From 9922e7ffdcc41ea593593b75e711f06b2309741a Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Sun, 16 May 2021 13:48:47 +0200
Subject: [PATCH 01/10] PERF: load plotting entrypoint only when necessary

---
 pandas/plotting/_core.py | 35 +++++++++++++++++------------------
 1 file changed, 17 insertions(+), 18 deletions(-)

diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index 27f8835968b54..5c912a8865f44 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -865,7 +865,7 @@ def _get_call_args(backend_name, data, args, kwargs):
         if args and isinstance(data, ABCSeries):
             positional_args = str(args)[1:-1]
             keyword_args = ", ".join(
-                f"{name}={repr(value)}" for (name, default), value in zip(arg_def, args)
+                f"{name}={repr(value)}" for (name, _), value in zip(arg_def, args)
             )
             msg = (
                 "`Series.plot()` should not be called with positional "
@@ -876,7 +876,7 @@ def _get_call_args(backend_name, data, args, kwargs):
             )
             raise TypeError(msg)
 
-        pos_args = {name: value for value, (name, _) in zip(args, arg_def)}
+        pos_args = {name: value for (name, _), value in zip(arg_def, args)}
         if backend_name == "pandas.plotting._matplotlib":
             kwargs = dict(arg_def, **pos_args, **kwargs)
         else:
@@ -1729,7 +1729,7 @@ def hexbin(self, x, y, C=None, reduce_C_function=None, gridsize=None, **kwargs):
 
 def _find_backend(backend: str):
     """
-    Find a pandas plotting backend>
+    Find a pandas plotting backend.
 
     Parameters
     ----------
@@ -1749,11 +1749,8 @@ def _find_backend(backend: str):
     import pkg_resources  # Delay import for performance.
 
     for entry_point in pkg_resources.iter_entry_points("pandas_plotting_backends"):
-        if entry_point.name == "matplotlib":
-            # matplotlib is an optional dependency. When
-            # missing, this would raise.
-            continue
-        _backends[entry_point.name] = entry_point.load()
+        if entry_point.name == backend:
+            _backends[entry_point.name] = entry_point.load()
 
     try:
         return _backends[backend]
@@ -1778,21 +1775,26 @@ def _find_backend(backend: str):
     )
 
 
-def _get_plot_backend(backend=None):
+def _get_plot_backend(backend: str | None = None):
     """
     Return the plotting backend to use (e.g. `pandas.plotting._matplotlib`).
 
-    The plotting system of pandas has been using matplotlib, but the idea here
-    is that it can also work with other third-party backends. In the future,
-    this function will return the backend from a pandas option, and all the
-    rest of the code in this file will use the backend specified there for the
-    plotting.
+    The plotting system of pandas uses matplotlib by default, but the idea here
+    is that it can also work with other third-party backends. This function
+    returns the module which provides a top-level `.plot` method that will
+    actually do the plotting. The backend is specified from a string, which
+    either comes from the keyword argument `backend`, or, if not specified, from
+    the option `pandas.options.plotting.backend`. All the rest of the code in
+    this file uses the backend specified there for the plotting.
 
     The backend is imported lazily, as matplotlib is a soft dependency, and
     pandas can be used without it being installed.
     """
     backend = backend or get_option("plotting.backend")
 
+    if backend in _backends:
+        return _backends[backend]
+
     if backend == "matplotlib":
         # Because matplotlib is an optional dependency and first-party backend,
         # we need to attempt an import here to raise an ImportError if needed.
@@ -1805,10 +1807,7 @@ def _get_plot_backend(backend=None):
             ) from None
 
         _backends["matplotlib"] = module
-
-    if backend in _backends:
-        return _backends[backend]
+        return module
 
     module = _find_backend(backend)
-    _backends[backend] = module
     return module

From 69c9206ff4f7ee72f0399c5e1473146261f7c850 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Mon, 17 May 2021 10:50:03 +0200
Subject: [PATCH 02/10] Better separate backend loading.

Introduces `_load_backend` instead of `_find_backend`, which has a
better defined role, and can also handle backend='matplotlib'
---
 pandas/plotting/_core.py | 80 ++++++++++++++++++++--------------------
 1 file changed, 39 insertions(+), 41 deletions(-)

diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index 5c912a8865f44..a75126ffe4cef 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -1727,15 +1727,15 @@ def hexbin(self, x, y, C=None, reduce_C_function=None, gridsize=None, **kwargs):
 _backends = {}
 
 
-def _find_backend(backend: str):
+def _load_backend(backend: str):
     """
-    Find a pandas plotting backend.
+    Load a pandas plotting backend.
 
     Parameters
     ----------
     backend : str
         The identifier for the backend. Either an entrypoint item registered
-        with pkg_resources, or a module name.
+        with pkg_resources, "matplotlib", or a module name.
 
     Notes
     -----
@@ -1746,32 +1746,45 @@ def _find_backend(backend: str):
     types.ModuleType
         The imported backend.
     """
-    import pkg_resources  # Delay import for performance.
-
-    for entry_point in pkg_resources.iter_entry_points("pandas_plotting_backends"):
-        if entry_point.name == backend:
-            _backends[entry_point.name] = entry_point.load()
-
-    try:
-        return _backends[backend]
-    except KeyError:
-        # Fall back to unregistered, module name approach.
+    if backend == "matplotlib":
+        # Because matplotlib is an optional dependency and first-party backend,
+        # we need to attempt an import here to raise an ImportError if needed.
         try:
-            module = importlib.import_module(backend)
+            import pandas.plotting._matplotlib as module
         except ImportError:
-            # We re-raise later on.
-            pass
-        else:
-            if hasattr(module, "plot"):
-                # Validate that the interface is implemented when the option
-                # is set, rather than at plot time.
-                _backends[backend] = module
-                return module
+            raise ImportError(
+                "matplotlib is required for plotting when the "
+                'default backend "matplotlib" is selected.'
+            ) from None
+
+    else:
+        module = None
+        # Delay import for performance.
+        # TODO: replace with `importlib.metadata` when python_requires >= 3.8.
+        from pkg_resources import iter_entry_points
+
+        for entry_point in iter_entry_points("pandas_plotting_backends"):
+            if entry_point.name == backend:
+                module = entry_point.load()
+
+        if module is None:
+            # Fall back to unregistered, module name approach.
+            try:
+                module = importlib.import_module(backend)
+            except ImportError:
+                # We re-raise later on.
+                pass
+
+    if hasattr(module, "plot"):
+        # Validate that the interface is implemented when the option is set,
+        # rather than at plot time.
+        _backends[backend] = module
+        return module
 
     raise ValueError(
-        f"Could not find plotting backend '{backend}'. Ensure that you've installed "
-        f"the package providing the '{backend}' entrypoint, or that the package has a "
-        "top-level `.plot` method."
+        f"Could not find plotting backend '{backend}'. Ensure that you've "
+        f"installed the package providing the '{backend}' entrypoint, or that "
+        "the package has a top-level `.plot` method."
     )
 
 
@@ -1795,19 +1808,4 @@ def _get_plot_backend(backend: str | None = None):
     if backend in _backends:
         return _backends[backend]
 
-    if backend == "matplotlib":
-        # Because matplotlib is an optional dependency and first-party backend,
-        # we need to attempt an import here to raise an ImportError if needed.
-        try:
-            import pandas.plotting._matplotlib as module
-        except ImportError:
-            raise ImportError(
-                "matplotlib is required for plotting when the "
-                'default backend "matplotlib" is selected.'
-            ) from None
-
-        _backends["matplotlib"] = module
-        return module
-
-    module = _find_backend(backend)
-    return module
+    return _load_backend(backend)

From 8315b388c0e07320915248936bf412675a55f257 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Mon, 17 May 2021 11:02:31 +0200
Subject: [PATCH 03/10] Update whatsnew

---
 doc/source/whatsnew/v1.3.0.rst | 1 +
 1 file changed, 1 insertion(+)

diff --git a/doc/source/whatsnew/v1.3.0.rst b/doc/source/whatsnew/v1.3.0.rst
index 409125b6d6691..a8ae36a1b17db 100644
--- a/doc/source/whatsnew/v1.3.0.rst
+++ b/doc/source/whatsnew/v1.3.0.rst
@@ -812,6 +812,7 @@ Performance improvements
 - Performance improvement in :meth:`.GroupBy.cummin` and :meth:`.GroupBy.cummax` with nullable data types (:issue:`37493`)
 - Performance improvement in :meth:`Series.nunique` with nan values (:issue:`40865`)
 - Performance improvement in :meth:`DataFrame.transpose`, :meth:`Series.unstack` with ``DatetimeTZDtype`` (:issue:`40149`)
+- Performance improvement in :meth:`Series.plot` and :meth:`DataFrame.plot` with entry point lazy loading (:issue:`41492`)
 
 .. ---------------------------------------------------------------------------
 

From 23d059e811c4881731ed82add8b93993e131c768 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Mon, 17 May 2021 15:59:03 +0200
Subject: [PATCH 04/10] Mutate _backends in_get_plot_backend import top

---
 pandas/plotting/_core.py | 43 ++++++++++++++++++++--------------------
 1 file changed, 22 insertions(+), 21 deletions(-)

diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index a75126ffe4cef..d8fe99c2c97ac 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -6,6 +6,9 @@
     Sequence,
 )
 
+# TODO: replace with `importlib.metadata` when python_requires >= 3.8.
+import pkg_resources
+
 from pandas._config import get_option
 
 from pandas._typing import IndexLabel
@@ -1737,10 +1740,6 @@ def _load_backend(backend: str):
         The identifier for the backend. Either an entrypoint item registered
         with pkg_resources, "matplotlib", or a module name.
 
-    Notes
-    -----
-    Modifies _backends with imported backends as a side effect.
-
     Returns
     -------
     types.ModuleType
@@ -1756,29 +1755,25 @@ def _load_backend(backend: str):
                 "matplotlib is required for plotting when the "
                 'default backend "matplotlib" is selected.'
             ) from None
+        return module
 
-    else:
-        module = None
-        # Delay import for performance.
-        # TODO: replace with `importlib.metadata` when python_requires >= 3.8.
-        from pkg_resources import iter_entry_points
+    module = None
 
-        for entry_point in iter_entry_points("pandas_plotting_backends"):
-            if entry_point.name == backend:
-                module = entry_point.load()
+    for entry_point in pkg_resources.iter_entry_points("pandas_plotting_backends"):
+        if entry_point.name == backend:
+            module = entry_point.load()
 
-        if module is None:
-            # Fall back to unregistered, module name approach.
-            try:
-                module = importlib.import_module(backend)
-            except ImportError:
-                # We re-raise later on.
-                pass
+    if module is None:
+        # Fall back to unregistered, module name approach.
+        try:
+            module = importlib.import_module(backend)
+        except ImportError:
+            # We re-raise later on.
+            pass
 
     if hasattr(module, "plot"):
         # Validate that the interface is implemented when the option is set,
         # rather than at plot time.
-        _backends[backend] = module
         return module
 
     raise ValueError(
@@ -1802,10 +1797,16 @@ def _get_plot_backend(backend: str | None = None):
 
     The backend is imported lazily, as matplotlib is a soft dependency, and
     pandas can be used without it being installed.
+
+    Notes
+    -----
+    Modifies `_backends` with imported backend as a side effect.
     """
     backend = backend or get_option("plotting.backend")
 
     if backend in _backends:
         return _backends[backend]
 
-    return _load_backend(backend)
+    module = _load_backend(backend)
+    _backends[backend] = module
+    return module

From a9f98991e75e1e2091e708a3d863ceeab9e23ed2 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Tue, 18 May 2021 13:07:57 +0200
Subject: [PATCH 05/10] Fix type annotations

---
 pandas/plotting/_core.py | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index d8fe99c2c97ac..cc0a15b5f98d7 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import importlib
+import types
 from typing import (
     TYPE_CHECKING,
     Sequence,
@@ -1727,7 +1728,7 @@ def hexbin(self, x, y, C=None, reduce_C_function=None, gridsize=None, **kwargs):
         return self(kind="hexbin", x=x, y=y, C=C, **kwargs)
 
 
-_backends = {}
+_backends: dict[str, types.ModuleType] = {}
 
 
 def _load_backend(backend: str):
@@ -1745,11 +1746,13 @@ def _load_backend(backend: str):
     types.ModuleType
         The imported backend.
     """
+    module: types.ModuleType | None = None
+
     if backend == "matplotlib":
         # Because matplotlib is an optional dependency and first-party backend,
         # we need to attempt an import here to raise an ImportError if needed.
         try:
-            import pandas.plotting._matplotlib as module
+            module = __import__("pandas.plotting._matplotlib")
         except ImportError:
             raise ImportError(
                 "matplotlib is required for plotting when the "
@@ -1757,8 +1760,6 @@ def _load_backend(backend: str):
             ) from None
         return module
 
-    module = None
-
     for entry_point in pkg_resources.iter_entry_points("pandas_plotting_backends"):
         if entry_point.name == backend:
             module = entry_point.load()

From 8808ecd714fc8a90adb49586ae1458c8e94952b7 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Tue, 18 May 2021 13:09:52 +0200
Subject: [PATCH 06/10] Add backend loading benchmark

---
 asv_bench/benchmarks/plotting.py | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/asv_bench/benchmarks/plotting.py b/asv_bench/benchmarks/plotting.py
index 11e43401f9395..7653520213d73 100644
--- a/asv_bench/benchmarks/plotting.py
+++ b/asv_bench/benchmarks/plotting.py
@@ -1,10 +1,15 @@
+import importlib
+import sys
+
 import matplotlib
 import numpy as np
+import pkg_resources
 
 from pandas import (
     DataFrame,
     DatetimeIndex,
     Series,
+    __file__ as pd_file,
     date_range,
 )
 
@@ -13,6 +18,8 @@
 except ImportError:
     from pandas.tools.plotting import andrews_curves
 
+from pandas.plotting._core import _get_plot_backend
+
 matplotlib.use("Agg")
 
 
@@ -99,4 +106,29 @@ def time_plot_andrews_curves(self):
         andrews_curves(self.df, "Name")
 
 
+class BackendLoading:
+    def setup(self):
+        dist = pkg_resources.get_distribution("pandas")
+        if dist.module_path not in pd_file:
+            # We are running from a non-installed pandas, and this benchmark is
+            # invalid
+            raise NotImplementedError("Testing a non-installed pandas")
+
+        spec = importlib.machinery.ModuleSpec("my_backend", None)
+        mod = importlib.util.module_from_spec(spec)
+        mod.plot = lambda *args, **kwargs: 1
+
+        backends = pkg_resources.get_entry_map("pandas")
+        my_entrypoint = pkg_resources.EntryPoint(
+            "pandas_plotting_backend", mod.__name__, dist=dist
+        )
+        backends["pandas_plotting_backends"][mod.__name__] = my_entrypoint
+        for i in range(1000):
+            backends["pandas_plotting_backends"][str(i)] = my_entrypoint
+        sys.modules["my_backend"] = mod
+
+    def time_get_plot_backend(self):
+        _get_plot_backend("my_backend")
+
+
 from .pandas_vb_common import setup  # noqa: F401 isort:skip

From 00f79d3caf0117dfe9ca4e7dac15a609e61faa6f Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Tue, 18 May 2021 14:21:42 +0200
Subject: [PATCH 07/10] Fix BackendLoading removing warmup

---
 asv_bench/benchmarks/plotting.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/asv_bench/benchmarks/plotting.py b/asv_bench/benchmarks/plotting.py
index 7653520213d73..d81a4019a7577 100644
--- a/asv_bench/benchmarks/plotting.py
+++ b/asv_bench/benchmarks/plotting.py
@@ -107,6 +107,10 @@ def time_plot_andrews_curves(self):
 
 
 class BackendLoading:
+    repeat = 1
+    number = 1
+    warmup_time = 0
+
     def setup(self):
         dist = pkg_resources.get_distribution("pandas")
         if dist.module_path not in pd_file:

From 25b1dced8f4c96fb6dab0b8987ec2b2eafbb1523 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Tue, 1 Jun 2021 18:29:02 +0200
Subject: [PATCH 08/10] Reduce number of dummy backends

---
 asv_bench/benchmarks/plotting.py | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/asv_bench/benchmarks/plotting.py b/asv_bench/benchmarks/plotting.py
index d81a4019a7577..249a8f3f556a1 100644
--- a/asv_bench/benchmarks/plotting.py
+++ b/asv_bench/benchmarks/plotting.py
@@ -9,7 +9,6 @@
     DataFrame,
     DatetimeIndex,
     Series,
-    __file__ as pd_file,
     date_range,
 )
 
@@ -113,11 +112,6 @@ class BackendLoading:
 
     def setup(self):
         dist = pkg_resources.get_distribution("pandas")
-        if dist.module_path not in pd_file:
-            # We are running from a non-installed pandas, and this benchmark is
-            # invalid
-            raise NotImplementedError("Testing a non-installed pandas")
-
         spec = importlib.machinery.ModuleSpec("my_backend", None)
         mod = importlib.util.module_from_spec(spec)
         mod.plot = lambda *args, **kwargs: 1
@@ -127,7 +121,7 @@ def setup(self):
             "pandas_plotting_backend", mod.__name__, dist=dist
         )
         backends["pandas_plotting_backends"][mod.__name__] = my_entrypoint
-        for i in range(1000):
+        for i in range(10):
             backends["pandas_plotting_backends"][str(i)] = my_entrypoint
         sys.modules["my_backend"] = mod
 

From be6f4f363de1e21c54f6660fa77be3a6967362f4 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Tue, 1 Jun 2021 18:30:04 +0200
Subject: [PATCH 09/10] Return type in _load_backend, changes for mypy

---
 pandas/plotting/_core.py | 24 +++++++++++++-----------
 1 file changed, 13 insertions(+), 11 deletions(-)

diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index cc0a15b5f98d7..a4db8018eb293 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -7,7 +7,6 @@
     Sequence,
 )
 
-# TODO: replace with `importlib.metadata` when python_requires >= 3.8.
 import pkg_resources
 
 from pandas._config import get_option
@@ -1731,7 +1730,7 @@ def hexbin(self, x, y, C=None, reduce_C_function=None, gridsize=None, **kwargs):
 _backends: dict[str, types.ModuleType] = {}
 
 
-def _load_backend(backend: str):
+def _load_backend(backend: str) -> types.ModuleType:
     """
     Load a pandas plotting backend.
 
@@ -1746,13 +1745,11 @@ def _load_backend(backend: str):
     types.ModuleType
         The imported backend.
     """
-    module: types.ModuleType | None = None
-
     if backend == "matplotlib":
         # Because matplotlib is an optional dependency and first-party backend,
         # we need to attempt an import here to raise an ImportError if needed.
         try:
-            module = __import__("pandas.plotting._matplotlib")
+            module = importlib.import_module("pandas.plotting._matplotlib")
         except ImportError:
             raise ImportError(
                 "matplotlib is required for plotting when the "
@@ -1760,22 +1757,27 @@ def _load_backend(backend: str):
             ) from None
         return module
 
+    found_backend = False
+
     for entry_point in pkg_resources.iter_entry_points("pandas_plotting_backends"):
-        if entry_point.name == backend:
+        found_backend = entry_point.name == backend
+        if found_backend:
             module = entry_point.load()
 
-    if module is None:
+    if not found_backend:
         # Fall back to unregistered, module name approach.
         try:
             module = importlib.import_module(backend)
+            found_backend = True
         except ImportError:
             # We re-raise later on.
             pass
 
-    if hasattr(module, "plot"):
-        # Validate that the interface is implemented when the option is set,
-        # rather than at plot time.
-        return module
+    if found_backend:
+        if hasattr(module, "plot"):
+            # Validate that the interface is implemented when the option is set,
+            # rather than at plot time.
+            return module
 
     raise ValueError(
         f"Could not find plotting backend '{backend}'. Ensure that you've "

From 54a54cae745c43d28de18633b714fa317a593f83 Mon Sep 17 00:00:00 2001
From: TLouf <loufthomas@gmail.com>
Date: Thu, 3 Jun 2021 10:09:07 +0200
Subject: [PATCH 10/10] Break iteration if found entry point

---
 pandas/plotting/_core.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/pandas/plotting/_core.py b/pandas/plotting/_core.py
index a4db8018eb293..5d3db13610845 100644
--- a/pandas/plotting/_core.py
+++ b/pandas/plotting/_core.py
@@ -1763,6 +1763,7 @@ def _load_backend(backend: str) -> types.ModuleType:
         found_backend = entry_point.name == backend
         if found_backend:
             module = entry_point.load()
+            break
 
     if not found_backend:
         # Fall back to unregistered, module name approach.