Skip to content

Matplotlib 3.3 compatibility fixups #35393

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.

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 11 commits into from
Jul 23, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
2 changes: 1 addition & 1 deletion ci/deps/azure-37-locale.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ dependencies:
- ipython
- jinja2
- lxml
- matplotlib <3.3.0
- matplotlib>=3.3.0
- moto
- nomkl
- numexpr
Expand Down
1 change: 1 addition & 0 deletions doc/source/whatsnew/v1.1.0.rst
Original file line number Diff line number Diff line change
Expand Up @@ -274,6 +274,7 @@ change, as ``fsspec`` will still bring in the same packages as before.
Other enhancements
^^^^^^^^^^^^^^^^^^

- Compatibility with matplotlib 3.3.0 (:issue:`34850`)
- :meth:`IntegerArray.astype` now supports ``datetime64`` dtype (:issue:`32538`)
- :class:`IntegerArray` now implements the ``sum`` operation (:issue:`33172`)
- Added :class:`pandas.errors.InvalidIndexError` (:issue:`34570`).
Expand Down
5 changes: 5 additions & 0 deletions pandas/plotting/_matplotlib/boxplot.py
Original file line number Diff line number Diff line change
Expand Up @@ -299,6 +299,11 @@ def plot_group(keys, values, ax):
if fontsize is not None:
ax.tick_params(axis="both", labelsize=fontsize)
if kwds.get("vert", 1):
ticks = ax.get_xticks()
Copy link
Contributor Author

Choose a reason for hiding this comment

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

This comes from a change in MPL being stricter about set_xticks. This worked for matplotlib 3.2

import pandas as pd
import matplotlib.pyplot as plt
import string
import numpy as np

fig, (ax1, ax2) = plt.subplots(ncols=2, sharex=True)
ax1.boxplot([
    np.array([1, 2, 3, 4]),
    np.array([1, 2, 3, 4])
])
ax2.boxplot([
    np.array([1, 2, 2, 3]),
    np.array([1, 2, 3, 4])
])

ax1.set_xticklabels(['A', 'B'])

We set 2 x ticklabels despite there being 4.

Copy link
Contributor

Choose a reason for hiding this comment

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

Yes, this was a deliberate change... matplotlib/matplotlib#17266 The obvious problem if you only give us 2 tick labels is that its ambiguous which of the 4 ticks you wanted labeled.

MPL now has categorical axes which I think were added because pandas has them. Probably too big a project to homogenize, but....

if len(ticks) != len(keys):
i, remainder = divmod(len(ticks), len(keys))
assert remainder == 0, remainder
keys *= i
ax.set_xticklabels(keys, rotation=rot)
else:
ax.set_yticklabels(keys, rotation=rot)
Expand Down
1 change: 1 addition & 0 deletions pandas/plotting/_matplotlib/compat.py
Original file line number Diff line number Diff line change
Expand Up @@ -21,3 +21,4 @@ def inner():
_mpl_ge_3_0_0 = _mpl_version("3.0.0", operator.ge)
_mpl_ge_3_1_0 = _mpl_version("3.1.0", operator.ge)
_mpl_ge_3_2_0 = _mpl_version("3.2.0", operator.ge)
_mpl_ge_3_3_0 = _mpl_version("3.3.0", operator.ge)
34 changes: 7 additions & 27 deletions pandas/plotting/_matplotlib/converter.py
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,6 @@
from pandas._libs.tslibs.offsets import BaseOffset

from pandas.core.dtypes.common import (
is_datetime64_ns_dtype,
is_float,
is_float_dtype,
is_integer,
Expand Down Expand Up @@ -246,19 +245,6 @@ def get_datevalue(date, freq):
raise ValueError(f"Unrecognizable date '{date}'")


def _dt_to_float_ordinal(dt):
"""
Convert :mod:`datetime` to the Gregorian date as UTC float days,
preserving hours, minutes, seconds and microseconds. Return value
is a :func:`float`.
"""
if isinstance(dt, (np.ndarray, Index, Series)) and is_datetime64_ns_dtype(dt):
base = dates.epoch2num(dt.asi8 / 1.0e9)
else:
base = dates.date2num(dt)
return base


# Datetime Conversion
class DatetimeConverter(dates.DateConverter):
@staticmethod
Expand All @@ -274,15 +260,11 @@ def convert(values, unit, axis):
def _convert_1d(values, unit, axis):
def try_parse(values):
try:
return _dt_to_float_ordinal(tools.to_datetime(values))
return dates.date2num(tools.to_datetime(values))
except Exception:
return values

if isinstance(values, (datetime, pydt.date)):
return _dt_to_float_ordinal(values)
elif isinstance(values, np.datetime64):
return _dt_to_float_ordinal(Timestamp(values))
elif isinstance(values, pydt.time):
if isinstance(values, (datetime, pydt.date, np.datetime64, pydt.time)):
return dates.date2num(values)
elif is_integer(values) or is_float(values):
return values
Expand All @@ -303,12 +285,10 @@ def try_parse(values):

try:
values = tools.to_datetime(values)
if isinstance(values, Index):
values = _dt_to_float_ordinal(values)
else:
values = [_dt_to_float_ordinal(x) for x in values]
except Exception:
values = _dt_to_float_ordinal(values)
pass

values = dates.date2num(values)

return values

Expand Down Expand Up @@ -411,8 +391,8 @@ def __call__(self):
interval = self._get_interval()
freq = f"{interval}L"
tz = self.tz.tzname(None)
st = _from_ordinal(dates.date2num(dmin)) # strip tz
ed = _from_ordinal(dates.date2num(dmax))
st = dmin.replace(tzinfo=None)
Copy link
Contributor Author

Choose a reason for hiding this comment

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

If the comment is correct that this is really about stripping the tz, then this is cleaner since it doesn't rely on matplotlib's Datetime <-> numeric conversion.

ed = dmin.replace(tzinfo=None)
all_dates = date_range(start=st, end=ed, freq=freq, tz=tz).astype(object)

try:
Expand Down
6 changes: 2 additions & 4 deletions pandas/tests/plotting/test_converter.py
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@
pass

pytest.importorskip("matplotlib.pyplot")
dates = pytest.importorskip("matplotlib.dates")


def test_registry_mpl_resets():
Expand Down Expand Up @@ -146,7 +147,7 @@ def test_convert_accepts_unicode(self):

def test_conversion(self):
rs = self.dtc.convert(["2012-1-1"], None, None)[0]
xp = datetime(2012, 1, 1).toordinal()
xp = dates.date2num(datetime(2012, 1, 1))
assert rs == xp

rs = self.dtc.convert("2012-1-1", None, None)
Expand All @@ -155,9 +156,6 @@ def test_conversion(self):
rs = self.dtc.convert(date(2012, 1, 1), None, None)
assert rs == xp

rs = self.dtc.convert(datetime(2012, 1, 1).toordinal(), None, None)
assert rs == xp

rs = self.dtc.convert("2012-1-1", None, None)
assert rs == xp

Expand Down
14 changes: 9 additions & 5 deletions pandas/tests/plotting/test_datetimelike.py
Original file line number Diff line number Diff line change
Expand Up @@ -331,7 +331,7 @@ def test_freq_with_no_period_alias(self):
bts = tm.makeTimeSeries(5).asfreq(freq)
_, ax = self.plt.subplots()
bts.plot(ax=ax)
assert ax.get_lines()[0].get_xydata()[0, 0] == bts.index[0].toordinal()
Copy link
Contributor Author

Choose a reason for hiding this comment

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

This test, and some similar, are invalid now. The relationship between the plotted xy data and ordinals isn't necessarily stable / valuable.


idx = ax.get_lines()[0].get_xdata()
msg = "freq not specified and cannot be inferred"
with pytest.raises(ValueError, match=msg):
Expand Down Expand Up @@ -1279,6 +1279,8 @@ def test_mpl_nopandas(self):
@pytest.mark.slow
def test_irregular_ts_shared_ax_xlim(self):
# GH 2960
from pandas.plotting._matplotlib.converter import DatetimeConverter

ts = tm.makeTimeSeries()[:20]
ts_irregular = ts[[1, 4, 5, 6, 8, 9, 10, 12, 13, 14, 15, 17, 18]]

Expand All @@ -1289,8 +1291,8 @@ def test_irregular_ts_shared_ax_xlim(self):

# check that axis limits are correct
left, right = ax.get_xlim()
assert left <= ts_irregular.index.min().toordinal()
assert right >= ts_irregular.index.max().toordinal()
assert left <= DatetimeConverter.convert(ts_irregular.index.min(), "", ax)
assert right >= DatetimeConverter.convert(ts_irregular.index.max(), "", ax)

@pytest.mark.slow
def test_secondary_y_non_ts_xlim(self):
Expand Down Expand Up @@ -1345,6 +1347,8 @@ def test_secondary_y_mixed_freq_ts_xlim(self):
@pytest.mark.slow
def test_secondary_y_irregular_ts_xlim(self):
# GH 3490 - irregular-timeseries with secondary y
from pandas.plotting._matplotlib.converter import DatetimeConverter

ts = tm.makeTimeSeries()[:20]
ts_irregular = ts[[1, 4, 5, 6, 8, 9, 10, 12, 13, 14, 15, 17, 18]]

Expand All @@ -1356,8 +1360,8 @@ def test_secondary_y_irregular_ts_xlim(self):
ts_irregular[:5].plot(ax=ax)

left, right = ax.get_xlim()
assert left <= ts_irregular.index.min().toordinal()
assert right >= ts_irregular.index.max().toordinal()
assert left <= DatetimeConverter.convert(ts_irregular.index.min(), "", ax)
assert right >= DatetimeConverter.convert(ts_irregular.index.max(), "", ax)

def test_plot_outofbounds_datetime(self):
# 2579 - checking this does not raise
Expand Down
1 change: 1 addition & 0 deletions pandas/tests/plotting/test_frame.py
Original file line number Diff line number Diff line change
Expand Up @@ -1563,6 +1563,7 @@ def test_boxplot(self):
ax.xaxis.get_ticklocs(), np.arange(1, len(numeric_cols) + 1)
)
assert len(ax.lines) == self.bp_n_objects * len(numeric_cols)
tm.close()

axes = series.plot.box(rot=40)
self._check_ticks_props(axes, xrot=40, yrot=0)
Expand Down
10 changes: 7 additions & 3 deletions pandas/tests/plotting/test_series.py
Original file line number Diff line number Diff line change
Expand Up @@ -274,12 +274,14 @@ def test_rotation(self):
self._check_ticks_props(axes, xrot=30)

def test_irregular_datetime(self):
from pandas.plotting._matplotlib.converter import DatetimeConverter

rng = date_range("1/1/2000", "3/1/2000")
rng = rng[[0, 1, 2, 3, 5, 9, 10, 11, 12]]
ser = Series(randn(len(rng)), rng)
_, ax = self.plt.subplots()
ax = ser.plot(ax=ax)
xp = datetime(1999, 1, 1).toordinal()
xp = DatetimeConverter.convert(datetime(1999, 1, 1), "", ax)
ax.set_xlim("1/1/1999", "1/1/2001")
assert xp == ax.get_xlim()[0]

Expand Down Expand Up @@ -684,11 +686,13 @@ def test_kind_both_ways(self):
kinds = (
plotting.PlotAccessor._common_kinds + plotting.PlotAccessor._series_kinds
)
_, ax = self.plt.subplots()
for kind in kinds:

_, ax = self.plt.subplots()
Copy link
Contributor Author

Choose a reason for hiding this comment

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

Some strange interaction / issue when plotting on existing axes. IIRC it's from MPL being stricter about things. I didn't investigate too much.

Copy link
Contributor

Choose a reason for hiding this comment

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

Its preferable to keep the reference to the figure and axes and then reuse, rather than depend on the implicit ability of plt.subplots() to pick up the same axes...

s.plot(kind=kind, ax=ax)
self.plt.close()
_, ax = self.plt.subplots()
getattr(s.plot, kind)()
self.plt.close()

@pytest.mark.slow
def test_invalid_plot_data(self):
Expand Down