#!/usr/bin/env python3 import difflib import filecmp import glob import hashlib import os import subprocess import sys import ujson from typing import Optional, List, Tuple TOOLS_DIR = os.path.abspath(os.path.dirname(__file__)) ZULIP_PATH = os.path.dirname(TOOLS_DIR) REQS_DIR = os.path.join(ZULIP_PATH, 'requirements') CACHE_DIR = os.path.join(ZULIP_PATH, 'var', 'tmp') TMP_DIR = '/var/tmp' if 'TRAVIS' in os.environ: CACHE_DIR = os.path.join(os.environ['HOME'], 'misc') CACHE_FILE = os.path.join(CACHE_DIR, 'requirements_hashes') LOCKED_REQS_FILE_NAMES = ['dev.txt', 'docs.txt', 'mypy.txt', 'prod.txt', 'thumbor.txt'] def print_diff(path_file1, path_file2): # type: (str, str) -> None with open(path_file1) as file1: with open(path_file2) as file2: diff = difflib.unified_diff( file1.readlines(), file2.readlines(), fromfile=path_file1, tofile=path_file2, ) for line in diff: print(line) def test_locked_requirements(): # type: () -> bool # `pip-compile` tries to avoid unnecessarily updating recursive dependencies # if lock files are present already. If we don't copy these files to the tmp # dir then recursive dependencies will get updated to their latest version # without any change in the input requirements file and the test will not pass. for fn in LOCKED_REQS_FILE_NAMES: locked_file = os.path.join(REQS_DIR, fn) test_locked_file = os.path.join(TMP_DIR, fn) subprocess.check_call(['cp', locked_file, test_locked_file]) subprocess.check_call([os.path.join(TOOLS_DIR, 'update-locked-requirements'), '--output-dir', TMP_DIR], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) same = True for fn in LOCKED_REQS_FILE_NAMES: locked_file = os.path.join(REQS_DIR, fn) test_locked_file = os.path.join(TMP_DIR, fn) same = same and filecmp.cmp(test_locked_file, locked_file, shallow=False) return same def get_requirements_hash(use_test_lock_files=False): # type: (Optional[bool]) -> str sha1 = hashlib.sha1() reqs_files = glob.glob(os.path.join(ZULIP_PATH, "requirements", "*.in")) lock_files_path = REQS_DIR if use_test_lock_files: lock_files_path = TMP_DIR reqs_files.extend([os.path.join(lock_files_path, fn) for fn in LOCKED_REQS_FILE_NAMES]) for file_path in reqs_files: with open(file_path) as fp: sha1.update(fp.read().encode("utf-8")) return sha1.hexdigest() def may_be_setup_cache(): # type: () -> None if not os.path.exists(CACHE_DIR): subprocess.check_call(['mkdir', CACHE_DIR]) if not os.path.exists(CACHE_FILE): with open(CACHE_FILE, 'w') as fp: ujson.dump([], fp) def load_cache(): # type: () -> List[str] with open(CACHE_FILE) as fp: hash_list = ujson.load(fp) return hash_list def update_cache(hash_list): # type: (List[str]) -> None # We store last 100 hash entries. Aggressive caching is # not a problem as it is cheap to do. if len(hash_list) > 100: hash_list = hash_list[-100:] with open(CACHE_FILE, 'w') as fp: ujson.dump(hash_list, fp) def main(): # type: () -> None may_be_setup_cache() hash_list = load_cache() curr_hash = get_requirements_hash() if curr_hash in hash_list: # We have already checked this set of requirements and they # were consistent so no need to check again. return requirements_are_consistent = test_locked_requirements() # Cache the hash so that we need not to run the `update_locked_requirements` # tool again for checking this set of requirements. valid_hash = get_requirements_hash(use_test_lock_files=True) hash_list.append(valid_hash) update_cache(hash_list) if not requirements_are_consistent: for fn in LOCKED_REQS_FILE_NAMES: locked_file = os.path.join(REQS_DIR, fn) test_locked_file = os.path.join(TMP_DIR, fn) print_diff(test_locked_file, locked_file) # Flush the output to ensure we print the error at the end. sys.stdout.flush() raise Exception("It looks like you have updated some python dependencies but haven't " "updated locked requirements files. Please update them by running " "`tools/update-locked-requirements`. For more information please " "refer to `requirements/README.md`.") if __name__ == '__main__': main()