[Bug]: Axes.bar_label() on log scale does not center the label. #23688

Activity

1. tacaswell commented on Aug 21, 2022

   

   tacaswell

   on Aug 21, 2022

   Member

   More actions

   https://matplotlib.org/stable/tutorials/text/annotations.html#using-complex-coordinates-with-annotations I think you want to use option 2 in that section and place the text relative to the bounding box of the rectangle.

   The back-compatibility on this may require some thought.
2. stefmolin commented on Aug 21, 2022

   

   stefmolin

   on Aug 21, 2022

   ContributorAuthor

   More actions

   @tacaswell - Did you mean something like this?

   import matplotlib.pyplot as plt
   
   fig, ax = plt.subplots()
   bar_container = ax.barh(['c', 'b', 'a'], [1_000, 5_000, 7_000])
   ax.set_xscale('log')
   ax.set_xlim(1, None)
   for bar in bar_container:
       ax.annotate(f'{bar.get_width():,d}',
                   xy=(0.5, 0.5), xycoords=bar,
                   xytext=(0, 0), textcoords='offset points',
                   ha='center', va='center')

   This only seems to work if the x in xy is 1, as I can move the text vertically by changing the y and the result is as expected. However, if I change the x value to 0.5 to place it halfway along the bar, nothing shows up, and if I do 0.99 the text is way beyond the beginning of the x-axis:
   
3. tacaswell commented on Aug 21, 2022

   

   tacaswell

   on Aug 21, 2022

   Member

   More actions

   Oh, that makes sense. The bounding bax it is using is not clipped so the left edge is way off the the left at where ever the very small number we clip 0 up to in log scale is. Same log problem, but from the other side.....
4. anntzer commented on Aug 21, 2022

   

   anntzer

   on Aug 21, 2022

   Contributor

   More actions

   The following (which was made easy by your debugging above!) seems to work:

   from matplotlib.transforms import Bbox
   import matplotlib.pyplot as plt
   
   fig, ax = plt.subplots()
   bar_container = ax.barh(['c', 'b', 'a'], [1_000, 5_000, 7_000])
   ax.set_xscale('log')
   ax.set_xlim(1, None)
   for bar in bar_container:
       ax.annotate(
           f'{bar.get_width():,d}',
           xy=(0.5, 0.5),
           xycoords=lambda r, b=bar: Bbox.intersection(b.get_window_extent(r), b.get_clip_box()),
           xytext=(0, 0), textcoords='offset points',
           ha='center', va='center')
   
   plt.show()
5. stefmolin commented on Aug 21, 2022

   

   stefmolin

   on Aug 21, 2022

   ContributorAuthor

   More actions

   Thanks @anntzer – I'm glad my debugging helped get to the bottom of this. Can I put in a PR to update Axes.bar_label() to properly handle the label_type='center' going forward? Unless this is something you don't want in the API, it seems more robust than what is already there and I imagine the logic will work for any axis scale.
6. anntzer commented on Aug 21, 2022

   

   anntzer

   on Aug 21, 2022

   Contributor

   More actions

   Seems good to me (although I admittedly haven't followed the whole discussion).
7. stefmolin

   added a commit that references this issue on Aug 21, 2022

   Implement solution from matplotlib#23688 to properly center bar label…

   ...

   d18d7d7
8. tacaswell

   added this to the v3.7.0 milestone on Aug 24, 2022
9. greglucas

   closed this as completedin #23698on Sep 11, 2022
10. melissawm

    added a commit that references this issue on Dec 19, 2022

    Implement solution from matplotlib#23688 to properly center bar label…

    ...

    f5ebe5f