def _get_url(self, path): annex_repo, annex_key = _get_annex_repo_key(path) if annex_key: # so we do not have it yet! repo = AnnexRepo(annex_repo) whereis = repo.whereis(annex_key, output='full', key=True) # , batch=True) # TODO: support also regular http remotes etc urls = list(chain(*(x.get('urls', []) for x in whereis.values()))) # TODO: some kind of analysis/fallback and not just taking the first one # TODO: Opened connections already might even have already some "equivalent" URL if urls: return urls[0]
def get_remote_urls(rec, remote): for k, v in rec.items(): if v.get('description', '') in [remote, '[%s]' % remote]: return v.get('urls', []) return [] if __name__ == '__main__': annex = AnnexRepo('.', create=False, init=False) # enable datalad special remote urls_to_register = defaultdict(list) # key: urls try: annex.call_annex(["enableremote", "datalad"]) # go through each and see where urls aren't yet under web # seems might have also --in=datalad to restrict w = annex.whereis([], options=['--all'], output='full') lgr.info("Got %d entries", len(w)) for k, rec in tqdm(w.items()): datalad_urls = get_remote_urls(rec, 'datalad') web_urls = set(get_remote_urls(rec, 'web')) for url in datalad_urls: if url not in web_urls: if 'openneuro.s3' in url or 'openfmri.s3' in url: urls_to_register[k].append(url) else: lgr.warning("Found unexpected url %s" % url) finally: # disable datalad special remote annex.remove_remote("datalad") # need to disable it first lgr.info("Got %d entries which could get new urls", len(urls_to_register))
def test_add_archive_content(path_orig, url, repo_path): with chpwd(repo_path): # TODO we need to be able to pass path into add_archive_content # We could mock but I mean for the API assert_raises(RuntimeError, add_archive_content, "nonexisting.tar.gz") # no repo yet repo = AnnexRepo(repo_path, create=True) assert_raises(ValueError, add_archive_content, "nonexisting.tar.gz") # we can't add a file from outside the repo ATM assert_raises(FileNotInRepositoryError, add_archive_content, opj(path_orig, '1.tar.gz')) # Let's add first archive to the repo so we could test with swallow_outputs(): repo.add_urls([opj(url, '1.tar.gz')], options=["--pathdepth", "-1"]) for s in range(1, 5): repo.add_urls([opj(url, '%du/1.tar.gz' % s)], options=["--pathdepth", "-2"]) repo.commit("added 1.tar.gz") key_1tar = repo.get_file_key( '1.tar.gz') # will be used in the test later def d1_basic_checks(): ok_(exists('1')) ok_file_under_git('1', '1 f.txt', annexed=True) ok_file_under_git(opj('1', 'd', '1d'), annexed=True) ok_archives_caches(repo_path, 0) # and by default it just does it, everything goes to annex repo_ = add_archive_content('1.tar.gz') eq_(repo.path, repo_.path) d1_basic_checks() # If ran again, should proceed just fine since the content is the same so no changes would be made really add_archive_content('1.tar.gz') # But that other one carries updated file, so should fail due to overwrite with assert_raises(RuntimeError) as cme: add_archive_content(opj('1u', '1.tar.gz'), use_current_dir=True) # TODO: somewhat not precise since we have two possible "already exists" # -- in caching and overwrite check assert_in("already exists", str(cme.exception)) # but should do fine if overrides are allowed add_archive_content(opj('1u', '1.tar.gz'), existing='overwrite', use_current_dir=True) add_archive_content(opj('2u', '1.tar.gz'), existing='archive-suffix', use_current_dir=True) add_archive_content(opj('3u', '1.tar.gz'), existing='archive-suffix', use_current_dir=True) add_archive_content(opj('4u', '1.tar.gz'), existing='archive-suffix', use_current_dir=True) # rudimentary test assert_equal(sorted(map(basename, glob(opj(repo_path, '1', '1*')))), ['1 f-1.1.txt', '1 f-1.2.txt', '1 f-1.txt', '1 f.txt']) whereis = repo.whereis(glob(opj(repo_path, '1', '1*'))) # they all must be the same assert (all([x == whereis[0] for x in whereis[1:]])) # and we should be able to reference it while under subdirectory subdir = opj(repo_path, 'subdir') with chpwd(subdir, mkdir=True): add_archive_content(opj(pardir, '1.tar.gz'), use_current_dir=True) d1_basic_checks() # or we could keep relative path and also demand to keep the archive prefix # while extracting under original (annex root) dir add_archive_content(opj(pardir, '1.tar.gz'), add_archive_leading_dir=True) with chpwd(opj(repo_path, '1')): d1_basic_checks() with chpwd(repo_path): # test with excludes and renames and annex options add_archive_content('1.tar.gz', exclude=['d'], rename=['/ /_', '/^1/2'], annex_options="-c annex.largefiles=exclude=*.txt", delete=True) # no conflicts since new name ok_file_under_git('2', '1_f.txt', annexed=False) assert_false(exists(opj('2', 'd'))) assert_false(exists('1.tar.gz')) # delete was in effect # now test ability to extract within subdir with chpwd(opj(repo_path, 'd1'), mkdir=True): # Let's add first archive to the repo so we could test # named the same way but different content with swallow_outputs(): repo.add_urls([opj(url, 'd1', '1.tar.gz')], options=["--pathdepth", "-1"], cwd=getpwd()) # invoke under current subdir repo.commit("added 1.tar.gz in d1") def d2_basic_checks(): ok_(exists('1')) ok_file_under_git('1', '2 f.txt', annexed=True) ok_file_under_git(opj('1', 'd2', '2d'), annexed=True) ok_archives_caches(repo.path, 0) add_archive_content('1.tar.gz') d2_basic_checks() # in manual tests ran into the situation of inability to obtain on a single run # a file from an archive which was coming from a dropped key. I thought it was # tested in custom remote tests, but I guess not sufficiently well enough repo.drop(opj('1', '1 f.txt')) # should be all kosher repo.get(opj('1', '1 f.txt')) ok_archives_caches(repo.path, 1, persistent=True) ok_archives_caches(repo.path, 0, persistent=False) repo.drop(opj('1', '1 f.txt')) # should be all kosher repo.drop(key_1tar, key=True) # is available from the URL -- should be kosher repo.get(opj('1', '1 f.txt')) # that what managed to not work # TODO: check if persistent archive is there for the 1.tar.gz # We should be able to drop everything since available online with swallow_outputs(): clean(dataset=repo.path) repo.drop(key_1tar, key=True) # is available from the URL -- should be kosher repo.drop(opj('1', '1 f.txt')) # should be all kosher repo.get(opj('1', '1 f.txt')) # and should be able to get it again # bug was that dropping didn't work since archive was dropped first repo.call_annex(["drop", "--all"]) # verify that we can't drop a file if archive key was dropped and online archive was removed or changed size! ;) repo.get(key_1tar, key=True) unlink(opj(path_orig, '1.tar.gz')) with assert_raises(CommandError) as e: repo.drop(key_1tar, key=True) assert_equal(e.kwargs['stdout_json'][0]['success'], False) assert_result_values_cond( e.kwargs['stdout_json'], 'note', lambda x: '(Use --force to override this check, or adjust numcopies.)' in x) assert exists(opj(repo.path, repo.get_contentlocation(key_1tar)))
def get_remote_urls(rec, remote): for k, v in rec.items(): if v.get('description', '') in [remote, '[%s]' % remote]: return v.get('urls', []) return [] if __name__ == '__main__': annex = AnnexRepo('.', create=False, init=False) # enable datalad special remote urls_to_register = defaultdict(list) # key: urls try: annex._annex_custom_command([], ["git", "annex", "enableremote", "datalad"]) # go through each and see where urls aren't yet under web # seems might have also --in=datalad to restrict w = annex.whereis([], options=['--all'], output='full') lgr.info("Got %d entries", len(w)) for k, rec in tqdm(w.items()): datalad_urls = get_remote_urls(rec, 'datalad') web_urls = set(get_remote_urls(rec, 'web')) for url in datalad_urls: if url not in web_urls: if 'openneuro.s3' in url or 'openfmri.s3' in url: urls_to_register[k].append(url) else: lgr.warning("Found unexpected url %s" % url) finally: # disable datalad special remote annex.remove_remote("datalad") # need to disable it first lgr.info(