From ddebf171a437e5034542266e2a4fe0ee0a007c30 Mon Sep 17 00:00:00 2001
From: Mark Harfouche <mark.harfouche@gmail.com>
Date: Tue, 3 Oct 2023 15:18:12 -0400
Subject: [PATCH] Speed up import time by lazy loading requests (#328)

Pooch is often used as an import addon but not as a core feature.
As such it is a little silly to expand the import time of dependent
packages by a lot. While the gain here is small, it is not
insignificant when considering that multiple packages are using
Pooch and would any small amount of time would add up.
---
 pooch/core.py        |  5 +++--
 pooch/downloaders.py | 22 +++++++++++++++-------
 2 files changed, 18 insertions(+), 9 deletions(-)

diff --git a/pooch/core.py b/pooch/core.py
index fb7a3cda..7eeab6f0 100644
--- a/pooch/core.py
+++ b/pooch/core.py
@@ -14,8 +14,6 @@
 import shlex
 import shutil
 
-import requests
-import requests.exceptions
 
 from .hashes import hash_matches, file_hash
 from .utils import (
@@ -792,6 +790,9 @@ def stream_download(url, fname, known_hash, downloader, pooch=None, retry_if_fai
     will retry the download the specified number of times in case the failure
     was due to a network error.
     """
+    # Lazy import requests to speed up import time
+    import requests.exceptions  # pylint: disable=C0415
+
     # Ensure the parent directory exists in case the file is in a subdirectory.
     # Otherwise, move will cause an error.
     if not fname.parent.exists():
diff --git a/pooch/downloaders.py b/pooch/downloaders.py
index 383f2a12..1cae6b1e 100644
--- a/pooch/downloaders.py
+++ b/pooch/downloaders.py
@@ -12,7 +12,6 @@
 import ftplib
 
 import warnings
-import requests
 
 from .utils import parse_url
 
@@ -192,6 +191,9 @@ def __call__(self, url, output_file, pooch, check_only=False):
             is available on the server. Otherwise, returns ``None``.
 
         """
+        # Lazy import requests to speed up import time
+        import requests  # pylint: disable=C0415
+
         if check_only:
             response = requests.head(url, allow_redirects=True)
             available = bool(response.status_code == 200)
@@ -626,6 +628,9 @@ def doi_to_url(doi):
         The URL of the archive in the data repository.
 
     """
+    # Lazy import requests to speed up import time
+    import requests  # pylint: disable=C0415
+
     # Use doi.org to resolve the DOI to the repository website.
     response = requests.get(f"https://doi.org/{doi}")
     url = response.url
@@ -777,8 +782,10 @@ def initialize(cls, doi, archive_url):
     @property
     def api_response(self):
         """Cached API response from Zenodo"""
-
         if self._api_response is None:
+            # Lazy import requests to speed up import time
+            import requests  # pylint: disable=C0415
+
             article_id = self.archive_url.split("/")[-1]
             self._api_response = requests.get(
                 f"https://zenodo.org/api/records/{article_id}"
@@ -801,7 +808,6 @@ def download_url(self, file_name):
         download_url : str
             The HTTP URL that can be used to download the file.
         """
-
         files = {item["key"]: item for item in self.api_response["files"]}
         if file_name not in files:
             raise ValueError(
@@ -875,8 +881,10 @@ def _parse_version_from_doi(self):
     @property
     def api_response(self):
         """Cached API response from Figshare"""
-
         if self._api_response is None:
+            # Lazy import requests to speed up import time
+            import requests  # pylint: disable=C0415
+
             # Use the figshare API to find the article ID from the DOI
             article = requests.get(
                 f"https://api.figshare.com/v2/articles?doi={self.doi}"
@@ -927,7 +935,6 @@ def download_url(self, file_name):
         download_url : str
             The HTTP URL that can be used to download the file.
         """
-
         files = {item["name"]: item for item in self.api_response}
         if file_name not in files:
             raise ValueError(
@@ -974,7 +981,6 @@ def initialize(cls, doi, archive_url):
         archive_url : str
             The resolved URL for the DOI
         """
-
         # Access the DOI as if this was a DataVerse instance
         response = cls._get_api_response(doi, archive_url)
 
@@ -995,6 +1001,9 @@ def _get_api_response(cls, doi, archive_url):
         This has been separated into a separate ``classmethod``, as it can be
         used prior and after the initialization.
         """
+        # Lazy import requests to speed up import time
+        import requests  # pylint: disable=C0415
+
         parsed = parse_url(archive_url)
         response = requests.get(
             f"{parsed['protocol']}://{parsed['netloc']}/api/datasets/"
@@ -1034,7 +1043,6 @@ def download_url(self, file_name):
         download_url : str
             The HTTP URL that can be used to download the file.
         """
-
         parsed = parse_url(self.archive_url)
 
         # Iterate over the given files until we find one of the requested name