ENH: Styler.background_gradient to accept vmin vmax and dtype Int64 #29245
Conversation
Resolve pandas-dev#12145 and pandas-dev#28869 For `vmin` and `vmax` use the same implementation in `Styler.bar` For dtype `Int64` issue, deprecated `.values` and use `.to_numpy` instead Here explicitly assign the dtype to float since we are doing normalize
pandas/io/formats/style.py
Outdated
| @@ -958,6 +958,8 @@ def background_gradient( | |||
| axis=0, | |||
| subset=None, | |||
| text_color_threshold=0.408, | |||
| vmin=None, | |||
There was a problem hiding this comment.
Can you annotate the new parameters?
There was a problem hiding this comment.
Thanks for your review @WillAyd, but I don't quite get it what do you mean by "annotate", can you elaborate a little more? (I have docstring already so I suppose you don't mean docstring right?)
There was a problem hiding this comment.
I got it! you mean the type-hints annotation right?
pandas/io/formats/style.py
Outdated
| smin = s.values.min() | ||
| smax = s.values.max() | ||
| smin = s.min() if vmin is None else vmin | ||
| if isinstance(smin, ABCSeries): |
There was a problem hiding this comment.
Is this not already covered by the line directly preceding it?
There was a problem hiding this comment.
Here depends on the axis argument provided to Styler.background_gradient, s can be a DataFrame or a Series. if s is a DataFrame and vmin not provided, smin will first become a Series then become the min value of the full data, the code here is consistent with (aka. direct copy from) Styler.bar method, do we want to change it?
There was a problem hiding this comment.
side note: try to use s.values.min() to get overall minima will produce AttributeError: 'IntegerArray' object has no attribute 'min' as described in #28869. OTOH, use np.min(s.values) will also produce NotImplementedError: The 'reduce' method is not supported. I thought IntegerArray might be a WIP and .min().min() is the workaround.
There was a problem hiding this comment.
Can you move this into a helper function? I think this affects highlight_extrema as well but seems to tackle slightly differently; should keep unified
squeeze might also be more appropriate to use here
There was a problem hiding this comment.
Thanks! didn't notice the highlight_extrema, they really should be consistent. turned-out a fairly simple approach can be used: np.nanmin(s.to_numpy()), s can be any shape
| rng = smax - smin | ||
| # extend lower / upper bounds, compresses color range | ||
| norm = colors.Normalize(smin - (rng * low), smax + (rng * high)) | ||
| # matplotlib colors.Normalize modifies inplace? | ||
| # https://github.com/matplotlib/matplotlib/issues/5427 | ||
| rgbas = plt.cm.get_cmap(cmap)(norm(s.values)) | ||
| rgbas = plt.cm.get_cmap(cmap)(norm(s.to_numpy(dtype=float))) |
There was a problem hiding this comment.
OK with this but can just use default constructor without dtype argument I think? As it it seems like this is trying to coerce, which I'm not sure why that would be needed
There was a problem hiding this comment.
unfortunately, .to_numpy() alone will produce dtype=object and rejected by matplotlib.
OTOH, it will immediately convert to float anyway since cmap require normalizing to [0,1], nothing to loose for coercing in the first place.
…_gradient-to-accept-vmin-vmax-and-int64 sync with the latest upstream/master
|
Thanks @immaxchen! |
…andas-dev#29245) * ENH: Styler.background_gradient to accept vmin vmax and dtype Int64 Resolve pandas-dev#12145 and pandas-dev#28869 For `vmin` and `vmax` use the same implementation in `Styler.bar` For dtype `Int64` issue, deprecated `.values` and use `.to_numpy` instead Here explicitly assign the dtype to float since we are doing normalize
…andas-dev#29245) * ENH: Styler.background_gradient to accept vmin vmax and dtype Int64 Resolve pandas-dev#12145 and pandas-dev#28869 For `vmin` and `vmax` use the same implementation in `Styler.bar` For dtype `Int64` issue, deprecated `.values` and use `.to_numpy` instead Here explicitly assign the dtype to float since we are doing normalize
…andas-dev#29245) * ENH: Styler.background_gradient to accept vmin vmax and dtype Int64 Resolve pandas-dev#12145 and pandas-dev#28869 For `vmin` and `vmax` use the same implementation in `Styler.bar` For dtype `Int64` issue, deprecated `.values` and use `.to_numpy` instead Here explicitly assign the dtype to float since we are doing normalize
For
vminvmaxuse the same implementation asStyler.barFor dtype
Int64issue, deprecated.valuesand use.to_numpyinsteadHere explicitly assign the dtype to float since we are doing normalize
style.background_gradienton anInt64column #28869black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff