Indicate color clipping and control clipping colors #686

philippjfr · 2016-05-12T19:46:53Z

As the title says, adds a plot option to extend the colorbar in a particular. Here's an example:

%%opts Image [colorbar=True]
hv.Layout([hv.Image(np.random.rand(10,10))(plot=dict(cbar_extend=opt))
           for opt in ['min', 'max', 'both']])

Can also consider automatically enabling this if your data falls outside the specified range but that's another discussion.

jbednar · 2016-05-12T20:03:33Z

I don't get it -- what does it mean to extend the colorbar? It doesn't seem extended, it just now has a pointy end. Do you mean, make the colorbar visually indicate that it extends beyond the area shown, without actually changing the colorbar or the colormapping? If so that's not what the option name or the name of this issue suggests, and seems very confusing.

philippjfr · 2016-05-12T20:04:35Z

True, the name isn't great, it's what matplotlib uses.

philippjfr · 2016-05-12T20:34:54Z

This would also be more useful with an option to supply different colors for values that are above and below the range (and for NaN values). We can brainstorm about a nice way to specify that.

jlstevens · 2016-05-12T20:46:09Z

pointy_cbar sounds more descriptive to me! ;-p

jbednar · 2016-05-12T20:50:42Z

I'm not sure it makes sense to provide direct control over this option, in HoloViews. Seems like HoloViews should just automatically turn it on if the user has selected a colormap that ends up clipping at a particular end of the colorbar, so that the visual representation is a reliable indicator of something meaningful. As it is, there's no relation between whether the user has explicitly enabled a pointy colorbar and anything about the actual plot, so indeed Jean-Luc's suggestion of calling it pointy_cbar is accurate!

jlstevens · 2016-05-12T20:54:43Z

Using this to indicate clipping when clipping occurs does make a lot more sense. We would still need a parameter to turn this behaviour on and off though...

jlstevens · 2016-05-12T20:56:28Z

Also I assume one problem this this suggestion is we might not know if clipping has occurred without inspecting the data range first? Sounds tricky...you might have to compare the data min/max with the limits to know if it should be pointy or not.

philippjfr · 2016-05-12T21:07:53Z

Sounds tricky...you might have to compare the data min/max with the limits to know if it should be pointy or not.

It might be fairly straightforward and the overhead wouldn't be huge, so might be worth doing. Coupled with a way to specify special colors for min, max and NaN values that would be a nice addition.

jlstevens · 2016-05-12T21:26:41Z

It might be fairly straightforward and the overhead wouldn't be huge..

Sure. As long as there is a way to turn it off and keep the current behaviour in case it ever becomes a problem. I agree it would be a nice feature and I'm wondering if there might be a way to specify the NaN values and clipping colors at the same time...

jbednar · 2016-05-12T22:32:10Z

That makes sense: the user can optionally enable having pointy cbars indicate such clipping. That way, whenever there is such a pointy cbar, it does accurately reflect something about the current plot, rather than just being some meaningless style option. Yet those who don't ever want pointy cbars can turn them off.

philippjfr · 2016-05-13T10:14:39Z

So since I actually needed this for some plots of mine I've gone ahead and implemented it. If you've got a better suggestion for the clipping_colors parameter please suggest it and tell me and I also still need a name for the option that turns the clipping indicators on and off in general.

jlstevens · 2016-05-13T10:27:32Z

Apparently ('w', 1) is used because the colorbar code in matplotlib takes the color and a separate alpha. I would much prefer to pretend matplotlib is consistent and support standard matplotlib colors, including the RGBA tuples.

philippjfr · 2016-05-13T10:28:35Z

Sure, shouldn't be hard.

jlstevens · 2016-05-13T10:29:08Z

holoviews/plotting/mpl/element.py

@@ -497,8 +497,10 @@ class ColorbarPlot(ElementPlot):
    colorbar = param.Boolean(default=False, doc="""
        Whether to draw a colorbar.""")

-    cbar_width = param.Number(default=0.05, doc="""
-        Width of the colorbar as a fraction of the main plot""")
+    clipping_colors = param.Dict(default={'NaN': ('w', 1)}, doc="""


I understand what I expect to happen to NaN values but an example of how to set the min and max colors would be nice (in the docstring).

jlstevens · 2016-05-13T10:33:23Z

Philipp tells me you would specify a max clipping like this:

{'min': 'r', 'Nan': 'w'}

Would setting these colors make the colorbar pointy as appropriate (if the data is clipped?). If so wouldn't an empty dictionary correspond to the current behaviour without pointy colorbars?

philippjfr · 2016-05-13T10:34:31Z

If so wouldn't an empty dictionary correspond to the current behaviour without pointy colorbars?

No, not specifying a special color just sets the min and max color in the cmap as the clipping color, as in the example above.

jlstevens · 2016-05-13T11:01:59Z

So why not have something like:

cbar_extension = {'min': 'default', 'max':'default'}

To indicate pointy colorbars with the min/max of the colormap where you can then specify your own custom colors instead of 'default'?

philippjfr · 2016-05-13T11:03:17Z

Sounds good, if we can come up with a better name for the parameter.

jbednar · 2016-05-13T14:18:45Z

Maybe cbar_clip?

jlstevens · 2016-05-13T17:36:57Z

Given that colorbar is a parameter name, colorbar_clip would appear next to it. For that reason I might like the longer name.

jbednar · 2016-05-13T17:39:00Z

Sure, if "clip" is accurate as a description about what it does (which I've somewhat lost track of by this point. :-)

philippjfr · 2016-05-13T17:39:02Z

We already have cbar_width and cbar_padding so it should at least be consistent.

Edit: Of course we could decide to rename those as well.

jlstevens · 2016-05-13T18:06:53Z

We already have cbar_width and cbar_padding so it should at least be consistent.

In that case we should alias colorbar with cbar to enable/disable color bars.

philippjfr · 2016-05-13T18:08:26Z

I'd suggest just calling it clipping_colors or color_clipping. Adding the pointy bit to the color bar is just a secondary side effect. Even without the colorbar enabled this will control the out-of-range colors so it's not really a colorbar option.

Edit: I guess why I was hesitant about the {'max': 'default'} suggestion is for the same reason, this parameter doesn't necessarily have anything to do with the colorbar, so enabling it that way feels a bit weird.

jlstevens · 2016-05-14T10:34:01Z

clipping_colors seems fine. So could you still use it to disable pointy colorbars?

jbednar · 2016-05-24T17:58:10Z

See #689 for proposed UI.

philippjfr · 2016-10-07T14:59:19Z

Ready to merge once clipping color support is fixed in bokeh: bokeh/bokeh#5323

jlstevens · 2016-10-07T15:24:42Z

holoviews/plotting/bokeh/element.py

@@ -748,6 +748,11 @@ class ColorbarPlot(ElementPlot):
        location, orientation, height, width, scale_alpha, title, title_props,
        margin, padding, background_fill_color and more.""")

+    clipping_colors = param.Dict(default={'NaN': (0, 0, 0, 1)}, doc="""
+        Dictionary to specify colors for clipped values, allows setting


RGBA tuples are fine though are any other color specifications allowed? E.g hex strings?

Yes, docstring should be updated to reflect that.

jlstevens · 2016-10-12T20:23:48Z

Looks good. Happy to merge when the tests pass.

jlstevens · 2016-10-12T20:37:13Z

Tests have all passed. Merging.

philippjfr added the in progress label May 12, 2016

philippjfr force-pushed the cbar_extend branch from 717261d to 5c7b351 Compare May 13, 2016 10:12

philippjfr changed the title ~~Added plot option to extend the colorbar in matplotlib~~ Indicate color clipping and control clipping colors in matplotlib May 13, 2016

philippjfr force-pushed the cbar_extend branch from 5c7b351 to df0e0d5 Compare May 13, 2016 10:15

jlstevens reviewed May 13, 2016
View reviewed changes

philippjfr force-pushed the cbar_extend branch 2 times, most recently from 25945f8 to 1bc85cf Compare May 13, 2016 10:45

philippjfr changed the title ~~Indicate color clipping and control clipping colors in matplotlib~~ Indicate color clipping and control clipping colors Oct 7, 2016

jlstevens reviewed Oct 7, 2016

View reviewed changes

philippjfr added 3 commits October 12, 2016 19:30

Detecting color clipping and added control over clipping_colors

dc24ecf

Added support for clipping colors for bokeh and mpl

7c15969

Do not set clipping_colors by default

eebdaad

philippjfr force-pushed the cbar_extend branch from e09e0b2 to eebdaad Compare October 12, 2016 18:30

Made clipping_color specification match between backends

9b3fc6a

philippjfr force-pushed the cbar_extend branch from 4db8ea6 to 9b3fc6a Compare October 12, 2016 19:01

philippjfr added 2 commits October 12, 2016 20:23

Fixed matplotlib colormap picking

d26a140

Restored default HeatMapPlot nan color to white

b306613

philippjfr force-pushed the cbar_extend branch from 94735b4 to b306613 Compare October 12, 2016 20:16

jlstevens merged commit 3c29b0b into master Oct 12, 2016

philippjfr deleted the cbar_extend branch October 14, 2016 01:55

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Indicate color clipping and control clipping colors #686

Indicate color clipping and control clipping colors #686

philippjfr commented May 12, 2016 •

edited

Loading

jbednar commented May 12, 2016

philippjfr commented May 12, 2016

philippjfr commented May 12, 2016 •

edited

Loading

jlstevens commented May 12, 2016

jbednar commented May 12, 2016

jlstevens commented May 12, 2016

jlstevens commented May 12, 2016

philippjfr commented May 12, 2016 •

edited

Loading

jlstevens commented May 12, 2016

jbednar commented May 12, 2016 •

edited

Loading

philippjfr commented May 13, 2016

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016

jlstevens May 13, 2016

philippjfr May 13, 2016

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016 •

edited

Loading

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016

jbednar commented May 13, 2016

jlstevens commented May 13, 2016

jbednar commented May 13, 2016

philippjfr commented May 13, 2016 •

edited

Loading

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016 •

edited

Loading

jlstevens commented May 14, 2016

jbednar commented May 24, 2016

philippjfr commented Oct 7, 2016

jlstevens Oct 7, 2016

philippjfr Oct 7, 2016

jlstevens commented Oct 12, 2016

jlstevens commented Oct 12, 2016

Indicate color clipping and control clipping colors #686

Indicate color clipping and control clipping colors #686

Conversation

philippjfr commented May 12, 2016 • edited Loading

jbednar commented May 12, 2016

philippjfr commented May 12, 2016

philippjfr commented May 12, 2016 • edited Loading

jlstevens commented May 12, 2016

jbednar commented May 12, 2016

jlstevens commented May 12, 2016

jlstevens commented May 12, 2016

philippjfr commented May 12, 2016 • edited Loading

jlstevens commented May 12, 2016

jbednar commented May 12, 2016 • edited Loading

philippjfr commented May 13, 2016

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016

jlstevens May 13, 2016

Choose a reason for hiding this comment

philippjfr May 13, 2016

Choose a reason for hiding this comment

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016 • edited Loading

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016

jbednar commented May 13, 2016

jlstevens commented May 13, 2016

jbednar commented May 13, 2016

philippjfr commented May 13, 2016 • edited Loading

jlstevens commented May 13, 2016

philippjfr commented May 13, 2016 • edited Loading

jlstevens commented May 14, 2016

jbednar commented May 24, 2016

philippjfr commented Oct 7, 2016

jlstevens Oct 7, 2016

Choose a reason for hiding this comment

philippjfr Oct 7, 2016

Choose a reason for hiding this comment

jlstevens commented Oct 12, 2016

jlstevens commented Oct 12, 2016

philippjfr commented May 12, 2016 •

edited

Loading

philippjfr commented May 12, 2016 •

edited

Loading

philippjfr commented May 12, 2016 •

edited

Loading

jbednar commented May 12, 2016 •

edited

Loading

philippjfr commented May 13, 2016 •

edited

Loading

philippjfr commented May 13, 2016 •

edited

Loading

philippjfr commented May 13, 2016 •

edited

Loading