def test_uninstall_multiple_paths(path):
ds = Dataset(path).create(force=True)
subds = ds.create('deep', force=True)
subds.save(recursive=True)
ok_clean_git(subds.path)
# needs to be able to add a combination of staged files, modified submodule,
# and untracked files
ds.save(recursive=True)
ok_clean_git(ds.path)
# drop content of all 'kill' files
topfile = 'kill'
deepfile = opj('deep', 'dir', 'kill')
# use a tuple not a list! should also work
ds.drop((topfile, deepfile), check=False)
ok_clean_git(ds.path)
files_left = glob(opj(ds.path, '*', '*', '*')) + glob(opj(ds.path, '*'))
ok_(all([f.endswith('keep') for f in files_left if exists(f) and not isdir(f)]))
ok_(not ds.repo.file_has_content(topfile))
ok_(not subds.repo.file_has_content(opj(*psplit(deepfile)[1:])))
# remove handles for all 'kill' files
ds.remove([topfile, deepfile], check=False)
ok_clean_git(ds.path)
files_left = glob(opj(ds.path, '*', '*', '*')) + glob(opj(ds.path, '*'))
ok_(all([f.endswith('keep') for f in files_left if exists(f) and not isdir(f)]))
ok_(not any([f.endswith(topfile) for f in files_left]))
def test_kill(path):
# nested datasets with load
ds = Dataset(path).create()
testfile = opj(ds.path, "file.dat")
with open(testfile, 'w') as f:
f.write("load")
ds.save("file.dat")
subds = ds.create('deep1')
eq_(sorted(ds.subdatasets(result_xfm='relpaths')), ['deep1'])
ok_clean_git(ds.path)
# and we fail to remove since content can't be dropped
res = ds.remove(on_failure='ignore')
assert_result_count(
res, 1,
status='error', path=testfile)
# Following two assertions on message are relying on the actual error.
# We have a second result with status 'impossible' for the ds, that we need
# to filter out for those assertions:
err_result = [r for r in res if r['status'] == 'error'][0]
assert_result_values_cond(
[err_result], 'message',
lambda x: "configured minimum number of copies not found" in x or
"Could only verify the existence of 0 out of 1 necessary copies" in x
)
eq_(ds.remove(recursive=True, check=False, result_xfm='datasets'),
[subds, ds])
ok_(not exists(path))
def test_install_into_dataset(source, top_path):
ds = create(top_path)
ok_clean_git(ds.path)
subds = ds.install("sub", source=source, save=False)
ok_(isdir(opj(subds.path, '.git')))
ok_(subds.is_installed())
assert_in('sub', ds.subdatasets(result_xfm='relpaths'))
# sub is clean:
ok_clean_git(subds.path, annex=None)
# top is too:
ok_clean_git(ds.path, annex=None)
ds.save(message='addsub')
# now it is:
ok_clean_git(ds.path, annex=None)
# but we could also save while installing and there should be no side-effect
# of saving any other changes if we state to not auto-save changes
# Create a dummy change
create_tree(ds.path, {'dummy.txt': 'buga'})
ok_clean_git(ds.path, untracked=['dummy.txt'])
subds_ = ds.install("sub2", source=source)
eq_(subds_.path, opj(ds.path, "sub2")) # for paranoid yoh ;)
ok_clean_git(ds.path, untracked=['dummy.txt'])
# and we should achieve the same behavior if we create a dataset
# and then decide to add it
create(_path_(top_path, 'sub3'))
ok_clean_git(ds.path, untracked=['dummy.txt', 'sub3/'])
ds.save('sub3')
ok_clean_git(ds.path, untracked=['dummy.txt'])
def test_clean_subds_removal(path):
ds = Dataset(path).create()
subds1 = ds.create('one')
subds2 = ds.create('two')
eq_(sorted(ds.subdatasets(result_xfm='relpaths')), ['one', 'two'])
ok_clean_git(ds.path)
# now kill one
res = ds.remove('one', result_xfm=None)
# subds1 got uninstalled, and ds got the removal of subds1 saved
assert_result_count(res, 1, path=subds1.path, action='uninstall', status='ok')
assert_result_count(res, 1, path=subds1.path, action='remove', status='ok')
assert_result_count(res, 1, path=ds.path, action='save', status='ok')
ok_(not subds1.is_installed())
ok_clean_git(ds.path)
# two must remain
eq_(ds.subdatasets(result_xfm='relpaths'), ['two'])
# one is gone
assert(not exists(subds1.path))
# and now again, but this time remove something that is not installed
ds.create('three')
eq_(sorted(ds.subdatasets(result_xfm='relpaths')), ['three', 'two'])
ds.uninstall('two')
ok_clean_git(ds.path)
eq_(sorted(ds.subdatasets(result_xfm='relpaths')), ['three', 'two'])
ok_(not subds2.is_installed())
assert(exists(subds2.path))
res = ds.remove('two', result_xfm='datasets')
ok_clean_git(ds.path)
# subds2 was already uninstalled, now ds got the removal of subds2 saved
assert(not exists(subds2.path))
eq_(ds.subdatasets(result_xfm='relpaths'), ['three'])
eq_(res, [subds2, ds])
def test_ssh_manager_close():
manager = SSHManager()
# check for previously existing sockets:
existed_before_1 = exists(opj(manager.socket_dir, 'localhost'))
existed_before_2 = exists(opj(manager.socket_dir, 'datalad-test'))
manager.get_connection('ssh://localhost').open()
manager.get_connection('ssh://datalad-test').open()
if existed_before_1 and existed_before_2:
# we need one connection to be closed and therefore being opened
# by `manager`
manager.get_connection('ssh://localhost').close()
manager.get_connection('ssh://localhost').open()
ok_(exists(opj(manager.socket_dir, get_connection_hash('localhost'))))
ok_(exists(opj(manager.socket_dir, get_connection_hash('datalad-test'))))
manager.close()
still_exists_1 = exists(opj(manager.socket_dir, 'localhost'))
still_exists_2 = exists(opj(manager.socket_dir, 'datalad-test'))
eq_(existed_before_1, still_exists_1)
eq_(existed_before_2, still_exists_2)
def test_GitRepo_ssh_pull(remote_path, repo_path):
from datalad import ssh_manager
remote_repo = GitRepo(remote_path, create=True)
url = "ssh://localhost" + op.abspath(remote_path)
socket_path = op.join(ssh_manager.socket_dir, get_connection_hash('localhost'))
repo = GitRepo(repo_path, create=True)
repo.add_remote("ssh-remote", url)
# modify remote:
remote_repo.checkout("ssh-test", ['-b'])
with open(op.join(remote_repo.path, "ssh_testfile.dat"), "w") as f:
f.write("whatever")
remote_repo.add("ssh_testfile.dat")
remote_repo.commit("ssh_testfile.dat added.")
# file is not locally known yet:
assert_not_in("ssh_testfile.dat", repo.get_indexed_files())
# pull changes:
repo.pull(remote="ssh-remote", refspec=remote_repo.get_active_branch())
ok_clean_git(repo.path, annex=False)
# the connection is known to the SSH manager, since fetch() requested it:
assert_in(socket_path, ssh_manager._connections)
# and socket was created:
ok_(op.exists(socket_path))
# we actually pulled the changes
assert_in("ssh_testfile.dat", repo.get_indexed_files())
def test_add_subdataset(path, other):
subds = create(opj(path, 'dir'), force=True)
ds = create(path, force=True)
ok_(subds.repo.dirty)
ok_(ds.repo.dirty)
assert_not_in('dir', ds.subdatasets(result_xfm='relpaths'))
# without a base dataset the next is interpreted as "add everything
# in subds to subds"
add(subds.path)
ok_clean_git(subds.path)
assert_not_in('dir', ds.subdatasets(result_xfm='relpaths'))
# but with a base directory we add the dataset subds as a subdataset
# to ds
ds.add(subds.path)
assert_in('dir', ds.subdatasets(result_xfm='relpaths'))
# create another one
other = create(other)
# install into superdataset, but don't add
other_clone = install(source=other.path, path=opj(ds.path, 'other'))
ok_(other_clone.is_installed)
assert_not_in('other', ds.subdatasets(result_xfm='relpaths'))
# now add, it should pick up the source URL
ds.add('other')
# and that is why, we can reobtain it from origin
ds.uninstall('other')
ok_(other_clone.is_installed)
ds.get('other')
ok_(other_clone.is_installed)
def test_url_base():
# Basic checks
assert_raises(ValueError, URL, "http://example.com", hostname='example.com')
url = URL("http://example.com")
eq_(url.hostname, 'example.com')
eq_(url.scheme, 'http')
eq_(url.port, '') # not specified -- empty strings
eq_(url.username, '') # not specified -- empty strings
eq_(repr(url), "URL(hostname='example.com', scheme='http')")
eq_(url, "http://example.com") # automagic coercion in __eq__
neq_(URL(), URL(hostname='x'))
smth = URL('smth')
eq_(smth.hostname, '')
ok_(bool(smth))
nok_(bool(URL()))
assert_raises(ValueError, url._set_from_fields, unknown='1')
with swallow_logs(new_level=logging.WARNING) as cml:
# we don't "care" about params ATM so there is a warning if there are any
purl = URL("http://example.com/;param")
eq_(str(purl), 'http://example.com/;param') # but we do maintain original string
assert_in('ParseResults contains params', cml.out)
eq_(purl.as_str(), 'http://example.com/')
def test_install_list(path, top_path):
# we want to be able to install several things, if these are known
# (no 'source' allowed). Therefore first toplevel:
ds = install(top_path, source=path, recursive=False)
assert_not_in('annex.hardlink', ds.config)
ok_(ds.is_installed())
sub1 = Dataset(opj(top_path, 'subm 1'))
sub2 = Dataset(opj(top_path, '2'))
ok_(not sub1.is_installed())
ok_(not sub2.is_installed())
# fails, when `source` is passed:
assert_raises(ValueError, ds.install,
path=['subm 1', '2'],
source='something')
# now should work:
result = ds.install(path=['subm 1', '2'], result_xfm='paths')
ok_(sub1.is_installed())
ok_(sub2.is_installed())
eq_(set(result), {sub1.path, sub2.path})
# and if we request it again via get, result should be empty
get_result = ds.get(path=['subm 1', '2'], get_data=False)
assert_status('notneeded', get_result)
def test_get_mixed_hierarchy(src, path):
origin = Dataset(src).create(no_annex=True)
origin_sub = origin.create('subds')
with open(opj(origin.path, 'file_in_git.txt'), "w") as f:
f.write('no idea')
with open(opj(origin_sub.path, 'file_in_annex.txt'), "w") as f:
f.write('content')
origin.add('file_in_git.txt', to_git=True)
origin_sub.add('file_in_annex.txt')
origin.save()
# now, install that thing:
ds, subds = install(
path, source=src, recursive=True,
result_xfm='datasets', return_type='item-or-list', result_filter=None)
ok_(subds.repo.file_has_content("file_in_annex.txt") is False)
# and get:
result = ds.get(curdir, recursive=True)
# git repo and subds
assert_status(['ok', 'notneeded'], result)
assert_result_count(
result, 1, path=opj(subds.path, "file_in_annex.txt"), status='ok')
ok_(subds.repo.file_has_content("file_in_annex.txt") is True)
def test_GitRepo_pull(test_path, orig_path, clone_path):
origin = GitRepo.clone(test_path, orig_path)
clone = GitRepo.clone(orig_path, clone_path)
filename = get_most_obscure_supported_name()
with open(op.join(orig_path, filename), 'w') as f:
f.write("New file.")
origin.add(filename)
origin.commit("new file added.")
clone.pull()
ok_(op.exists(op.join(clone_path, filename)))
# While at it, let's test _get_remotes_having_commit a bit
clone.add_remote("very_origin", test_path)
clone.fetch("very_origin")
eq_(
clone._get_remotes_having_commit(clone.get_hexsha()),
['origin']
)
prev_commit = clone.get_hexsha('HEAD^')
eq_(
set(clone._get_remotes_having_commit(prev_commit)),
{'origin', 'very_origin'}
)
def test_install_known_subdataset(src, path):
# get the superdataset:
ds = install(path, source=src)
# subdataset not installed:
subds = Dataset(opj(path, 'subm 1'))
assert_false(subds.is_installed())
assert_in('subm 1', ds.subdatasets(fulfilled=False, result_xfm='relpaths'))
assert_not_in('subm 1', ds.subdatasets(fulfilled=True, result_xfm='relpaths'))
# install it:
ds.install('subm 1')
ok_(subds.is_installed())
ok_(AnnexRepo.is_valid_repo(subds.path, allow_noninitialized=False))
# Verify that it is the correct submodule installed and not
# new repository initiated
eq_(set(subds.repo.get_indexed_files()),
{'test.dat', 'INFO.txt', 'test-annex.dat'})
assert_not_in('subm 1', ds.subdatasets(fulfilled=False, result_xfm='relpaths'))
assert_in('subm 1', ds.subdatasets(fulfilled=True, result_xfm='relpaths'))
# now, get the data by reinstalling with -g:
ok_(subds.repo.file_has_content('test-annex.dat') is False)
with chpwd(ds.path):
result = get(path='subm 1', dataset=os.curdir)
assert_in_results(result, path=opj(subds.path, 'test-annex.dat'))
ok_(subds.repo.file_has_content('test-annex.dat') is True)
ok_(subds.is_installed())
def test_help():
stdout, stderr = run_main(['--help'])
# Let's extract section titles:
sections = filter(re.compile('[a-zA-Z ]{4,50}:').match, stdout.split('\n'))
ok_(sections[0].startswith('Usage:')) # == Usage: nosetests [-h] if running using nose
assert_equal(sections[1:], ['Positional arguments:', 'Options:'])
def test_ssh_get_connection():
manager = SSHManager()
assert manager._socket_dir is None, \
"Should be unset upon initialization. Got %s" % str(manager._socket_dir)
c1 = manager.get_connection('ssh://localhost')
assert manager._socket_dir, "Should be set after interactions with the manager"
assert_is_instance(c1, SSHConnection)
# subsequent call returns the very same instance:
ok_(manager.get_connection('ssh://localhost') is c1)
# fail on malformed URls (meaning: our fancy URL parser can't correctly
# deal with them):
#assert_raises(ValueError, manager.get_connection, 'localhost')
# we now allow those simple specifications of host to get_connection
c2 = manager.get_connection('localhost')
assert_is_instance(c2, SSHConnection)
# but should fail if it looks like something else
assert_raises(ValueError, manager.get_connection, 'localhost/')
assert_raises(ValueError, manager.get_connection, ':localhost')
# we can do what urlparse cannot
# assert_raises(ValueError, manager.get_connection, 'someone@localhost')
# next one is considered a proper url by urlparse (netloc:'',
# path='/localhost), but eventually gets turned into SSHRI(hostname='ssh',
# path='/localhost') -- which is fair IMHO -> invalid test
# assert_raises(ValueError, manager.get_connection, 'ssh:/localhost')
manager.close()
def test_GitRepo_add(src, path):
gr = GitRepo.clone(src, path)
filename = get_most_obscure_supported_name()
with open(op.join(path, filename), 'w') as f:
f.write("File to add to git")
added = gr.add(filename)
eq_(added, {'success': True, 'file': filename})
assert_in(filename, gr.get_indexed_files(),
"%s not successfully added to %s" % (filename, path))
# uncommitted:
ok_(gr.dirty)
filename = "another.txt"
with open(op.join(path, filename), 'w') as f:
f.write("Another file to add to git")
# include committing:
added2 = gr.add(filename)
gr.commit(msg="Add two files.")
eq_(added2, {'success': True, 'file': filename})
assert_in(filename, gr.get_indexed_files(),
"%s not successfully added to %s" % (filename, path))
ok_clean_git(path)
def test_ssh_custom_identity_file():
ifile = "/tmp/dl-test-ssh-id" # Travis
if not op.exists(ifile):
raise SkipTest("Travis-specific '{}' identity file does not exist"
.format(ifile))
from datalad import cfg
try:
with patch.dict("os.environ", {"DATALAD_SSH_IDENTITYFILE": ifile}):
cfg.reload(force=True)
with swallow_logs(new_level=logging.DEBUG) as cml:
manager = SSHManager()
ssh = manager.get_connection('ssh://localhost')
cmd_out, _ = ssh("echo blah")
expected_socket = op.join(
text_type(manager.socket_dir),
get_connection_hash("localhost", identity_file=ifile,
bundled=True))
ok_(exists(expected_socket))
manager.close()
assert_in("-i", cml.out)
assert_in(ifile, cml.out)
finally:
# Prevent overridden DATALAD_SSH_IDENTITYFILE from lingering.
cfg.reload(force=True)
def test_submodule_deinit(path):
from datalad.support.annexrepo import AnnexRepo
top_repo = AnnexRepo(path, create=False)
eq_({'subm 1', '2'}, {s.name for s in top_repo.get_submodules()})
# note: here init=True is ok, since we are using it just for testing
with swallow_logs(new_level=logging.WARN) as cml:
top_repo.update_submodule('subm 1', init=True)
assert_in('Do not use update_submodule with init=True', cml.out)
top_repo.update_submodule('2', init=True)
# ok_(all([s.module_exists() for s in top_repo.get_submodules()]))
# TODO: old assertion above if non-bare? (can't use "direct mode" in test_gitrepo)
# Alternatively: New testrepo (plain git submodules) and have a dedicated
# test for annexes in addition
ok_(all([GitRepo.is_valid_repo(op.join(top_repo.path, s.path))
for s in top_repo.get_submodules()]))
# modify submodule:
with open(op.join(top_repo.path, 'subm 1', 'file_ut.dat'), "w") as f:
f.write("some content")
assert_raises(CommandError, top_repo.deinit_submodule, 'sub1')
# using force should work:
top_repo.deinit_submodule('subm 1', force=True)
ok_(not top_repo.repo.submodule('subm 1').module_exists())
def test_add_subdataset(path, other):
subds = create(op.join(path, 'dir'), force=True)
ds = create(path, force=True)
ok_(subds.repo.dirty)
ok_(ds.repo.dirty)
assert_not_in('dir', ds.subdatasets(result_xfm='relpaths'))
# "add everything in subds to subds"
save(dataset=subds.path)
assert_repo_status(subds.path)
assert_not_in('dir', ds.subdatasets(result_xfm='relpaths'))
# but with a base directory we add the dataset subds as a subdataset
# to ds
res = ds.save(subds.path)
assert_in_results(res, action="add", path=subds.path, refds=ds.path)
assert_in('dir', ds.subdatasets(result_xfm='relpaths'))
# create another one
other = create(other)
# install into superdataset, but don't add
other_clone = install(source=other.path, path=op.join(ds.path, 'other'))
# little dance to get the revolution-type dataset
other_clone = Dataset(other_clone.path)
ok_(other_clone.is_installed)
assert_not_in('other', ds.subdatasets(result_xfm='relpaths'))
# now add, it should pick up the source URL
ds.save('other')
# and that is why, we can reobtain it from origin
ds.uninstall('other')
ok_(not other_clone.is_installed())
ds.get('other')
ok_(other_clone.is_installed())
def test_notclone_known_subdataset(src, path):
# get the superdataset:
ds = clone(src, path,
result_xfm='datasets', return_type='item-or-list')
# subdataset not installed:
subds = Dataset(opj(path, 'subm 1'))
assert_false(subds.is_installed())
assert_in('subm 1', ds.subdatasets(fulfilled=False, result_xfm='relpaths'))
assert_not_in('subm 1', ds.subdatasets(fulfilled=True, result_xfm='relpaths'))
# clone is not meaningful
res = ds.clone('subm 1', on_failure='ignore')
assert_status('error', res)
assert_message('Failed to clone from any candidate source URL. '
'Encountered errors per each url were: %s',
res)
# get does the job
res = ds.get(path='subm 1', get_data=False)
assert_status('ok', res)
ok_(subds.is_installed())
ok_(AnnexRepo.is_valid_repo(subds.path, allow_noninitialized=False))
# Verify that it is the correct submodule installed and not
# new repository initiated
eq_(set(subds.repo.get_indexed_files()),
{'test.dat', 'INFO.txt', 'test-annex.dat'})
assert_not_in('subm 1', ds.subdatasets(fulfilled=False, result_xfm='relpaths'))
assert_in('subm 1', ds.subdatasets(fulfilled=True, result_xfm='relpaths'))
def _test_match_basic(matcher, query):
extracts = dict(
xpaths={'text': 'text()'},
csss={'favorite': '.class1::text'}
)
m = matcher(query, **extracts)
mg = m(dict(response="<div></div>"))
ok_(inspect.isgenerator(mg))
eq_(list(mg), []) # there should be no hits
mg = m(dict(response=sample1.response))
ok_(inspect.isgenerator(mg))
hits = list(mg)
eq_(len(hits), 3)
for hit, a_html, a_text, class1_text in zip(
hits, sample1.a_htmls, sample1.a_texts, sample1.class1_texts):
ok_(hit['response'])
eq_(hit['match'], a_html)
eq_(hit['text'], a_text)
eq_(hit.get('favorite', None), class1_text)
m = matcher(query, min_count=4, **extracts)
mg = m(dict(response=sample1.response))
ok_(inspect.isgenerator(mg))
assert_raises(ValueError, list, mg)
m = matcher(query, max_count=2, **extracts)
mg = m(dict(response=sample1.response))
ok_(inspect.isgenerator(mg))
assert_raises(ValueError, list, mg)
def test_ssh_open_close(tfile1):
manager = SSHManager()
path = opj(manager.socket_dir, get_connection_hash('localhost'))
# TODO: facilitate the test when it didn't exist
existed_before = exists(path)
print("%s existed: %s" % (path, existed_before))
c1 = manager.get_connection('ssh://localhost')
c1.open()
# control master exists for sure now
ok_(exists(path))
# use connection to execute remote command:
local_home = os.path.expanduser('~')
# we list explicitly local HOME since we override it in module_setup
out, err = c1('ls -a %r' % local_home)
remote_ls = [entry for entry in out.splitlines()
if entry != '.' and entry != '..']
local_ls = os.listdir(local_home)
eq_(set(remote_ls), set(local_ls))
# now test for arguments containing spaces and other pleasant symbols
out, err = c1('ls -l {}'.format(sh_quote(tfile1)))
assert_in(tfile1, out)
eq_(err, '')
c1.close()
# control master doesn't exist anymore:
ok_(exists(path) == existed_before)
def test_clone_isnot_recursive(src, path_nr, path_r):
ds = clone(src, path_nr, result_xfm='datasets', return_type='item-or-list')
ok_(ds.is_installed())
# check nothin is unintentionally installed
subdss = ds.subdatasets(recursive=True)
assert_result_count(subdss, len(subdss), state='absent')
# this also means, subdatasets to be listed as not fulfilled:
eq_(set(ds.subdatasets(recursive=True, fulfilled=False, result_xfm='relpaths')),
{'subm 1', '2'})
def test_create_test_dataset():
# rudimentary smoke test
from datalad.api import create_test_dataset
with swallow_logs(), swallow_outputs():
dss = create_test_dataset(spec='2/1-2')
ok_(5 <= len(dss) <= 7) # at least five - 1 top, two on top level, 1 in each
for ds in dss:
ok_clean_git(ds, annex=None) # some of them are annex but we just don't check
ok_(len(glob(opj(ds, 'file*'))))
def test_GitRepo_equals(path1, path2):
repo1 = GitRepo(path1)
repo2 = GitRepo(path1)
ok_(repo1 == repo2)
eq_(repo1, repo2)
repo2 = GitRepo(path2)
neq_(repo1, repo2)
ok_(repo1 != repo2)
def test_install_dataset_from_just_source(url, path):
with chpwd(path, mkdir=True):
ds = install(source=url)
ok_startswith(ds.path, path)
ok_(ds.is_installed())
ok_(GitRepo.is_valid_repo(ds.path))
ok_clean_git(ds.path, annex=None)
assert_in('INFO.txt', ds.repo.get_indexed_files())
def test_autoresolve_multiple_datasets(src, path):
with chpwd(path):
ds1 = install('ds1', source=src)
ds2 = install('ds2', source=src)
results = get([opj('ds1', 'test-annex.dat')] + glob(opj('ds2', '*.dat')))
# each ds has one file
eq_(len(results), 2)
ok_(ds1.repo.file_has_content('test-annex.dat') is True)
ok_(ds2.repo.file_has_content('test-annex.dat') is True)
def test_clone_dataset_from_just_source(url, path):
with chpwd(path, mkdir=True):
ds = clone(url, result_xfm='datasets', return_type='item-or-list')
ok_startswith(ds.path, path)
ok_(ds.is_installed())
ok_(GitRepo.is_valid_repo(ds.path))
ok_clean_git(ds.path, annex=None)
assert_in('INFO.txt', ds.repo.get_indexed_files())
def test_install_dataset_from_instance(src, dst):
origin = Dataset(src)
clone = install(source=origin, path=dst)
assert_is_instance(clone, Dataset)
ok_startswith(clone.path, dst)
ok_(clone.is_installed())
ok_(GitRepo.is_valid_repo(clone.path))
ok_clean_git(clone.path, annex=None)
assert_in('INFO.txt', clone.repo.get_indexed_files())
def test_inject(path):
ds = Dataset(path).create(force=True)
ok_(ds.repo.is_dirty())
list(run_command("nonsense command",
dataset=ds,
inject=True,
extra_info={"custom_key": "custom_field"}))
msg = ds.repo.format_commit("%B")
assert_in("custom_key", msg)
assert_in("nonsense command", msg)
def test_no_interaction_with_untracked_content(path):
# extracted from what was a metadata test originally
ds = Dataset(opj(path, 'origin')).create(force=True)
create_tree(ds.path, {'sub': {'subsub': {'dat': 'lots of data'}}})
subds = ds.create('sub', force=True)
subds.remove(opj('.datalad', 'config'), if_dirty='ignore')
ok_(not exists(opj(subds.path, '.datalad', 'config')))
# this will only work, if `remove` didn't do anything stupid and
# caused all content to be saved
subds.create('subsub', force=True)
def test_is_url():
ok_(is_url('file://localhost/some'))
ok_(is_url('http://localhost'))
ok_(is_url('ssh://me@localhost'))
# in current understanding it is indeed a url but an 'ssh', implicit=True, not just
# a useless scheme=weired with a hope to point to a netloc
with swallow_logs():
ok_(is_url('weired://'))
nok_(is_url('relative'))
nok_(is_url('/absolute'))
ok_(is_url('like@sshlogin')) # actually we do allow ssh:implicit urls ATM
nok_(is_url(''))
nok_(is_url(' '))
nok_(is_url(123)) # stuff of other types wouldn't be considered a URL
# we can pass RI instance directly
ok_(is_url(RI('file://localhost/some')))
nok_(is_url(RI('relative')))
def test_target_ssh_simple(origin, src_path, target_rootpath):
# prepare src
source = install(src_path,
source=origin,
result_xfm='datasets',
return_type='item-or-list')
target_path = opj(target_rootpath, "basic")
with swallow_logs(new_level=logging.ERROR) as cml:
create_sibling(dataset=source,
name="local_target",
sshurl="ssh://*****:*****@with_testsui(responses=["yes"])
def interactive_assert_create_sshwebserver():
assert_create_sshwebserver(
dataset=source,
name="local_target",
sshurl="ssh://localhost" + target_path,
publish_by_default='master',
existing='replace',
ui=True,
)
interactive_assert_create_sshwebserver()
eq_("ssh://localhost" + urlquote(target_path),
source.repo.get_remote_url("local_target"))
ok_(source.repo.get_remote_url("local_target", push=True) is None)
# ensure target tree actually replaced by source
assert_false(exists(opj(target_path, 'random')))
if src_is_annex:
lclcfg = AnnexRepo(src_path).config
eq_(lclcfg.get('remote.local_target.annex-ignore'), 'false')
# valid uuid
eq_(lclcfg.get('remote.local_target.annex-uuid').count('-'), 4)
# should be added too, even if URL matches prior state
eq_(lclcfg.get('remote.local_target.push'), 'master')
# again, by explicitly passing urls. Since we are on localhost, the
# local path should work:
cpkwargs = dict(
dataset=source,
name="local_target",
sshurl="ssh://localhost",
target_dir=target_path,
target_url=target_path,
target_pushurl="ssh://localhost" + target_path,
ui=True,
)
@with_testsui(responses=['yes'])
def interactive_assert_create_sshwebserver():
assert_create_sshwebserver(existing='replace', **cpkwargs)
interactive_assert_create_sshwebserver()
if src_is_annex:
target_description = AnnexRepo(target_path,
create=False).get_description()
eq_(target_description, target_path)
eq_(target_path, source.repo.get_remote_url("local_target"))
eq_("ssh://localhost" + target_path,
source.repo.get_remote_url("local_target", push=True))
assert_publish_with_ui(target_path)
# now, push should work:
publish(dataset=source, to="local_target")
# and we should be able to 'reconfigure'
def process_digests_mtimes(digests, mtimes):
# it should have triggered a hook, which would have created log and metadata files
check_metadata = False
for part in 'logs', 'metadata':
metafiles = [
k for k in digests
if k.startswith(_path_('.git/datalad/%s/' % part))
]
# This is in effect ONLY if we have "compatible" datalad installed on remote
# end. ATM we don't have easy way to guarantee that AFAIK (yoh),
# so let's not check/enforce (TODO)
# assert(len(metafiles) >= 1) # we might have 2 logs if timestamps do not collide ;)
# Let's actually do it to some degree
if part == 'logs':
# always should have those:
assert (len(metafiles) >= 1)
with open(opj(target_path, metafiles[0])) as f:
if 'no datalad found' not in f.read():
check_metadata = True
if part == 'metadata':
eq_(len(metafiles), bool(check_metadata))
for f in metafiles:
digests.pop(f)
mtimes.pop(f)
# and just pop some leftovers from annex
for f in list(digests):
if f.startswith('.git/annex/mergedrefs'):
digests.pop(f)
mtimes.pop(f)
orig_digests, orig_mtimes = get_mtimes_and_digests(target_path)
process_digests_mtimes(orig_digests, orig_mtimes)
import time
time.sleep(0.1) # just so that mtimes change
assert_create_sshwebserver(existing='reconfigure', **cpkwargs)
digests, mtimes = get_mtimes_and_digests(target_path)
process_digests_mtimes(digests, mtimes)
assert_dict_equal(orig_digests,
digests) # nothing should change in terms of content
# but some files should have been modified
modified_files = {
k
for k in mtimes if orig_mtimes.get(k, 0) != mtimes.get(k, 0)
}
# collect which files were expected to be modified without incurring any changes
ok_modified_files = {
_path_('.git/hooks/post-update'),
'index.html',
}
ok_modified_files.add(_path_('.git/config'))
ok_modified_files.update(
{f
for f in digests if f.startswith(_path_('.git/datalad/web'))})
# it seems that with some recent git behavior has changed a bit
# and index might get touched
if _path_('.git/index') in modified_files:
ok_modified_files.add(_path_('.git/index'))
ok_(modified_files.issuperset(ok_modified_files))
def _test_target_ssh_inherit(standardgroup, ui, use_ssh, src_path,
target_path):
ds = Dataset(src_path).create()
if use_ssh:
target_url = 'localhost:%s' % target_path
else:
target_url = target_path
remote = "magical"
# for the test of setting a group, will just smoke test while using current
# user's group
ds.create_sibling(target_url,
name=remote,
shared='group',
group=os.getgid(),
ui=ui) # not doing recursively
if standardgroup:
ds.repo.set_preferred_content('wanted', 'standard', remote)
ds.repo.set_preferred_content('group', standardgroup, remote)
ds.publish(to=remote)
# now a month later we created a new subdataset... a few of the nested ones
# A known hiccup happened when there
# is also subsub ds added - we might incorrectly traverse and not prepare
# sub first for subsub to inherit etc
parent_ds = ds
subdss = []
nlevels = 2 # gets slow: 1 - 43 sec, 2 - 49 sec , 3 - 69 sec
for levels in range(nlevels):
subds = parent_ds.create('sub')
create_tree(subds.path, {'sub.dat': 'lots of data'})
parent_ds.save('sub', recursive=True)
ok_file_under_git(subds.path, 'sub.dat', annexed=True)
parent_ds = subds
subdss.append(subds)
target_subdss = [
Dataset(opj(*([target_path] + ['sub'] * (i + 1))))
for i in range(nlevels)
]
# since we do not have yet/thus have not used an option to record to publish
# to that sibling by default (e.g. --set-upstream), if we run just ds.publish
# -- should fail
assert_result_count(
ds.publish(on_failure='ignore'),
1,
status='impossible',
message=
'No target sibling configured for default publication, please specific via --to'
)
ds.publish(
to=remote) # should be ok, non recursive; BUT it (git or us?) would
# create an empty sub/ directory
assert_postupdate_hooks(target_path, installed=ui)
for target_sub in target_subdss:
ok_(not target_sub.is_installed()) # still not there
res = ds.publish(to=remote, recursive=True, on_failure='ignore')
assert_result_count(res, 1 + len(subdss))
assert_status(('error', 'notneeded'), res)
assert_result_count(res,
len(subdss),
status='error',
message=("Unknown target sibling '%s' for publication",
'magical'))
# Finally publishing with inheritance
ds.publish(to=remote, recursive=True, missing='inherit')
assert_postupdate_hooks(target_path, installed=ui)
def check_dss():
# we added the remote and set all the
for subds in subdss:
eq_(subds.repo.get_preferred_content('wanted', remote),
'standard' if standardgroup else '')
eq_(subds.repo.get_preferred_content('group', remote),
standardgroup or '')
for target_sub in target_subdss:
ok_(target_sub.is_installed()) # it is there now
eq_(target_sub.repo.config.get('core.sharedrepository'), '1')
# and we have transferred the content
if standardgroup and standardgroup == 'backup':
# only then content should be copied
ok_file_has_content(opj(target_sub.path, 'sub.dat'),
'lots of data')
else:
# otherwise nothing is copied by default
assert_false(target_sub.repo.file_has_content('sub.dat'))
check_dss()
# and it should be ok to reconfigure the full hierarchy of datasets
# while "inheriting". No URL must be specified, and we must not blow
# but just issue a warning for the top level dataset which has no super,
# so cannot inherit anything - use case is to fixup/establish the full
# hierarchy on the remote site
with swallow_logs(logging.WARNING) as cml:
out = ds.create_sibling(None,
name=remote,
existing="reconfigure",
inherit=True,
ui=ui,
recursive=True)
eq_(len(out), 1 + len(subdss))
assert_in("Cannot determine super dataset", cml.out)
check_dss()
def test_local_path_target_dir(path):
path = Path(path)
ds_main = Dataset(path / "main").create()
ds_main.create_sibling(name="abspath-targetdir",
sshurl=str(path / "a"),
target_dir="tdir")
ok_((path / "a" / "tdir").exists())
ds_main.create_sibling(name="relpath-bound-targetdir",
sshurl=os.path.relpath(str(path / "b"),
ds_main.path),
target_dir="tdir")
ok_((path / "b" / "tdir").exists())
with chpwd(path):
create_sibling(dataset=ds_main.path,
name="relpath-unbound-targetdir",
sshurl="c",
target_dir="tdir")
ok_((path / "c" / "tdir").exists())
ds_main.create("subds")
ds_main.create_sibling(name="rec-plain-targetdir",
recursive=True,
sshurl=str(path / "d"),
target_dir="tdir")
ok_((path / "d" / "tdir" / "subds").exists())
ds_main.create_sibling(name="rec-template-targetdir",
recursive=True,
sshurl=str(path / "e"),
target_dir="d%RELNAME")
ok_((path / "e" / "d").exists())
ok_((path / "e" / "d-subds").exists())
def test_same_website():
ok_(same_website("http://a.b", "http://a.b/2014/01/xxx/"))
ok_(same_website("http://a.b/page/2/", "http://a.b/2014/01/xxx/"))
ok_(same_website("https://a.b/page/2/", "http://a.b/2014/01/xxx/"))
ok_(same_website("http://a.b/page/2/", "https://a.b/2014/01/xxx/"))
