# -*- coding: utf-8 -*- # Copyright (C) 2018 Anaconda, Inc # SPDX-License-Identifier: Proprietary from __future__ import absolute_import, division, print_function, unicode_literals import bz2 from collections import OrderedDict import copy from datetime import datetime import json from numbers import Number import os from os.path import abspath, basename, getmtime, getsize, isdir, isfile, join, splitext, dirname import subprocess import sys import tarfile from tempfile import gettempdir import time from uuid import uuid4 # Lots of conda internals here. Should refactor to use exports. from conda.common.compat import ensure_binary # from conda.resolve import dashlist import pytz from jinja2 import Environment, PackageLoader from tqdm import tqdm import yaml from yaml.constructor import ConstructorError from yaml.parser import ParserError from yaml.scanner import ScannerError from yaml.reader import ReaderError import fnmatch from functools import partial import logging import libarchive import conda_package_handling.api from concurrent.futures import ProcessPoolExecutor, as_completed from concurrent.futures import Future, Executor from threading import Lock # BAD BAD BAD - conda internals from conda.core.subdir_data import SubdirData from conda.models.channel import Channel from . import conda_interface, utils from .conda_interface import MatchSpec, VersionOrder, human_bytes, context from .conda_interface import CondaError, CondaHTTPError, get_index, url_path from .conda_interface import TemporaryDirectory from .conda_interface import Resolve from .utils import glob, get_logger, FileNotFoundError, JSONDecodeError # try: # from conda.base.constants import CONDA_TARBALL_EXTENSIONS # except Exception: # from conda.base.constants import CONDA_TARBALL_EXTENSION # CONDA_TARBALL_EXTENSIONS = (CONDA_TARBALL_EXTENSION,) # TODO: better to define this in conda; doing it here because we're implementing it in conda-build first CONDA_TARBALL_EXTENSIONS = ('.conda', '.tar.bz2') log = get_logger(__name__) # use this for debugging, because ProcessPoolExecutor isn't pdb/ipdb friendly class DummyExecutor(Executor): def __init__(self): self._shutdown = False self._shutdownLock = Lock() def submit(self, fn, *args, **kwargs): with self._shutdownLock: if self._shutdown: raise RuntimeError('cannot schedule new futures after shutdown') f = Future() try: result = fn(*args, **kwargs) except BaseException as e: f.set_exception(e) else: f.set_result(result) return f def shutdown(self, wait=True): with self._shutdownLock: self._shutdown = True try: from conda.base.constants import NAMESPACES_MAP, NAMESPACE_PACKAGE_NAMES except ImportError: NAMESPACES_MAP = { # base package name, namespace "python": "python", "r": "r", "r-base": "r", "mro-base": "r", "mro-base_impl": "r", "erlang": "erlang", "java": "java", "openjdk": "java", "julia": "julia", "latex": "latex", "lua": "lua", "nodejs": "js", "perl": "perl", "php": "php", "ruby": "ruby", "m2-base": "m2", "msys2-conda-epoch": "m2w64", } NAMESPACE_PACKAGE_NAMES = frozenset(NAMESPACES_MAP) NAMESPACES = frozenset(NAMESPACES_MAP.values()) local_index_timestamp = 0 cached_index = None local_subdir = "" cached_channels = [] channel_data = {} # TODO: support for libarchive seems to have broken ability to use multiple threads here. # The new conda format is so much faster that it more than makes up for it. However, it # would be nice to fix this at some point. MAX_THREADS_DEFAULT = os.cpu_count() if (hasattr(os, "cpu_count") and os.cpu_count() > 1) else 1 if sys.platform == 'win32': # see https://github.com/python/cpython/commit/8ea0fd85bc67438f679491fae29dfe0a3961900a MAX_THREADS_DEFAULT = min(61, MAX_THREADS_DEFAULT) LOCK_TIMEOUT_SECS = 3 * 3600 LOCKFILE_NAME = ".lock" # TODO: this is to make sure that the index doesn't leak tokens. It breaks use of private channels, though. # os.environ['CONDA_ADD_ANACONDA_TOKEN'] = "false" try: from cytoolz.itertoolz import concat, concatv, groupby except ImportError: # pragma: no cover from conda._vendor.toolz.itertoolz import concat, concatv, groupby # NOQA def get_build_index(subdir, bldpkgs_dir, output_folder=None, clear_cache=False, omit_defaults=False, channel_urls=None, debug=False, verbose=True, **kwargs): global local_index_timestamp global local_subdir global cached_index global cached_channels global channel_data mtime = 0 channel_urls = list(utils.ensure_list(channel_urls)) if not output_folder: output_folder = dirname(bldpkgs_dir) # check file modification time - this is the age of our local index. index_file = os.path.join(output_folder, subdir, 'repodata.json') if os.path.isfile(index_file): mtime = os.path.getmtime(index_file) if (clear_cache or not os.path.isfile(index_file) or local_subdir != subdir or mtime > local_index_timestamp or cached_channels != channel_urls): # priority: (local as either croot or output_folder IF NOT EXPLICITLY IN CHANNEL ARGS), # then channels passed as args (if local in this, it remains in same order), # then channels from condarc. urls = list(channel_urls) loggers = utils.LoggingContext.default_loggers + [__name__] if debug: log_context = partial(utils.LoggingContext, logging.DEBUG, loggers=loggers) elif verbose: log_context = partial(utils.LoggingContext, logging.WARN, loggers=loggers) else: log_context = partial(utils.LoggingContext, logging.CRITICAL + 1, loggers=loggers) with log_context(): # this is where we add the "local" channel. It's a little smarter than conda, because # conda does not know about our output_folder when it is not the default setting. if os.path.isdir(output_folder): local_path = url_path(output_folder) # replace local with the appropriate real channel. Order is maintained. urls = [url if url != 'local' else local_path for url in urls] if local_path not in urls: urls.insert(0, local_path) _ensure_valid_channel(output_folder, subdir) update_index(output_folder, verbose=debug) # replace noarch with native subdir - this ends up building an index with both the # native content and the noarch content. if subdir == 'noarch': subdir = conda_interface.subdir try: cached_index = get_index(channel_urls=urls, prepend=not omit_defaults, use_local=False, use_cache=context.offline, platform=subdir) # HACK: defaults does not have the many subfolders we support. Omit it and # try again. except CondaHTTPError: if 'defaults' in urls: urls.remove('defaults') cached_index = get_index(channel_urls=urls, prepend=omit_defaults, use_local=False, use_cache=context.offline, platform=subdir) expanded_channels = {rec.channel for rec in cached_index.values()} superchannel = {} # we need channeldata.json too, as it is a more reliable source of run_exports data for channel in expanded_channels: if channel.scheme == "file": location = channel.location if utils.on_win: location = location.lstrip("/") elif (not os.path.isabs(channel.location) and os.path.exists(os.path.join(os.path.sep, channel.location))): location = os.path.join(os.path.sep, channel.location) channeldata_file = os.path.join(location, channel.name, 'channeldata.json') retry = 0 max_retries = 1 if os.path.isfile(channeldata_file): while retry < max_retries: try: with open(channeldata_file, "r+") as f: channel_data[channel.name] = json.load(f) break except (IOError, JSONDecodeError): time.sleep(0.2) retry += 1 else: # download channeldata.json for url if not context.offline: try: channel_data[channel.name] = utils.download_channeldata(channel.base_url + '/channeldata.json') except CondaHTTPError: continue # collapse defaults metachannel back into one superchannel, merging channeldata if channel.base_url in context.default_channels and channel_data.get(channel.name): packages = superchannel.get('packages', {}) packages.update(channel_data[channel.name]) superchannel['packages'] = packages channel_data['defaults'] = superchannel local_index_timestamp = os.path.getmtime(index_file) local_subdir = subdir cached_channels = channel_urls return cached_index, local_index_timestamp, channel_data def _ensure_valid_channel(local_folder, subdir): for folder in {subdir, 'noarch'}: path = os.path.join(local_folder, folder) if not os.path.isdir(path): os.makedirs(path) def update_index(dir_path, check_md5=False, channel_name=None, patch_generator=None, threads=MAX_THREADS_DEFAULT, verbose=False, progress=False, hotfix_source_repo=None, subdirs=None, warn=True, current_index_versions=None, debug=False): """ If dir_path contains a directory named 'noarch', the path tree therein is treated as though it's a full channel, with a level of subdirs, each subdir having an update to repodata.json. The full channel will also have a channeldata.json file. If dir_path does not contain a directory named 'noarch', but instead contains at least one '*.tar.bz2' file, the directory is assumed to be a standard subdir, and only repodata.json information will be updated. """ base_path, dirname = os.path.split(dir_path) if dirname in utils.DEFAULT_SUBDIRS: if warn: log.warn("The update_index function has changed to index all subdirs at once. You're pointing it at a single subdir. " "Please update your code to point it at the channel root, rather than a subdir.") return update_index(base_path, check_md5=check_md5, channel_name=channel_name, threads=threads, verbose=verbose, progress=progress, hotfix_source_repo=hotfix_source_repo, current_index_versions=current_index_versions) return ChannelIndex(dir_path, channel_name, subdirs=subdirs, threads=threads, deep_integrity_check=check_md5, debug=debug).index( patch_generator=patch_generator, verbose=verbose, progress=progress, hotfix_source_repo=hotfix_source_repo, current_index_versions=current_index_versions) def _determine_namespace(info): if info.get('namespace'): namespace = info['namespace'] else: depends_names = set() for spec in info.get('depends', []): try: depends_names.add(MatchSpec(spec).name) except CondaError: pass spaces = depends_names & NAMESPACE_PACKAGE_NAMES if len(spaces) == 1: namespace = NAMESPACES_MAP[spaces.pop()] else: namespace = "global" info['namespace'] = namespace if not info.get('namespace_in_name') and '-' in info['name']: namespace_prefix, reduced_name = info['name'].split('-', 1) if namespace_prefix == namespace: info['name_in_channel'] = info['name'] info['name'] = reduced_name return namespace, info.get('name_in_channel', info['name']), info['name'] def _make_seconds(timestamp): timestamp = int(timestamp) if timestamp > 253402300799: # 9999-12-31 timestamp //= 1000 # convert milliseconds to seconds; see conda/conda-build#1988 return timestamp # ========================================================================== REPODATA_VERSION = 1 CHANNELDATA_VERSION = 1 REPODATA_JSON_FN = 'repodata.json' REPODATA_FROM_PKGS_JSON_FN = 'repodata_from_packages.json' CHANNELDATA_FIELDS = ( "description", "dev_url", "doc_url", "doc_source_url", "home", "license", "reference_package", "source_url", "source_git_url", "source_git_tag", "source_git_rev", "summary", "version", "subdirs", "icon_url", "icon_hash", # "md5:abc123:12" "run_exports", "binary_prefix", "text_prefix", "activate.d", "deactivate.d", "pre_link", "post_link", "pre_unlink", "tags", "identifiers", "keywords", "recipe_origin", "commits", ) def _clear_newline_chars(record, field_name): if field_name in record: try: record[field_name] = record[field_name].strip().replace('\n', ' ') except AttributeError: # sometimes description gets added as a list instead of just a string record[field_name] = record[field_name][0].strip().replace('\n', ' ') def _apply_instructions(subdir, repodata, instructions): repodata.setdefault("removed", []) utils.merge_or_update_dict(repodata.get('packages', {}), instructions.get('packages', {}), merge=False, add_missing_keys=False) # we could have totally separate instructions for .conda than .tar.bz2, but it's easier if we assume # that a similarly-named .tar.bz2 file is the same content as .conda, and shares fixes new_pkg_fixes = {k.replace('.tar.bz2', '.conda'): v for k, v in instructions.get('packages', {}).items()} utils.merge_or_update_dict(repodata.get('packages.conda', {}), new_pkg_fixes, merge=False, add_missing_keys=False) utils.merge_or_update_dict(repodata.get('packages.conda', {}), instructions.get('packages.conda', {}), merge=False, add_missing_keys=False) for fn in instructions.get('revoke', ()): for key in ('packages', 'packages.conda'): if fn.endswith('.tar.bz2') and key == 'packages.conda': fn = fn.replace('.tar.bz2', '.conda') if fn in repodata[key]: repodata[key][fn]['revoked'] = True repodata[key][fn]['depends'].append('package_has_been_revoked') for fn in instructions.get('remove', ()): for key in ('packages', 'packages.conda'): if fn.endswith('.tar.bz2') and key == 'packages.conda': fn = fn.replace('.tar.bz2', '.conda') popped = repodata[key].pop(fn, None) if popped: repodata["removed"].append(fn) repodata["removed"].sort() return repodata def _get_jinja2_environment(): def _filter_strftime(dt, dt_format): if isinstance(dt, Number): if dt > 253402300799: # 9999-12-31 dt //= 1000 # convert milliseconds to seconds; see #1988 dt = datetime.utcfromtimestamp(dt).replace(tzinfo=pytz.timezone("UTC")) return dt.strftime(dt_format) def _filter_add_href(text, link, **kwargs): if link: kwargs_list = ['href="{0}"'.format(link)] kwargs_list.append('alt="{0}"'.format(text)) kwargs_list += ['{0}="{1}"'.format(k, v) for k, v in kwargs.items()] return '{1}'.format(' '.join(kwargs_list), text) else: return text environment = Environment( loader=PackageLoader('conda_build', 'templates'), ) environment.filters['human_bytes'] = human_bytes environment.filters['strftime'] = _filter_strftime environment.filters['add_href'] = _filter_add_href environment.trim_blocks = True environment.lstrip_blocks = True return environment def _maybe_write(path, content, write_newline_end=False, content_is_binary=False): temp_path = join(gettempdir(), str(uuid4())) if not content_is_binary: content = ensure_binary(content) with open(temp_path, 'wb') as fh: fh.write(content) if write_newline_end: fh.write(b'\n') if isfile(path): if utils.md5_file(temp_path) == utils.md5_file(path): # No need to change mtimes. The contents already match. os.unlink(temp_path) return False # log.info("writing %s", path) utils.move_with_fallback(temp_path, path) return True def _make_build_string(build, build_number): build_number_as_string = str(build_number) if build.endswith(build_number_as_string): build = build[:-len(build_number_as_string)] build = build.rstrip("_") build_string = build return build_string def _warn_on_missing_dependencies(missing_dependencies, patched_repodata): """ The following dependencies do not exist in the channel and are not declared as external dependencies: dependency1: - subdir/fn1.tar.bz2 - subdir/fn2.tar.bz2 dependency2: - subdir/fn3.tar.bz2 - subdir/fn4.tar.bz2 The associated packages are being removed from the index. """ if missing_dependencies: builder = [ "WARNING: The following dependencies do not exist in the channel", " and are not declared as external dependencies:" ] for dep_name in sorted(missing_dependencies): builder.append(" %s" % dep_name) for subdir_fn in sorted(missing_dependencies[dep_name]): builder.append(" - %s" % subdir_fn) subdir, fn = subdir_fn.split("/") popped = patched_repodata["packages"].pop(fn, None) if popped: patched_repodata["removed"].append(fn) builder.append("The associated packages are being removed from the index.") builder.append('') log.warn("\n".join(builder)) def _cache_post_install_details(paths_cache_path, post_install_cache_path): post_install_details_json = {'binary_prefix': False, 'text_prefix': False, 'activate.d': False, 'deactivate.d': False, 'pre_link': False, 'post_link': False, 'pre_unlink': False} if os.path.lexists(paths_cache_path): with open(paths_cache_path) as f: paths = json.load(f).get('paths', []) # get embedded prefix data from paths.json for f in paths: if f.get('prefix_placeholder'): if f.get('file_mode') == 'binary': post_install_details_json['binary_prefix'] = True elif f.get('file_mode') == 'text': post_install_details_json['text_prefix'] = True # check for any activate.d/deactivate.d scripts for k in ('activate.d', 'deactivate.d'): if not post_install_details_json.get(k) and f['_path'].startswith('etc/conda/%s' % k): post_install_details_json[k] = True # check for any link scripts for pat in ('pre-link', 'post-link', 'pre-unlink'): if not post_install_details_json.get(pat) and fnmatch.fnmatch(f['_path'], '*/.*-%s.*' % pat): post_install_details_json[pat.replace("-", "_")] = True with open(post_install_cache_path, 'w') as fh: json.dump(post_install_details_json, fh) def _cache_recipe(tmpdir, recipe_cache_path): recipe_path_search_order = ( 'info/recipe/meta.yaml.rendered', 'info/recipe/meta.yaml', 'info/meta.yaml', ) for path in recipe_path_search_order: recipe_path = os.path.join(tmpdir, path) if os.path.lexists(recipe_path): break recipe_path = None recipe_json = {} if recipe_path: with open(recipe_path) as f: try: recipe_json = yaml.safe_load(f) except (ConstructorError, ParserError, ScannerError, ReaderError): pass try: recipe_json_str = json.dumps(recipe_json) except TypeError: recipe_json.get('requirements', {}).pop('build') recipe_json_str = json.dumps(recipe_json) with open(recipe_cache_path, 'w') as fh: fh.write(recipe_json_str) return recipe_json def _cache_run_exports(tmpdir, run_exports_cache_path): run_exports = {} try: with open(os.path.join(tmpdir, 'info', 'run_exports.json')) as f: run_exports = json.load(f) except (IOError, FileNotFoundError): try: with open(os.path.join(tmpdir, 'info', 'run_exports.yaml')) as f: run_exports = yaml.safe_load(f) except (IOError, FileNotFoundError): log.debug("%s has no run_exports file (this is OK)" % tmpdir) with open(run_exports_cache_path, 'w') as fh: json.dump(run_exports, fh) def _cache_icon(tmpdir, recipe_json, icon_cache_path): # If a conda package contains an icon, also extract and cache that in an .icon/ # directory. The icon file name is the name of the package, plus the extension # of the icon file as indicated by the meta.yaml `app/icon` key. # apparently right now conda-build renames all icons to 'icon.png' # What happens if it's an ico file, or a svg file, instead of a png? Not sure! app_icon_path = recipe_json.get('app', {}).get('icon') if app_icon_path: icon_path = os.path.join(tmpdir, 'info', 'recipe', app_icon_path) if not os.path.lexists(icon_path): icon_path = os.path.join(tmpdir, 'info', 'icon.png') if os.path.lexists(icon_path): icon_cache_path += splitext(app_icon_path)[-1] utils.move_with_fallback(icon_path, icon_cache_path) def _make_subdir_index_html(channel_name, subdir, repodata_packages, extra_paths): environment = _get_jinja2_environment() template = environment.get_template('subdir-index.html.j2') rendered_html = template.render( title="%s/%s" % (channel_name or '', subdir), packages=repodata_packages, current_time=datetime.utcnow().replace(tzinfo=pytz.timezone("UTC")), extra_paths=extra_paths, ) return rendered_html def _make_channeldata_index_html(channel_name, channeldata): environment = _get_jinja2_environment() template = environment.get_template('channeldata-index.html.j2') rendered_html = template.render( title=channel_name, packages=channeldata['packages'], subdirs=channeldata['subdirs'], current_time=datetime.utcnow().replace(tzinfo=pytz.timezone("UTC")), ) return rendered_html def _get_source_repo_git_info(path): is_repo = subprocess.check_output(["git", "rev-parse", "--is-inside-work-tree"], cwd=path) if is_repo.strip().decode('utf-8') == "true": output = subprocess.check_output(['git', 'log', "--pretty=format:'%h|%ad|%an|%s'", "--date=unix"], cwd=path) commits = [] for line in output.decode("utf-8").strip().splitlines(): _hash, _time, _author, _desc = line.split("|") commits.append({"hash": _hash, "timestamp": int(_time), "author": _author, "description": _desc}) return commits def _cache_info_file(tmpdir, info_fn, cache_path): info_path = os.path.join(tmpdir, 'info', info_fn) if os.path.lexists(info_path): utils.move_with_fallback(info_path, cache_path) def _alternate_file_extension(fn): cache_fn = fn for ext in CONDA_TARBALL_EXTENSIONS: cache_fn = cache_fn.replace(ext, '') other_ext = set(CONDA_TARBALL_EXTENSIONS) - set([fn.replace(cache_fn, '')]) return cache_fn + next(iter(other_ext)) def _get_resolve_object(subdir, file_path=None, precs=None, repodata=None): packages = {} conda_packages = {} if file_path: with open(file_path) as fi: packages = json.load(fi) recs = json.load(fi) for k, v in recs.items(): if k.endswith('.tar.bz2'): packages[k] = v elif k.endswith('.conda'): conda_packages[k] = v if not repodata: repodata = { "info": { "subdir": subdir, "arch": context.arch_name, "platform": context.platform, }, "packages": packages, "packages.conda": conda_packages, } channel = Channel('https://conda.anaconda.org/dummy-channel/%s' % subdir) sd = SubdirData(channel) sd._process_raw_repodata_str(json.dumps(repodata)) sd._loaded = True SubdirData._cache_[channel.url(with_credentials=True)] = sd index = {prec: prec for prec in precs or sd._package_records} r = Resolve(index, channels=(channel,)) return r def _get_newest_versions(r, pins={}): groups = {} for g_name, g_recs in r.groups.items(): if g_name in pins: matches = [] for pin in pins[g_name]: version = r.find_matches(MatchSpec('%s=%s' % (g_name, pin)))[0].version matches.extend(r.find_matches(MatchSpec('%s=%s' % (g_name, version)))) else: version = r.groups[g_name][0].version matches = r.find_matches(MatchSpec('%s=%s' % (g_name, version))) groups[g_name] = matches return [pkg for group in groups.values() for pkg in group] def _add_missing_deps(new_r, original_r): """For each package in new_r, if any deps are not satisfiable, backfill them from original_r.""" expanded_groups = copy.deepcopy(new_r.groups) seen_specs = set() for g_name, g_recs in new_r.groups.items(): for g_rec in g_recs: for dep_spec in g_rec.depends: if dep_spec in seen_specs: continue ms = MatchSpec(dep_spec) if not new_r.find_matches(ms): matches = original_r.find_matches(ms) if matches: version = matches[0].version expanded_groups[ms.name] = ( set(expanded_groups.get(ms.name, [])) | set(original_r.find_matches(MatchSpec('%s=%s' % (ms.name, version))))) seen_specs.add(dep_spec) return [pkg for group in expanded_groups.values() for pkg in group] def _shard_newest_packages(subdir, r, pins=None): """Captures only the newest versions of software in the resolve object. For things where more than one version is supported simultaneously (like Python), pass pins as a dictionary, with the key being the package name, and the value being a list of supported versions. For example: {'python': ["2.7", "3.6"]} """ groups = {} pins = pins or {} for g_name, g_recs in r.groups.items(): # always do the latest implicitly version = r.groups[g_name][0].version matches = set(r.find_matches(MatchSpec('%s=%s' % (g_name, version)))) if g_name in pins: for pin_value in pins[g_name]: version = r.find_matches(MatchSpec('%s=%s' % (g_name, pin_value)))[0].version matches.update(r.find_matches(MatchSpec('%s=%s' % (g_name, version)))) groups[g_name] = matches new_r = _get_resolve_object(subdir, precs=[pkg for group in groups.values() for pkg in group]) return set(_add_missing_deps(new_r, r)) def _build_current_repodata(subdir, repodata, pins): r = _get_resolve_object(subdir, repodata=repodata) keep_pkgs = _shard_newest_packages(subdir, r, pins) new_repodata = {k: repodata[k] for k in set(repodata.keys()) - set(['packages', 'packages.conda'])} packages = {} conda_packages = {} for keep_pkg in keep_pkgs: if keep_pkg.fn.endswith('.conda'): conda_packages[keep_pkg.fn] = repodata['packages.conda'][keep_pkg.fn] # in order to prevent package churn we consider the md5 for the .tar.bz2 that matches the .conda file # This holds when .conda files contain the same files as .tar.bz2, which is an assumption we'll make # until it becomes more prevalent that people provide only .conda files and just skip .tar.bz2 counterpart = keep_pkg.fn.replace('.conda', '.tar.bz2') conda_packages[keep_pkg.fn]['legacy_bz2_md5'] = repodata['packages'].get(counterpart, {}).get('md5') elif keep_pkg.fn.endswith('.tar.bz2'): packages[keep_pkg.fn] = repodata['packages'][keep_pkg.fn] new_repodata['packages'] = packages new_repodata['packages.conda'] = conda_packages return new_repodata class ChannelIndex(object): def __init__(self, channel_root, channel_name, subdirs=None, threads=MAX_THREADS_DEFAULT, deep_integrity_check=False, debug=False): self.channel_root = abspath(channel_root) self.channel_name = channel_name or basename(channel_root.rstrip('/')) self._subdirs = subdirs self.thread_executor = (DummyExecutor() if (debug or sys.version_info.major == 2) else ProcessPoolExecutor(threads)) self.deep_integrity_check = deep_integrity_check def index(self, patch_generator, hotfix_source_repo=None, verbose=False, progress=False, current_index_versions=None): if verbose: level = logging.DEBUG else: level = logging.ERROR with utils.LoggingContext(level, loggers=[__name__]): if not self._subdirs: detected_subdirs = set(subdir for subdir in os.listdir(self.channel_root) if subdir in utils.DEFAULT_SUBDIRS and isdir(join(self.channel_root, subdir))) log.debug("found subdirs %s" % detected_subdirs) self.subdirs = subdirs = sorted(detected_subdirs | {'noarch'}) else: self.subdirs = subdirs = sorted(set(self._subdirs) | {'noarch'}) # Step 1. Lock local channel. with utils.try_acquire_locks([utils.get_lock(self.channel_root)], timeout=900): channel_data = {} channeldata_file = os.path.join(self.channel_root, 'channeldata.json') if os.path.isfile(channeldata_file): with open(channeldata_file) as f: channel_data = json.load(f) # Step 2. Collect repodata from packages, save to pkg_repodata.json file with tqdm(total=len(subdirs), disable=(verbose or not progress), leave=False) as t: for subdir in subdirs: t.set_description("Subdir: %s" % subdir) t.update() with tqdm(total=8, disable=(verbose or not progress), leave=False) as t2: t2.set_description("Gathering repodata") t2.update() _ensure_valid_channel(self.channel_root, subdir) repodata_from_packages = self.index_subdir( subdir, verbose=verbose, progress=progress) t2.set_description("Writing pre-patch repodata") t2.update() self._write_repodata(subdir, repodata_from_packages, REPODATA_FROM_PKGS_JSON_FN) # Step 3. Apply patch instructions. t2.set_description("Applying patch instructions") t2.update() patched_repodata, patch_instructions = self._patch_repodata( subdir, repodata_from_packages, patch_generator) # Step 4. Save patched and augmented repodata. # If the contents of repodata have changed, write a new repodata.json file. # Also create associated index.html. t2.set_description("Writing patched repodata") t2.update() self._write_repodata(subdir, patched_repodata, REPODATA_JSON_FN) t2.set_description("Building current_repodata subset") t2.update() current_repodata = _build_current_repodata(subdir, patched_repodata, pins=current_index_versions) t2.set_description("Writing current_repodata subset") t2.update() self._write_repodata(subdir, current_repodata, json_filename="current_repodata.json") t2.set_description("Writing subdir index HTML") t2.update() self._write_subdir_index_html(subdir, patched_repodata) t2.set_description("Updating channeldata") t2.update() self._update_channeldata(channel_data, patched_repodata, subdir) # Step 7. Create and write channeldata. self._write_channeldata_index_html(channel_data) self._write_channeldata(channel_data) def index_subdir(self, subdir, verbose=False, progress=False): subdir_path = join(self.channel_root, subdir) self._ensure_dirs(subdir) repodata_json_path = join(subdir_path, REPODATA_FROM_PKGS_JSON_FN) if verbose: log.info("Building repodata for %s" % subdir_path) # gather conda package filenames in subdir # we'll process these first, because reading their metadata is much faster fns_in_subdir = {fn for fn in os.listdir(subdir_path) if fn.endswith('.conda') or fn.endswith('.tar.bz2')} # load current/old repodata try: with open(repodata_json_path) as fh: old_repodata = json.load(fh) or {} except (EnvironmentError, JSONDecodeError): # log.info("no repodata found at %s", repodata_json_path) old_repodata = {} old_repodata_packages = old_repodata.get("packages", {}) old_repodata_conda_packages = old_repodata.get("packages.conda", {}) old_repodata_fns = set(old_repodata_packages) | set(old_repodata_conda_packages) # Load stat cache. The stat cache has the form # { # 'package_name.tar.bz2': { # 'mtime': 123456, # 'md5': 'abd123', # }, # } stat_cache_path = join(subdir_path, '.cache', 'stat.json') try: with open(stat_cache_path) as fh: stat_cache = json.load(fh) or {} except: stat_cache = {} stat_cache_original = stat_cache.copy() try: # calculate all the paths and figure out what we're going to do with them # add_set: filenames that aren't in the current/old repodata, but exist in the subdir add_set = fns_in_subdir - old_repodata_fns remove_set = old_repodata_fns - fns_in_subdir ignore_set = set(old_repodata.get('removed', [])) add_set -= ignore_set # update_set: Filenames that are in both old repodata and new repodata, # and whose contents have changed based on file size or mtime. We're # not using md5 here because it takes too long. If needing to do full md5 checks, # use the --deep-integrity-check flag / self.deep_integrity_check option. update_set = self._calculate_update_set( subdir, fns_in_subdir, old_repodata_fns, stat_cache, verbose=verbose, progress=progress ) # unchanged_set: packages in old repodata whose information can carry straight # across to new repodata unchanged_set = sorted(old_repodata_fns - update_set - remove_set - ignore_set) # clean up removed files removed_set = (old_repodata_fns - fns_in_subdir) for fn in removed_set: if fn in stat_cache: del stat_cache[fn] new_repodata_packages = {k: v for k, v in old_repodata.get('packages', {}).items() if k in unchanged_set} new_repodata_conda_packages = {k: v for k, v in old_repodata.get('packages.conda', {}).items() if k in unchanged_set} for k in unchanged_set: if not (k in new_repodata_packages or k in new_repodata_conda_packages): fn, rec = ChannelIndex._load_index_from_cache(self.channel_root, subdir, fn, stat_cache) # this is how we pass an exception through. When fn == rec, there's been a problem, # and we need to reload this file if fn == rec: update_set.add(fn) else: if fn.endswith('.tar.bz2'): new_repodata_packages[fn] = rec else: new_repodata_conda_packages[fn] = rec # Invalidate cached files for update_set. # Extract and cache update_set and add_set, then add to new_repodata_packages. # This is also where we update the contents of the stat_cache for successfully # extracted packages. # Sorting here prioritizes .conda files ('c') over .tar.bz2 files ('b') hash_extract_set = tuple(concatv(add_set, update_set)) # split up the set by .conda packages first, then .tar.bz2. This avoids race conditions # with execution in parallel that would end up in the same place. for conda_format in tqdm(CONDA_TARBALL_EXTENSIONS, desc="File format", disable=(verbose or not progress), leave=False): futures = tuple(self.thread_executor.submit( ChannelIndex._extract_to_cache, self.channel_root, subdir, fn ) for fn in hash_extract_set if fn.endswith(conda_format)) with tqdm(desc="hash & extract packages for %s" % subdir, total=len(futures), disable=(verbose or not progress), leave=False) as t: for future in as_completed(futures): fn, mtime, size, index_json = future.result() # fn can be None if the file was corrupt or no longer there if fn and mtime: # the progress bar shows package names, but we don't know what their name is before they complete. t.set_description("Updated: %s" % fn) t.update() stat_cache[fn] = {'mtime': int(mtime), 'size': size} if index_json: if fn.endswith(".conda"): new_repodata_conda_packages[fn] = index_json else: new_repodata_packages[fn] = index_json new_repodata = { 'packages': new_repodata_packages, 'packages.conda': new_repodata_conda_packages, 'info': { 'subdir': subdir, }, 'repodata_version': REPODATA_VERSION, 'removed': sorted(list(ignore_set)) } finally: if stat_cache != stat_cache_original: # log.info("writing stat cache to %s", stat_cache_path) with open(stat_cache_path, 'w') as fh: json.dump(stat_cache, fh) return new_repodata def _ensure_dirs(self, subdir): # Create all cache directories in the subdir. ensure = lambda path: isdir(path) or os.makedirs(path) cache_path = join(self.channel_root, subdir, '.cache') ensure(cache_path) ensure(join(cache_path, 'index')) ensure(join(cache_path, 'about')) ensure(join(cache_path, 'paths')) ensure(join(cache_path, 'recipe')) ensure(join(cache_path, 'run_exports')) ensure(join(cache_path, 'post_install')) ensure(join(cache_path, 'icon')) ensure(join(self.channel_root, 'icons')) ensure(join(cache_path, 'recipe_log')) def _calculate_update_set(self, subdir, fns_in_subdir, old_repodata_fns, stat_cache, verbose=False, progress=True): # Determine the packages that already exist in repodata, but need to be updated. # We're not using md5 here because it takes too long. candidate_fns = fns_in_subdir & old_repodata_fns subdir_path = join(self.channel_root, subdir) update_set = set() for fn in tqdm(iter(candidate_fns), desc="Finding updated files", disable=(verbose or not progress), leave=False): if fn not in stat_cache: update_set.add(fn) else: stat_result = os.stat(join(subdir_path, fn)) if (int(stat_result.st_mtime) != int(stat_cache[fn]['mtime']) or stat_result.st_size != stat_cache[fn]['size']): update_set.add(fn) return update_set @staticmethod def _extract_to_cache(channel_root, subdir, fn, second_try=False): # This method WILL reread the tarball. Probably need another one to exit early if # there are cases where it's fine not to reread. Like if we just rebuild repodata # from the cached files, but don't use the existing repodata.json as a starting point. subdir_path = join(channel_root, subdir) # allow .conda files to reuse cache from .tar.bz2 and vice-versa. # Assumes that .tar.bz2 and .conda files have exactly the same # contents. This is convention, but not guaranteed, nor checked. alternate_cache_fn = _alternate_file_extension(fn) cache_fn = fn abs_fn = os.path.join(subdir_path, fn) stat_result = os.stat(abs_fn) size = stat_result.st_size mtime = stat_result.st_mtime retval = fn, mtime, size, None index_cache_path = join(subdir_path, '.cache', 'index', cache_fn + '.json') about_cache_path = join(subdir_path, '.cache', 'about', cache_fn + '.json') paths_cache_path = join(subdir_path, '.cache', 'paths', cache_fn + '.json') recipe_cache_path = join(subdir_path, '.cache', 'recipe', cache_fn + '.json') run_exports_cache_path = join(subdir_path, '.cache', 'run_exports', cache_fn + '.json') post_install_cache_path = join(subdir_path, '.cache', 'post_install', cache_fn + '.json') icon_cache_path = join(subdir_path, '.cache', 'icon', cache_fn) log.debug("hashing, extracting, and caching %s" % fn) alternate_cache = False if (not os.path.exists(index_cache_path) and os.path.exists(index_cache_path.replace(fn, alternate_cache_fn))): alternate_cache = True try: # allow .tar.bz2 files to use the .conda cache, but not vice-versa. # .conda readup is very fast (essentially free), but .conda files come from # converting .tar.bz2 files, which can go wrong. Forcing extraction for # .conda files gives us a check on the validity of that conversion. if not fn.endswith('.conda') and os.path.isfile(index_cache_path): with open(index_cache_path) as f: index_json = json.load(f) elif not alternate_cache and (second_try or not os.path.exists(index_cache_path)): with TemporaryDirectory() as tmpdir: conda_package_handling.api.extract(abs_fn, dest_dir=tmpdir, components="info") index_file = os.path.join(tmpdir, 'info', 'index.json') if not os.path.exists(index_file): return retval with open(index_file) as f: index_json = json.load(f) _cache_info_file(tmpdir, 'about.json', about_cache_path) _cache_info_file(tmpdir, 'paths.json', paths_cache_path) _cache_info_file(tmpdir, 'recipe_log.json', paths_cache_path) _cache_run_exports(tmpdir, run_exports_cache_path) _cache_post_install_details(paths_cache_path, post_install_cache_path) recipe_json = _cache_recipe(tmpdir, recipe_cache_path) _cache_icon(tmpdir, recipe_json, icon_cache_path) # decide what fields to filter out, like has_prefix filter_fields = { 'arch', 'has_prefix', 'mtime', 'platform', 'ucs', 'requires_features', 'binstar', 'target-triplet', 'machine', 'operatingsystem', } for field_name in filter_fields & set(index_json): del index_json[field_name] elif alternate_cache: # we hit the cache of the other file type. Copy files to this name, and replace # the size, md5, and sha256 values paths = [index_cache_path, about_cache_path, paths_cache_path, recipe_cache_path, run_exports_cache_path, post_install_cache_path, icon_cache_path] bizarro_paths = [_.replace(fn, alternate_cache_fn) for _ in paths] for src, dest in zip(bizarro_paths, paths): if os.path.exists(src): try: os.makedirs(os.path.dirname(dest)) except: pass utils.copy_into(src, dest) with open(index_cache_path) as f: index_json = json.load(f) else: with open(index_cache_path) as f: index_json = json.load(f) # calculate extra stuff to add to index.json cache, size, md5, sha256 # This is done always for all files, whether the cache is loaded or not, # because the cache may be from the other file type. We don't store this # info in the cache to avoid confusion. index_json.update(conda_package_handling.api.get_pkg_details(abs_fn)) with open(index_cache_path, 'w') as fh: json.dump(index_json, fh) retval = fn, mtime, size, index_json except (libarchive.exception.ArchiveError, tarfile.ReadError, KeyError, EOFError, JSONDecodeError) as e: if not second_try: return ChannelIndex._extract_to_cache(channel_root, subdir, fn, second_try=True) log.error("Package %s appears to be corrupt. Please remove it and re-download it" % abs_fn) log.error(str(e)) except FileNotFoundError as e: log.error("Package %s appears to be corrupt. Please remove it and re-download it" % abs_fn) log.error(str(e)) return retval @staticmethod def _load_index_from_cache(channel_root, subdir, fn, stat_cache): index_cache_path = join(channel_root, subdir, '.cache', 'index', fn + '.json') try: with open(index_cache_path) as fh: index_json = json.load(fh) except (IOError, JSONDecodeError): index_json = fn return fn, index_json @staticmethod def _load_all_from_cache(channel_root, subdir, fn): subdir_path = join(channel_root, subdir) try: mtime = getmtime(join(subdir_path, fn)) except FileNotFoundError: return {} # In contrast to self._load_index_from_cache(), this method reads up pretty much # all of the cached metadata, except for paths. It all gets dumped into a single map. index_cache_path = join(subdir_path, '.cache', 'index', fn + '.json') about_cache_path = join(subdir_path, '.cache', 'about', fn + '.json') recipe_cache_path = join(subdir_path, '.cache', 'recipe', fn + '.json') run_exports_cache_path = join(subdir_path, '.cache', 'run_exports', fn + '.json') post_install_cache_path = join(subdir_path, '.cache', 'post_install', fn + '.json') icon_cache_path_glob = join(subdir_path, '.cache', 'icon', fn + ".*") recipe_log_path = join(subdir_path, '.cache', 'recipe_log', fn + '.json') data = {} for path in (recipe_cache_path, about_cache_path, index_cache_path, post_install_cache_path, recipe_log_path): try: if os.path.getsize(path) != 0: with open(path) as fh: data.update(json.load(fh)) except (OSError, EOFError, IOError): pass icon_cache_paths = glob(icon_cache_path_glob) if icon_cache_paths: icon_cache_path = sorted(icon_cache_paths)[-1] icon_ext = icon_cache_path.rsplit('.', 1)[-1] channel_icon_fn = "%s.%s" % (data['name'], icon_ext) icon_url = "icons/" + channel_icon_fn icon_channel_path = join(channel_root, 'icons', channel_icon_fn) icon_md5 = utils.md5_file(icon_cache_path) icon_hash = "md5:%s:%s" % (icon_md5, getsize(icon_cache_path)) data.update(icon_hash=icon_hash, icon_url=icon_url) # log.info("writing icon from %s to %s", icon_cache_path, icon_channel_path) utils.move_with_fallback(icon_cache_path, icon_channel_path) # have to stat again, because we don't have access to the stat cache here data['mtime'] = mtime source = data.get("source", {}) try: data.update({"source_" + k: v for k, v in source.items()}) except AttributeError: # sometimes source is a list instead of a dict pass _clear_newline_chars(data, 'description') _clear_newline_chars(data, 'summary') try: with open(run_exports_cache_path) as fh: data["run_exports"] = json.load(fh) except (OSError, EOFError): data["run_exports"] = {} return data def _write_repodata(self, subdir, repodata, json_filename): repodata_json_path = join(self.channel_root, subdir, json_filename) new_repodata_binary = json.dumps(repodata, indent=2, sort_keys=True,).replace("':'", "': '").encode("utf-8") write_result = _maybe_write(repodata_json_path, new_repodata_binary, write_newline_end=True) if write_result: repodata_bz2_path = repodata_json_path + ".bz2" bz2_content = bz2.compress(new_repodata_binary) _maybe_write(repodata_bz2_path, bz2_content, content_is_binary=True) return write_result def _write_subdir_index_html(self, subdir, repodata): repodata_packages = repodata["packages"] subdir_path = join(self.channel_root, subdir) def _add_extra_path(extra_paths, path): if isfile(join(self.channel_root, path)): extra_paths[basename(path)] = { 'size': getsize(path), 'timestamp': int(getmtime(path)), 'sha256': utils.sha256_checksum(path), 'md5': utils.md5_file(path), } extra_paths = OrderedDict() _add_extra_path(extra_paths, join(subdir_path, REPODATA_JSON_FN)) _add_extra_path(extra_paths, join(subdir_path, REPODATA_JSON_FN + '.bz2')) _add_extra_path(extra_paths, join(subdir_path, REPODATA_FROM_PKGS_JSON_FN)) _add_extra_path(extra_paths, join(subdir_path, REPODATA_FROM_PKGS_JSON_FN + '.bz2')) # _add_extra_path(extra_paths, join(subdir_path, "repodata2.json")) _add_extra_path(extra_paths, join(subdir_path, "patch_instructions.json")) rendered_html = _make_subdir_index_html( self.channel_name, subdir, repodata_packages, extra_paths ) index_path = join(subdir_path, 'index.html') return _maybe_write(index_path, rendered_html) def _write_channeldata_index_html(self, channeldata): rendered_html = _make_channeldata_index_html( self.channel_name, channeldata ) index_path = join(self.channel_root, 'index.html') _maybe_write(index_path, rendered_html) def _update_channeldata(self, channel_data, repodata, subdir): legacy_packages = repodata["packages"] conda_packages = repodata["packages.conda"] use_these_legacy_keys = set(legacy_packages.keys()) - set(k[:-6] + '.tar.bz2' for k in conda_packages.keys()) all_repodata_packages = conda_packages.copy() all_repodata_packages.update({k: legacy_packages[k] for k in use_these_legacy_keys}) package_data = channel_data.get('packages', {}) def _append_group(groups, candidate): pkg_dict = candidate[1] pkg_name = pkg_dict['name'] run_exports = package_data.get(pkg_name, {}).get('run_exports', {}) if (pkg_name not in package_data or subdir not in package_data.get(pkg_name, {}).get('subdirs', []) or package_data.get(pkg_name, {}).get('timestamp', 0) < _make_seconds(pkg_dict.get('timestamp', 0)) or run_exports and pkg_dict['version'] not in run_exports): groups.append(candidate) groups = [] package_groups = groupby(lambda x: x[1]['name'], all_repodata_packages.items()) for groupname, group in package_groups.items(): if (groupname not in package_data or package_data[groupname].get('run_exports')): # pay special attention to groups that have run_exports - we need to process each version # group by version; take newest per version group. We handle groups that are not # in the index t all yet similarly, because we can't check if they have any run_exports for vgroup in groupby(lambda x: x[1]['version'], group).values(): candidate = next(iter(sorted(vgroup, key=lambda x: x[1].get('timestamp', 0), reverse=True))) _append_group(groups, candidate) else: # take newest per group candidate = next(iter(sorted(group, key=lambda x: x[1].get('timestamp', 0), reverse=True))) _append_group(groups, candidate) def _replace_if_newer_and_present(pd, data, erec, data_newer, k): if data.get(k) and (data_newer or not erec.get(k)): pd[k] = data[k] else: pd[k] = erec.get(k) # unzipping fns, fn_dicts = [], [] if groups: fns, fn_dicts = zip(*groups) futures = tuple(self.thread_executor.submit( ChannelIndex._load_all_from_cache, self.channel_root, subdir, fn ) for fn in fns) for fn_dict, future in zip(fn_dicts, futures): data = future.result() if data: data.update(fn_dict) name = data['name'] # existing record erec = package_data.get(name, {}) data_v = data.get('version', '0') erec_v = erec.get('version', '0') data_newer = VersionOrder(data_v) > VersionOrder(erec_v) package_data[name] = package_data.get(name, {}) # keep newer value for these for k in ('description', 'dev_url', 'doc_url', 'doc_source_url', 'home', 'license', 'source_url', 'source_git_url', 'summary', 'icon_url', 'icon_hash', 'tags', 'identifiers', 'keywords', 'recipe_origin', 'version'): _replace_if_newer_and_present(package_data[name], data, erec, data_newer, k) # keep any true value for these, since we don't distinguish subdirs for k in ("binary_prefix", "text_prefix", "activate.d", "deactivate.d", "pre_link", "post_link", "pre_unlink"): package_data[name][k] = any((data.get(k), erec.get(k))) package_data[name]['subdirs'] = sorted(list(set(erec.get('subdirs', []) + [subdir]))) # keep one run_exports entry per version of the package, since these vary by version run_exports = erec.get('run_exports', {}) exports_from_this_version = data.get('run_exports') if exports_from_this_version: run_exports[data_v] = data.get('run_exports') package_data[name]['run_exports'] = run_exports package_data[name]['timestamp'] = _make_seconds(max( data.get('timestamp', 0), channel_data.get(name, {}).get('timestamp', 0))) channel_data.update({ 'channeldata_version': CHANNELDATA_VERSION, 'subdirs': sorted(list(set(channel_data.get('subdirs', []) + [subdir]))), 'packages': package_data, }) def _write_channeldata(self, channeldata): # trim out commits, as they can take up a ton of space. They're really only for the RSS feed. for _pkg, pkg_dict in channeldata.get('packages', {}).items(): if "commits" in pkg_dict: del pkg_dict['commits'] channeldata_path = join(self.channel_root, 'channeldata.json') content = json.dumps(channeldata, indent=2, sort_keys=True).replace("':'", "': '") _maybe_write(channeldata_path, content, True) def _load_patch_instructions_tarball(self, subdir, patch_generator): instructions = {} with TemporaryDirectory() as tmpdir: conda_package_handling.api.extract(patch_generator, dest_dir=tmpdir) instructions_file = os.path.join(tmpdir, subdir, "patch_instructions.json") if os.path.isfile(instructions_file): with open(instructions_file) as f: instructions = json.load(f) return instructions def _create_patch_instructions(self, subdir, repodata, patch_generator=None): gen_patch_path = patch_generator or join(self.channel_root, 'gen_patch.py') if isfile(gen_patch_path): log.debug("using patch generator %s for %s" % (gen_patch_path, subdir)) # https://stackoverflow.com/a/41595552/2127762 try: from importlib.util import spec_from_file_location, module_from_spec spec = spec_from_file_location('a_b', gen_patch_path) mod = module_from_spec(spec) spec.loader.exec_module(mod) # older pythons except ImportError: import imp mod = imp.load_source('a_b', gen_patch_path) instructions = mod._patch_repodata(repodata, subdir) if instructions.get('patch_instructions_version', 0) > 1: raise RuntimeError("Incompatible patch instructions version") return instructions else: if patch_generator: raise ValueError("Specified metadata patch file '{}' does not exist. Please try an absolute " "path, or examine your relative path carefully with respect to your cwd." .format(patch_generator)) return {} def _write_patch_instructions(self, subdir, instructions): new_patch = json.dumps(instructions, indent=2, sort_keys=True).replace("':'", "': '") patch_instructions_path = join(self.channel_root, subdir, 'patch_instructions.json') _maybe_write(patch_instructions_path, new_patch, True) def _load_instructions(self, subdir): patch_instructions_path = join(self.channel_root, subdir, 'patch_instructions.json') if isfile(patch_instructions_path): log.debug("using patch instructions %s" % patch_instructions_path) with open(patch_instructions_path) as fh: instructions = json.load(fh) if instructions.get('patch_instructions_version', 0) > 1: raise RuntimeError("Incompatible patch instructions version") return instructions return {} def _patch_repodata(self, subdir, repodata, patch_generator=None): if patch_generator and any(patch_generator.endswith(ext) for ext in CONDA_TARBALL_EXTENSIONS): instructions = self._load_patch_instructions_tarball(subdir, patch_generator) else: instructions = self._create_patch_instructions(subdir, repodata, patch_generator) if instructions: self._write_patch_instructions(subdir, instructions) else: instructions = self._load_instructions(subdir) if instructions.get('patch_instructions_version', 0) > 1: raise RuntimeError("Incompatible patch instructions version") return _apply_instructions(subdir, repodata, instructions), instructions