Added Interactive Legend to forestplot Bokeh #1591

Rishabh261998 · 2021-03-02T11:06:21Z

Description

Added the Legend to forest plot in with backend Bokeh. I have also made it interactive but wasn't able to control the different axes with a single legend so had to provide a legend for each axes. See the below image

One more problem I am facing is that I am unable to hide the y_ticks corresponding to the model when the model plot is hidden. It seems that Bokeh supports interactivity with only glyphs.

Follows official PR format
Includes a sample plot to visually illustrate the changes (only for plot-related functions)
New features are properly documented (with an example if appropriate)?
Includes new or updated tests to cover the new feature
Code style correct (follows pylint and black guidelines)
Changes are listed in changelog

Rishabh261998 · 2021-03-02T11:09:00Z

@OriolAbril Can you look into this. I was also facing problem in rebasing so I used the git merge command instead of rebasing.

codecov · 2021-03-02T11:48:51Z

Codecov Report

Merging #1591 (0fae999) into main (43ceb17) will increase coverage by 0.02%.
The diff coverage is 98.50%.

@@            Coverage Diff             @@
##             main    #1591      +/-   ##
==========================================
+ Coverage   90.05%   90.07%   +0.02%     
==========================================
  Files         108      108              
  Lines       11584    11608      +24     
==========================================
+ Hits        10432    10456      +24     
  Misses       1152     1152

Impacted Files	Coverage Δ
arviz/plots/backends/bokeh/forestplot.py	`95.46% <98.50%> (+0.33%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 43ceb17...b7582b5. Read the comment docs.

OriolAbril

I had very little time to review, will continue afterwards.

Added the Legend to forest plot in with backend Bokeh. I have also made it interactive but wasn't able to control the different axes with a single legend so had to provide a legend for each axes. See the below image

This is amazing, don't worry about the one legend per axes issue. This is perfectly fine.

What is strange is that the ess/rhat dots should be aligned with the forestplot hdi lines, otherwise they are not very useful. Does this happen always or only with the legend?

One more problem I am facing is that I am unable to hide the y_ticks corresponding to the model when the model plot is hidden. It seems that Bokeh supports interactivity with only glyphs.

Again, this is no problem at all, I recommend users use the NoModel labeller (or mixtures of it) which should be the default already with legend=True. With everything working well, there is only a label per variable, see the last example in the label guide for instance: https://arviz-devs.github.io/arviz/user_guide/label_guide.html#labeller-mixtures. What should not be happening are the 0 and 12 as ticklabels, I will try to take a look afterwards, but wanted to point it out already.

I was also facing problem in rebasing so I used the git merge command instead of rebasing.

This is to be expected. You are using the same branch you used for the rope PR, which has the changes in the rope PR, but we merged it to master already so you have the same changes in different commits, which makes rebasing tricky.

We recommend keeping the main branch up to date with the main branch in arviz, and never work on it. And every time you need to make a PR, create a new branch for that feature. After merging, we delete the branch and prune the repo from time to time.

OriolAbril · 2021-03-02T12:25:57Z

arviz/plots/backends/bokeh/forestplot.py

+ legend_it_ess = []
+ for (model_name, glyphs) in plotted_ess.items():
+ legend_it_ess.append((model_name, glyphs))
+ legend_ess = Legend(items=legend_it_ess)
+ axes[0, idx].add_layout(legend_ess, "right")
+ axes[0, idx].legend.click_policy = "hide"


I think all this should only happen if legend=True. Same below

OriolAbril · 2021-03-02T12:35:43Z

arviz/plots/backends/bokeh/forestplot.py

- yield y, row_label, label, selection_list[idx], values, self.model_color[
- model_name
- ]
+ yield y, row_label, model_name, label, selection_list[


This is for the code above, not for here, but you need to update the label generation to use the labeller: https://github.com/arviz-devs/arviz/blob/main/arviz/plots/backends/matplotlib/forestplot.py#L555

By this, you just mean that I need to generate the row labels as done In matplotib?

Both row labels and model labels used in legend should be processed by the labeller, you'll see that in matplotlib there are two labeller methods called

Rishabh261998 · 2021-03-02T14:15:51Z

@OriolAbril Also the ess/rhat dots are not aligned in the current bokeh forestplot without legend, whereas the matplotlib version is perfect. Also the 0 and 12 problem also appears in the bokeh forestplot without legend.

Rishabh261998 · 2021-03-02T14:18:04Z

And also should I create a new branch or is it fine to work on this branch just for this PR? i.e is there any problem if I use git merge again if I need to sync. Thanks

OriolAbril · 2021-03-02T20:40:13Z

@OriolAbril Also the ess/rhat dots are not aligned in the current bokeh forestplot without legend, whereas the matplotlib version is perfect. Also the 0 and 12 problem also appears in the bokeh forestplot without legend.

Do you want to work on these here or open an issue for you or someone else to work on there at some point later in time?

And also should I create a new branch or is it fine to work on this branch just for this PR? i.e is there any problem if I use git merge again if I need to sync. Thanks

For now it is fine I think, if you ran into problems or the number of commits increased I's probably suggest a rebase, but let's go to than if we get there.

Rishabh261998 · 2021-03-02T20:59:44Z

I think for now we should create a new issue and when I have finished with the legend thing in foresplot and also the ppc _plot then I can take this or someone else who is interested to work can also take it up. Thanks

Rishabh261998 · 2021-03-03T20:12:01Z

@OriolAbril in the below plot why are there only orange ess dots, plot is from matplotlib backend

OriolAbril · 2021-03-03T22:57:04Z

@OriolAbril in the below plot why are there only orange ess dots, plot is from matplotlib backend

No idea. I am not sure ess and rhat are very useful for multiple models case, but it should not happen, and if it does then we should document that ess/rhat do not work with multiple models.

Rishabh261998 · 2021-03-04T10:44:17Z

@OriolAbril fixed the label generation and also fixed the 0, 12 error which was happening before:

This is the current plot with bokeh as backend and using NoModel labeller. As you can see after updating the label generation in bokeh as it was in matplotlib the ess dots are only shown for a single model. For this I will create a separate PR and fix this problem for both mpl and bokeh. Also see the below 2 plots 1 for mpl and 1 for bokeh with labeller as DimCoordLabeller. In this the ess dots are plotted for both the model.
For Matplolib:

For Bokeh:

OriolAbril · 2021-03-04T12:50:10Z

Also see the below 2 plots 1 for mpl and 1 for bokeh with labeller as DimCoordLabeller. In this the ess dots are plotted for both the model.

Oh, I see, good catch. The issue is that the code must be checking a dict or set to see if a dot has been plotted, and with the new labeller capabilities, labels don't need to be unique. I should have updated the code to stop using labels as ids but completely missed that. A new PR to fix this would be great.

OriolAbril

Thanks, this is great

OriolAbril · 2021-03-04T12:52:38Z

arviz/plots/backends/bokeh/forestplot.py

@@ -191,8 +202,10 @@ def plot_forest(
 ] = -all_plotters[ # pylint: disable=protected-access
 0
 ].group_offset
- axes[0, 0].y_range._property_values["end"] = y_max # pylint: disable=protected-access
+ axes[0, 0].y_range._property_values["end"] = y_max # pylint: disable=protected-


undo this change to have pyling keep ignoring this warning

Did this, still checks are failing:-C
Update: Got the error, docstring missing in a public method

* Included estimate of probability within ROPE * Positioned ROPE text and added to CHANGELOG * Changed color scheme in plot_posterior in bokeh * Changed size of rope text in posteriorplot bokeh * Added rope_color and ref_val_color arguments to plot_posterior * Minor Changes * Some minor changes * Added Interactive Legend to forest plot * Formatted Code * Updated label generation and fixed some errors * Updated CHANGELOG * Minor change

Rishabh261998 added 11 commits February 22, 2021 17:52

Included estimate of probability within ROPE

29580ac

Positioned ROPE text and added to CHANGELOG

86d36d4

Changed color scheme in plot_posterior in bokeh

54634d5

Changed size of rope text in posteriorplot bokeh

0556c5d

Added rope_color and ref_val_color arguments to plot_posterior

16ca922

Minor Changes

55f353e

Some minor changes

b4e8fb7

Merge remote-tracking branch 'upstream/main' into my-feature

97f2ffe

Added Interactive Legend to forest plot

2e1bd56

Syncing fork

4cb5581

Formatted Code

15a988a

Rishabh261998 changed the title ~~My feature~~ Added Interactive Legend to forestplot Bokeh Mar 2, 2021

OriolAbril reviewed Mar 2, 2021

View reviewed changes

Rishabh261998 added 2 commits March 4, 2021 15:55

Updated label generation and fixed some errors

aaeb3f4

Syncing remote

b7582b5

OriolAbril approved these changes Mar 4, 2021

View reviewed changes

Rishabh261998 added 2 commits March 4, 2021 19:56

Updated CHANGELOG

dc8c495

Minor change

d650344

OriolAbril mentioned this pull request Mar 4, 2021

Fix ess/rhat plotting in plot_forest #1596

Closed

OriolAbril merged commit ae3ce51 into arviz-devs:main Mar 4, 2021

Rishabh261998 deleted the my-feature branch March 4, 2021 19:40

Rishabh261998 mentioned this pull request Mar 8, 2021

Fix ess/rhat plot in plot_forest #1606

Merged

6 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Added Interactive Legend to forestplot Bokeh #1591

Added Interactive Legend to forestplot Bokeh #1591

Rishabh261998 commented Mar 2, 2021 •

edited

Loading

Rishabh261998 commented Mar 2, 2021

codecov bot commented Mar 2, 2021 •

edited

Loading

OriolAbril left a comment

OriolAbril Mar 2, 2021

OriolAbril Mar 2, 2021

Rishabh261998 Mar 2, 2021

OriolAbril Mar 2, 2021

Rishabh261998 commented Mar 2, 2021 •

edited

Loading

Rishabh261998 commented Mar 2, 2021

OriolAbril commented Mar 2, 2021

Rishabh261998 commented Mar 2, 2021

Rishabh261998 commented Mar 3, 2021

OriolAbril commented Mar 3, 2021

Rishabh261998 commented Mar 4, 2021

OriolAbril commented Mar 4, 2021

OriolAbril left a comment

OriolAbril Mar 4, 2021

Rishabh261998 Mar 4, 2021 •

edited

Loading

Added Interactive Legend to forestplot Bokeh #1591

Added Interactive Legend to forestplot Bokeh #1591

Conversation

Rishabh261998 commented Mar 2, 2021 • edited Loading

Description

Rishabh261998 commented Mar 2, 2021

codecov bot commented Mar 2, 2021 • edited Loading

Codecov Report

OriolAbril left a comment

Choose a reason for hiding this comment

OriolAbril Mar 2, 2021

Choose a reason for hiding this comment

OriolAbril Mar 2, 2021

Choose a reason for hiding this comment

Rishabh261998 Mar 2, 2021

Choose a reason for hiding this comment

OriolAbril Mar 2, 2021

Choose a reason for hiding this comment

Rishabh261998 commented Mar 2, 2021 • edited Loading

Rishabh261998 commented Mar 2, 2021

OriolAbril commented Mar 2, 2021

Rishabh261998 commented Mar 2, 2021

Rishabh261998 commented Mar 3, 2021

OriolAbril commented Mar 3, 2021

Rishabh261998 commented Mar 4, 2021

OriolAbril commented Mar 4, 2021

OriolAbril left a comment

Choose a reason for hiding this comment

OriolAbril Mar 4, 2021

Choose a reason for hiding this comment

Rishabh261998 Mar 4, 2021 • edited Loading

Choose a reason for hiding this comment

Rishabh261998 commented Mar 2, 2021 •

edited

Loading

codecov bot commented Mar 2, 2021 •

edited

Loading

Rishabh261998 commented Mar 2, 2021 •

edited

Loading

Rishabh261998 Mar 4, 2021 •

edited

Loading