Port to plotly.py v4_subplots and renderers #88

jonmmease · 2019-05-19T23:29:00Z

Overview

This PR updates plotly_express to use the plotly.py plotly.subplots.make_subplots functionality to build faceted figures, rather than the current layout.grid approach. By building on plotly.subplots.make_subplots, it will be much easier for users to modify faceted figures in plotly.py.

It also removes ExpressFigure and relies on the plotly.py plotly.io.renderers configuration to display figures.

Because this PR requires plotly.py to be imported with the v4_subplots and renderer_defaults future flags. If plotly_express is imported first then this is done automatically, but if plotly.py is import first, without these future flags, then an error will be raised when plotly_express is imported.

Examples

Update a y-axis based on row

import plotly_express as px
iris = px.data.iris()
fig = (px.scatter(iris, x="sepal_width", y="sepal_length", facet_row='species', range_x=[0, 5], log_y=True))
fig.update_yaxes(title_font_color='red', row=3)

Update a trace based on row

fig = (px.scatter(iris, x="sepal_width", y="sepal_length", facet_row='species', range_x=[0, 5], log_y=True))
fig.update_traces(marker_color='red', row=2)

Marginals and Faceting

Support for combining marginals and faceting

px.scatter(iris, x="sepal_width", y="sepal_length", facet_row='species', marginal_y='histogram')

fig = px.scatter(tips, x="total_bill", y="tip", facet_row="time", facet_col="day", # color="smoker",
                 category_orders={"day": ["Thur", "Fri", "Sat", "Sun"], "time": ["Lunch", "Dinner"]},
                 marginal_y='rug', marginal_x='histogram', color='sex'
                )

fig.update(layout_height=800)

Faceting non-cartesian trace types

This PR doesn't add facet_row/facet_col to any additional plot types, but when these args are added, faceting will "work" automatically. I didn't add it in this PR because the default spacing for other trace types can look pretty bad, so we'll need to work through how these other subplot types shoud look with faceting.

cc @nicolaskruchten

plotly_express/_core.py

…resent

plotly_express/_core.py

nicolaskruchten · 2019-05-21T02:08:40Z

plotly_express/_core.py

+                showgrid=args["marginal_x"] == 'histogram',
+                row=row
+            )
+            fig.update_xaxes(


has this been run through black ? :)

Thanks! I know the plotly.py repo doesn't have that in place but this one does for now ;)

Yeah, I'm interested in using black in plotly.py too. Have you looked into the git integration https://black.readthedocs.io/en/stable/version_control_integration.html?

nicolaskruchten · 2019-05-21T02:17:45Z

plotly_express/_core.py

+
+                # Find row for trace, handling facet_row and marginal_x
+                if m.facet == 'row':
+                    if has_marginal_x:


Is this case for marginals + facets? that's not supported right now in the grouping logic.

Yes. With these changes I think marginals + facets are working. See the screenshot in the description above.

aha I see, that's interesting. Cool that it works!

I think that for combining them, I was envisioning per-row and per-column marginals rather than per-subplot marginals, which isn't quite as straightforward (although not impossible, just requires a second group-by pass). Thoughts?

I liked per subplot marginals, but I suppose per row/col could useful to. Would it make sense for this to be the default and then later add an option to change the behavior?

nicolaskruchten · 2019-05-21T16:53:54Z

I didn't add it in this PR because the default spacing for other trace types can look pretty bad, so we'll need to work through how these other subplot types shoud look with faceting.

can I see an example? Is the spacing too big or too small?

jonmmease · 2019-05-22T12:14:08Z

Sure, here are some examples:

I see two issues:

The first is that different subplot types will need different default padding. This should just be a matter of taking the time to work through all of the subplot types.
The second is that subplot types with a fixed aspect ratio (ternary and polar) are centered in their domain, so the row/column labels don't line up well. I'm not really sure what to do about this.

# Conflicts: # plotly_express/_core.py

If faceting and marginal specified then faceting wins. It's still possible to facet columns and have marginals per columns or to facet rows and have marginals per row.

jonmmease · 2019-06-13T17:17:38Z

@nicolaskruchten, I just updated the PR with master and removed the marginals per facet.

If faceting and marginal specified then faceting wins. It's still
possible to facet columns and have marginals per columns or to facet rows
and have marginals per row.

e.g.

If this looks alright, I'll start pulling it into the v4_integration branch of plotly.py

nicolaskruchten · 2019-06-13T17:36:55Z

plotly_express/_core.py

 )
-TraceSpec = namedtuple("TraceSpec", ["constructor", "attrs", "trace_patch"])
+TraceSpec = namedtuple("TraceSpec", ["constructor", "attrs", "trace_patch", "marginal"])


This is a good idea, and maybe should be generalized later to model the "role" of the trace in the figure... Is it the main event, or a marginal, or a trendline, or some other future thing like the raw scatter that someone might want to add to a density heatmap?

nicolaskruchten · 2019-06-13T17:37:38Z

plotly_express/_core.py

    log_key = "log_" + letter
    range_key = "range_" + letter
    if log_key in args and args[log_key]:
-        layout[axis]["type"] = "log"
+        axis["type"] = "log"


if axis is a go then it would be nicer to have axis.type here right?

(not necessary, just noting)

nicolaskruchten · 2019-06-13T17:40:34Z

plotly_express/_core.py

@@ -38,25 +45,6 @@ def set_mapbox_access_token(token):
    MAPBOX_TOKEN = token


-class ExpressFigure(go.Figure):


should grep the docstrings for ExpressFigure and change that out :)

nicolaskruchten · 2019-06-13T17:48:12Z

This looks great! Having marginals vs facets work this way on a per-direction basis is really 👍 !!

nicolaskruchten · 2019-06-13T17:48:26Z

Let's pull this into v4

jonmmease · 2019-06-14T16:04:52Z

Thanks @nicolaskruchten, superseded by plotly/plotly.py#1613

Port to plotly.py v4_subplots and renderers

78f058f

jonmmease commented May 20, 2019

View reviewed changes

plotly_express/_core.py Show resolved Hide resolved

jonmmease commented May 20, 2019

View reviewed changes

plotly_express/_core.py Show resolved Hide resolved

jonmmease added 2 commits May 20, 2019 18:22

Remove ExpressFigure

82e1910

Fix marginal grid configuration when only one of x or y marginal is p…

f1e4b89

…resent

jonmmease changed the title ~~[WIP] Port to plotly.py v4_subplots and renderers~~ Port to plotly.py v4_subplots and renderers May 20, 2019

jonmmease requested a review from nicolaskruchten May 20, 2019 22:36

nicolaskruchten reviewed May 21, 2019

View reviewed changes

plotly_express/_core.py Show resolved Hide resolved

nicolaskruchten reviewed May 21, 2019

View reviewed changes

Black format files

a313ce5

jonmmease added 2 commits June 13, 2019 11:43

Merge branch 'master' into v4_subplots

e278265

# Conflicts: # plotly_express/_core.py

Remove support for marginals inside facets.

1f639bf

If faceting and marginal specified then faceting wins. It's still possible to facet columns and have marginals per columns or to facet rows and have marginals per row.

nicolaskruchten reviewed Jun 13, 2019

View reviewed changes

jonmmease mentioned this pull request Jun 14, 2019

plotly_express integration plotly/plotly.py#1613

Merged

jonmmease closed this Jun 14, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Port to plotly.py v4_subplots and renderers #88

Port to plotly.py v4_subplots and renderers #88

jonmmease commented May 19, 2019 •

edited

Loading

nicolaskruchten May 21, 2019

jonmmease May 21, 2019

nicolaskruchten May 21, 2019

jonmmease May 22, 2019

nicolaskruchten May 21, 2019

jonmmease May 21, 2019

nicolaskruchten May 21, 2019

jonmmease May 22, 2019

nicolaskruchten commented May 21, 2019

jonmmease commented May 22, 2019

jonmmease commented Jun 13, 2019

nicolaskruchten Jun 13, 2019

nicolaskruchten Jun 13, 2019

nicolaskruchten Jun 13, 2019

nicolaskruchten Jun 13, 2019

nicolaskruchten commented Jun 13, 2019

nicolaskruchten commented Jun 13, 2019

jonmmease commented Jun 14, 2019

		@@ -38,25 +45,6 @@ def set_mapbox_access_token(token):
		MAPBOX_TOKEN = token


		class ExpressFigure(go.Figure):

Port to plotly.py v4_subplots and renderers #88

Port to plotly.py v4_subplots and renderers #88

Conversation

jonmmease commented May 19, 2019 • edited Loading

Overview

Examples

Marginals and Faceting

Faceting non-cartesian trace types

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nicolaskruchten commented May 21, 2019

jonmmease commented May 22, 2019

jonmmease commented Jun 13, 2019

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nicolaskruchten commented Jun 13, 2019

nicolaskruchten commented Jun 13, 2019

jonmmease commented Jun 14, 2019

jonmmease commented May 19, 2019 •

edited

Loading