PERF: pandas' import time #16764
Labels
Performance
Memory or execution speed performance
Milestone
Comments
|
do these in independent processes |
|
Oh, right - |
|
or better yet import numpy and matplotlib first |
|
We do attempt to import matplotlib at import time. We could delay that with something like diff --git a/pandas/plotting/__init__.py b/pandas/plotting/__init__.py
index c3cbedb0f..8f98e297e 100644
--- a/pandas/plotting/__init__.py
+++ b/pandas/plotting/__init__.py
@@ -4,12 +4,6 @@ Plotting api
# flake8: noqa
-try: # mpl optional
- from pandas.plotting import _converter
- _converter.register() # needs to override so set_xlim works with str/number
-except ImportError:
- pass
-
from pandas.plotting._misc import (scatter_matrix, radviz,
andrews_curves, bootstrap_plot,
parallel_coordinates, lag_plot,
diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index 391fa377f..9821c89c4 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -37,12 +37,7 @@ from pandas.plotting._tools import (_subplots, _flatten, table,
_get_xlim, _set_ticks_props,
format_date_labels)
-
-if _mpl_ge_1_5_0():
- # Compat with mp 1.5, which uses cycler.
- import cycler
- colors = mpl_stylesheet.pop('axes.color_cycle')
- mpl_stylesheet['axes.prop_cycle'] = cycler.cycler('color', colors)
+_registered = False
def _get_standard_kind(kind):
@@ -92,6 +87,7 @@ class MPLPlot(object):
secondary_y=False, colormap=None,
table=False, layout=None, **kwds):
+ self._setup()
self.data = data
self.by = by
@@ -175,6 +171,20 @@ class MPLPlot(object):
self._validate_color_args()
+ def _setup(self):
+ global _registered
+ if not _registered:
+ from pandas.plotting import _converter
+ _converter.register()
+
+ if _mpl_ge_1_5_0():
+ # Compat with mp 1.5, which uses cycler.
+ import cycler
+ colors = mpl_stylesheet.pop('axes.color_cycle')
+ mpl_stylesheet['axes.prop_cycle'] = cycler.cycler('color', colors)
+
+ _registered = True
+
def _validate_color_args(self):
if 'color' not in self.kwds and 'colors' in self.kwds:
warnings.warn(("'colors' is being deprecated. Please use 'color'"That covers all the |
|
Looks like |
|
Oh, sorry, I was thinking of |
|
I did a profile with https://github.com/cournape/import-profiler From a very quick skim:
Not huge, but those two would already remove ca 20% of the import time. However I don't see Full output: |
|
Also see #7282, but seems like already more attention here. |
It's a bit different issue, this is in general about reducing import time, the other issue is about a specific case where the import takes many seconds (but also numpy takes seconds to import, so IMO it's not pandas specific issue) |
Merged
ghost
pushed a commit
to reef-technologies/pandas
that referenced
this issue
Oct 16, 2017
alanbato
pushed a commit
to alanbato/pandas
that referenced
this issue
Nov 10, 2017
No-Stream
pushed a commit
to No-Stream/pandas
that referenced
this issue
Nov 28, 2017
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I wouldn't normally be concerned about this, as of it course it only happens once, but our import time has gotten quite long, to the point I notice it hanging my
ipythonstartup.I don't have a good sense of what would be required to improve this, probably deferring more imports to be just in time?
on
0.20.2- each import in a separate processBelow is a single process, importing deps first
The text was updated successfully, but these errors were encountered: