def test_add_archive_single_file(repo_path=None):
    ds = Dataset(repo_path).create(force=True)
    with chpwd(repo_path):
        archives = glob('archives/*')
        ds.save(archives, message='Added archives')

        for archive in archives:
            archive_name = os.path.splitext(archive)[0]
            archive_content = os.path.basename(archive_name)
            ds.add_archive_content(archive)
            ok_file_has_content(archive_name, archive_content)
Beispiel #2
0
def test_annex_get_from_subdir(topdir=None):
    ds = Dataset(topdir)
    ds.create(force=True)
    ds.save('a.tar.gz')
    ds.add_archive_content('a.tar.gz', delete=True)
    fpath = op.join(topdir, 'a', 'd', fn_in_archive_obscure)

    with chpwd(op.join(topdir, 'a', 'd')):
        runner = WitlessRunner()
        runner.run(['git', 'annex', 'drop', '--', fn_in_archive_obscure],
                   protocol=KillOutput)  # run git annex drop
        assert_false(ds.repo.file_has_content(
            fpath))  # and verify if file deleted from directory
        runner.run(['git', 'annex', 'get', '--', fn_in_archive_obscure],
                   protocol=KillOutput)  # run git annex get
        assert_true(ds.repo.file_has_content(
            fpath))  # and verify if file got into directory
def test_add_archive_content(path_orig=None, url=None, repo_path=None):
    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

        # no repo yet
        assert_raises(NoDatasetFound, add_archive_content,
                      "nonexisting.tar.gz")
        ds = Dataset(repo_path).create()
        res = ds.add_archive_content("nonexisting.tar.gz", on_failure='ignore')
        assert_in_results(res,
                          action='add-archive-content',
                          status='impossible')
        repo = ds.repo

        # we can't add a file from outside the repo ATM
        res = ds.add_archive_content(Path(path_orig) / '1.tar.gz',
                                     on_failure='ignore')
        assert_in_results(res,
                          action='add-archive-content',
                          status='impossible',
                          type="dataset",
                          message="Can not add archive outside of the dataset")

        # Let's add first archive to the repo so we could test
        with swallow_outputs():
            repo.add_url_to_file('1.tar.gz', opj(url, '1.tar.gz'))
            for s in range(1, 5):
                repo.add_url_to_file('%du/1.tar.gz' % s,
                                     opj(url, '%du/1.tar.gz' % s))
            repo.commit("added 1.tar.gz")

        key_1tar = repo.get_file_annexinfo('1.tar.gz')[
            'key']  # 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
        res = add_archive_content('1.tar.gz')
        assert_in_results(res, action='add-archive-content', status='ok')
        d1_basic_checks()

        # If ran again, should proceed just fine since the content is the same
        # so no changes would be made really
        res = add_archive_content('1.tar.gz')
        assert_in_results(res, action='add-archive-content', status='ok')

        # But that other one carries updated file, so should fail due to
        # overwrite
        res = add_archive_content(Path('1u') / '1.tar.gz',
                                  use_current_dir=True,
                                  on_failure='ignore')
        assert_in_results(
            res,
            action='add-archive-content',
            status='error',
        )
        assert_in('exists, but would be overwritten by new file',
                  res[0]['message'])
        # but should do fine if overrides are allowed
        add_archive_content(Path('1u') / '1.tar.gz',
                            existing='overwrite',
                            use_current_dir=True)
        add_archive_content(Path('2u') / '1.tar.gz',
                            existing='archive-suffix',
                            use_current_dir=True)
        add_archive_content(Path('3u') / '1.tar.gz',
                            existing='archive-suffix',
                            use_current_dir=True)
        add_archive_content(Path('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'),
                            dataset=ds.path,
                            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'),
                            dataset=ds.path,
                            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
        ds.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_url_to_file('d1/1.tar.gz', opj(url, 'd1', '1.tar.gz'))
        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', dataset=ds.path)
        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=ds)
    repo.drop(key_1tar,
              key=True)  # is available from the URL -- should be kosher

    ds.drop(opj('1', '1 f.txt'))  # should be all kosher
    ds.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)))