aboutsummaryrefslogtreecommitdiff
path: root/wikiget/validations.py
diff options
context:
space:
mode:
Diffstat (limited to 'wikiget/validations.py')
-rw-r--r--wikiget/validations.py64
1 files changed, 0 insertions, 64 deletions
diff --git a/wikiget/validations.py b/wikiget/validations.py
deleted file mode 100644
index dc70df4..0000000
--- a/wikiget/validations.py
+++ /dev/null
@@ -1,64 +0,0 @@
-# wikiget - CLI tool for downloading files from Wikimedia sites
-# Copyright (C) 2018, 2019, 2020 Cody Logan
-# SPDX-License-Identifier: GPL-3.0-or-later
-#
-# Wikiget is free software: you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published by
-# the Free Software Foundation, either version 3 of the License, or
-# (at your option) any later version.
-#
-# Wikiget is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with Wikiget. If not, see <https://www.gnu.org/licenses/>.
-
-import hashlib
-import re
-
-from wikiget import BLOCKSIZE
-
-
-def valid_file(search_string):
- """
- Determines if the given string contains a valid file name, defined as a
- string ending with a '.' and at least one character, beginning with 'File:'
- or 'Image:', the standard file prefixes in MediaWiki.
- :param search_string: string to validate
- :returns: a regex Match object if there's a match or None otherwise
- """
- # second group could also restrict to file extensions with three or more
- # letters with ([^/\r\n\t\f\v]+\.\w{3,})
- file_regex = re.compile(r"(File:|Image:)([^/\r\n\t\f\v]+\.\w+)$", re.I)
- return file_regex.search(search_string)
-
-
-def valid_site(search_string):
- """
- Determines if the given string contains a valid site name, defined as a
- string ending with 'wikipedia.org' or 'wikimedia.org'. This covers all
- subdomains of those domains. Eventually, it should be possible to support
- any MediaWiki site, regardless of domain name.
- :param search_string: string to validate
- :returns: a regex Match object if there's a match or None otherwise
- """
- site_regex = re.compile(r"wiki[mp]edia\.org$", re.I)
- return site_regex.search(search_string)
-
-
-def verify_hash(filename):
- """
- Calculates the SHA1 hash of the given file for comparison with a known
- value.
- :param filename: name of the file to calculate a hash for
- :return: hash digest
- """
- hasher = hashlib.sha1() # noqa: S324
- with open(filename, "rb") as dl:
- buf = dl.read(BLOCKSIZE)
- while len(buf) > 0:
- hasher.update(buf)
- buf = dl.read(BLOCKSIZE)
- return hasher.hexdigest()