Skip to content

PLOT: Add option to specify the plotting backend #26753

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 16 commits into from
Jun 21, 2019
Merged

PLOT: Add option to specify the plotting backend #26753

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
776e82d
WIP/PLOT: Add option to specify the plotting backend
datapythonista Jun 9, 2019
7d89c5e
Merge remote-tracking branch 'upstream/master' into select_backend
Jun 10, 2019
fd36e1a
Moving the validation of the backend to when the backend is selected
Jun 10, 2019
3ae0662
Adding tests, doc and whatsnew
Jun 10, 2019
57f0119
Restoring plotting backend after tests, to not affect other tests
Jun 10, 2019
06e829c
avoid failing tests when matplotlib is not installed
Jun 10, 2019
f5233f3
Merging from master
Jun 11, 2019
f7c6e33
Removing checks to see if plotting backens implement the API
Jun 11, 2019
1095344
Removing tests related to previous commit
Jun 11, 2019
e832985
Merging from master
Jun 20, 2019
b13a74b
Fixing failing test, and unifying get_plot_backend code
Jun 20, 2019
001c57b
Fixing typo in whatsnew
Jun 20, 2019
231094e
Merge remote-tracking branch 'upstream/master' into select_backend
Jun 20, 2019
aa27d34
Merge remote-tracking branch 'upstream/master' into select_backend
datapythonista Jun 21, 2019
f31fc67
Merge branch 'select_backend' of github.com:datapythonista/pandas int…
datapythonista Jun 21, 2019
1f4e1c0
Adding import lost in the merge
datapythonista Jun 21, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 11 additions & 0 deletions pandas/core/config_init.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -460,6 +460,17 @@ def use_inf_as_na_cb(key):
# Plotting
# ---------

plotting_backend_doc = """
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One question: If I wanted to use the altar backend, I would be more like to use .set_option('plotting.backend', 'altair') than ..., 'pdvega'). @jakevdp what name would you prefer?

I think hvplot will just be hvplot, so that's fine.

Anyway, we might consider adding a dict here like plotting_backend_alias that maps the user-facing name like altair to the backend name like pdvega. When the backend library registers themselves, they can also register their aliases.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see your point, and I think it'd add value to the users, but not sure if I'm in favor of adding the extra complexity it's needed to manage aliases in a dynamic way.

I like the simplicity of the parameter being the name of the module. I guess in some cases will look nicer than others. May be hvplot will use hvplot.pandas, since hvplot contains other things besides our plugin, and the module to use may be hvplot.pandas.

In practice I guess backends will register themselves, and users will rarely switch backends manually. But I guess if they do, it'll be better if they know they need to use the name of the module:

import pandas
import hvplot.pandas

df.plot()

pandas.set_option('backend.plotting', 'matplotlib')
df.plot()

pandas.set_option('backend.plotting', 'hvplot.pandas')
df.plot()

I don't have a strong opinion, but I'd say let's start with the simplest option, and add aliases or something else if we think it's useful once we start using this.

Copy link
Contributor

@TomAugspurger TomAugspurger Jun 20, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it especially complex? I was thinking something like

_plotting_aliases = {}  # or define somewhere else

def register_plotting_backend_cb(key):
    backend_str = cf.get_option(key)
    backend_str = _plotting_aliases.get(backend_str, backend_str)
    ...

Indeed, I think this simplifies things already, since we can use 'matplotlib' as pandas.plotting._matplotlib. Though we may continue to special case matplotlib to provide a nice error message.

Copy link
Member Author

@datapythonista datapythonista Jun 20, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What I wouldn't do is to have the aliases in pandas itself. May be I'm being too strict, but if feels wrong.

But you're right, it's probably not as complex as I was thinking anyway. An simple option plotting.aliases with a dictionary may not be ideal, but would allow backends create an alias by simply:

pandas.set_option('plotting.aliases', dict(pandas.get_option('plotting.aliases'), hvplot='hvplot.pandas'))

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

but better in a follow up PR I think, so we can focus there on the exact syntax and approach

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Perfectly fine doing as a followup.

And my thinking may have been a bit muddled here. I was thinking that the backend library would have already been imported, and so would have a chance to register their own aliases. But as you say, it would be pandas managing them, which doesn't feel quite right.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

another simple option is that backends add an optional attribute alias = 'hvplot', and we simply do:

if hasattr(backend_mod, 'alias'):
    plotting_aliases[alias] = backend_mod.__name__

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes good idea. But still leaving this as a followup?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I prefer to keep the focus, the smaller the PRs, the better the content :)

: str
The plotting backend to use. The default value is "matplotlib", the
backend provided with pandas. Other backends can be specified by
prodiving the name of the module that implements the backend.
"""

with cf.config_prefix('plotting'):
cf.register_option('backend', defval='matplotlib',
doc=plotting_backend_doc)

register_converter_doc = """
: bool
Whether to register converters with matplotlib's units registry for
Expand Down
43 changes: 38 additions & 5 deletions pandas/plotting/_core.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
import importlib
from typing import List, Type # noqa

from pandas.util._decorators import Appender
Expand Down Expand Up @@ -625,11 +626,43 @@ def _get_plot_backend():
The backend is imported lazily, as matplotlib is a soft dependency, and
pandas can be used without it being installed.
"""
try:
import pandas.plotting._matplotlib as plot_backend
except ImportError:
raise ImportError("matplotlib is required for plotting.")
return plot_backend
backend_str = pandas.get_option('plotting.backend')
if backend_str == 'matplotlib':
try:
import pandas.plotting._matplotlib as backend_mod
except ImportError:
raise ImportError('matplotlib is required for plotting when the '
'default backend is selected.')
else:
try:
mod = importlib.import_module(backend_str)
except ImportError:
raise ValueError('"{}" does not seem to be an installed module.'
'A pandas plotting backend must be a module that '
'can be imported'.format(backend_str))

required_objs = ['LinePlot', 'BarPlot', 'BarhPlot', 'HistPlot',
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems like we’re being too opinionated about the backends implementation. IMO, we should provide a series and frame .plot accessor that dispatches to the right backend. We can provide an ABC for backends to subclass with the expected user API, but beyond that I don’t think we care.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't have a strong opinion about this at this point. I opened #26747 to have a discussion on what we expect from backends, this can surely be simplified.

At this stage I just tried to be conservative, and force the backends to raise NotImplementedError or whatever they consider if they don't implement something. My idea here is that if the user calls something like Series.plot() or any other functionality, we don't find with a weird error that doesn't provide relevant information for the user.

But happy raise exceptions from our side for some of those when the backend is not the default one, if that's the preferred option.

'BoxPlot', 'KdePlot', 'AreaPlot', 'PiePlot',
'ScatterPlot', 'HexBinPlot', 'hist_series',
'hist_frame', 'boxplot', 'boxplot_frame',
'boxplot_frame_groupby', 'tsplot', 'table',
'andrews_curves', 'autocorrelation_plot',
'bootstrap_plot', 'lag_plot', 'parallel_coordinates',
'radviz', 'scatter_matrix', 'register', 'deregister']
missing_objs = set(required_objs) - set(dir(mod))
if len(missing_objs) == len(required_objs):
raise ValueError(
'"{}" does not seem to be a valid backend. Valid backends are '
'modules that implement the next objects:\n{}'.format(
backend_str, '\n-'.join(required_objs)))
elif missing_objs:
raise ValueError(
'"{}" does not seem to be a complete backend. Valid backends '
'must implement the next objects:\n{}'.format(
backend_str, '\n-'.join(missing_objs)))
else:
backend_mod = importlib.import_module(backend_str)
return backend_mod


def _plot_classes():
Expand Down