fix: add tools directly to Overlay

[Azaya89]

fixes #6240
fixes #6654

This PR fixes the problem where tools specified via .opts(tools=[...]) on an Overlay were being ignored. The OverlayPlot._init_tools() method only collected tools from subplots but never processed tools from self.default_tools and self.tools.

Solution

• Added logic to iterate through self.default_tools + self.tools and add them to the plot
• Improved tool deduplication: Replaced simple type-based deduplication (tool_types list) with unique identifier-based deduplication (tool_ids set). This is better since most bokeh tools like PanTool have variants differentiated only by their dimensions (e.g., 'pan', 'xpan', 'ypan' are all of type PanTool)
• Preserved special handling for hover tools and subcoordinate_y zoom tools which was causing the initial test failures
• Added tests

import holoviews as hv
hv.extension('bokeh')

overlay = (hv.Curve([1, 2, 3]).opts(tools=['xpan']) * hv.Curve([2, 3, 4]).opts(tools=['ypan', 'zoom_out'])).opts(tools=['zoom_in', 'xpan', 'ypan'])
overlay

[Azaya89]

This works now but it seems a bit verbose to me. Will try to find a more elegant way to implement...

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.22%. Comparing base (47f696f) to head (cdfbfe3).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6713      +/-   ##
==========================================
+ Coverage   89.21%   89.22%   +0.01%     
==========================================
  Files         332      332              
  Lines       72027    72120      +93     
==========================================
+ Hits        64259    64352      +93     
  Misses       7768     7768              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[hoxbro]

Have pushed some changes. Left some todos and info with your name directly in the code.

One thing I noticed is that the tool ordering seems weird; it would be nice to have more consistency.

I also think we should check that this fixes #6654.

[Azaya89]

Suggested change

# INFO(Azaya): Can this really be right?

Hmm, I found this interesting...

The test passes because it validates the internal tool ordering (i.e, checking plot.state.tools) but visually when you actually plot the same code, the ordering is now different (['PanTool', 'BoxZoomTool', 'WheelZoomTool', 'ZoomOutTool', 'ZoomInTool', 'SaveTool', 'ResetTool']). Maybe this is a result of how Bokeh handles the visual arrangement of the tools?

[hoxbro]

I don't know why this is the case at the top of my head. Try to see if you can understand by looking at the code and trying to use pure bokeh examples. Maybe also related to merge_tools.

[Azaya89]

I think it's definitely from bokeh

from bokeh.plotting import figure, show
from bokeh.io import output_notebook

output_notebook()

x = [1, 2, 3, 4, 5]
y = [4, 5, 5, 7, 2]

default_tools = ['wheel_zoom', 'save', 'pan', 'box_zoom', 'reset']
extra_tools = ['zoom_out', 'zoom_in']

# create a plot
p = figure(
    title="Toolbars",
    tools=extra_tools+default_tools,
    width=400,
    height=300,
)

p.line(x, y)
show(p)

p.toolbar.tools

[ZoomOutTool(id='p1475', ...),
 ZoomInTool(id='p1476', ...),
 WheelZoomTool(id='p1477', ...),
 SaveTool(id='p1478', ...),
 PanTool(id='p1479', ...),
 BoxZoomTool(id='p1480', ...),
 ResetTool(id='p1488', ...)]

So while plot.toolbar.tools respects the set order of the tools, visually it still follows the other way (i.e, ['PanTool', 'BoxZoomTool', 'WheelZoomTool', 'ZoomOutTool', 'ZoomInTool', 'SaveTool', 'ResetTool'])

[Azaya89]

overlay_plot.mov

[hoxbro]

Suggested change

	if isinstance(identifier, list | tuple):
	if isinstance(identifier, list):

Don't use pipe for isinstance and no need to check for tuple.

[hoxbro]

What happens if the name is the same but the icon is not? They should likely not be identified as the same thing.

[Azaya89]

What happens if the name is the same but the icon is not? They should likely not be identified as the same thing.

I've added implementation to identify as different types if there are different disambiguation properties. This now means that something like this will produce multiple PanTool instances:

import holoviews as hv
hv.extension('bokeh')

hv.Curve([1, 2, 3]).opts(tools=[PanTool(description='New Pan')])