndim indexed by axis for all types.
this is equivalent to .loc[.....] and NOT repeated sectioning. I know that may be confusing, but I said that for your tests are doing incorrect indexing. pls make the changes.
|
@stahlous do you want to rebase / update. fyi, you might want to squash as well. |
|
I squashed the commits. Sorry for the delay - had other things going on the past couple weeks. Can you take a look and let me know if I've addressed everything? |
| @@ -2774,40 +2774,28 @@ def fillna(self, value=None, method=None, axis=None, inplace=False, | |||
| # set the default here, so functions examining the signaure | |||
| # can detect if something was set (e.g. in groupby) (GH9221) | |||
| if axis is None: | |||
| axis = 0 | |||
| axis = self._stat_axis_name | |||
There was a problem hiding this comment.
need to document this 'change' in the doc-string
|
closing, we are going to deprecate |
closes #3570
closes #8251
This is a re-submission of PR #8395 and addresses issue #8251. My apologies for the duplicate PR, but despite rebasing to pydata/pandas master, the original PR would not update with my new commits and I couldn't get rid of the
LooseVersion()errors in the Travis build.This PR may need some fleshing out still, but I wanted to submit this to see what the thoughts are on this approach. For this PR I've created a new interpolation mechanism at the block level to implement filling across 3 or more dimensions. This avoids the fiddly problems of trying to implement filling of 3+ dimensions at the frame level. However, it isn't possible with this technique to fill across blocks of different dtypes, although that seems like that should be a rare occurrence.
Incidentally, this will also address #3570, which is one of the issues referenced in #9862.