def __init__(self):
if not self.dir:
self.dir = pathlib.Path(inspect.getfile(self.__class__)).parent
self.dir = DataDir(self.dir)
md = self.dir / 'metadata.json'
self.metadata = self.metadata_cls.from_file(
md) if md.exists() else self.metadata_cls()
self.metadata.id = self.id
class Dataset(object):
"""
A cldfbench dataset ties together
- `raw` data, to be used as source for the
- `cldf` data, which is created using config data from
- `etc`.
To use the cldfbench infrastructure, one should sub-class `Dataset`.
cldfbench supports the following workflow:
- a `download` command populates a `Dataset`'s `raw` directory.
- a `makecldf` command (re)creates the CLDF dataset in `cldf`.
"""
dir = None
id = None
metadata_cls = Metadata
def __init__(self):
if not self.dir:
self.dir = pathlib.Path(inspect.getfile(self.__class__)).parent
self.dir = DataDir(self.dir)
md = self.dir / 'metadata.json'
self.metadata = self.metadata_cls.from_file(
md) if md.exists() else self.metadata_cls()
self.metadata.id = self.id
def __str__(self):
return '{0.__class__.__name__} "{0.id}" at {1}'.format(
self, self.dir.resolve())
def cldf_specs(self):
"""
A `Dataset` must declare all CLDF datasets that are derived from it.
:return: A single `CLDFSpec` instance, or a `dict`, mapping names to `CLDFSpec` \
instances, where the name will be used by `cldf_reader`/`cldf_writer` to look up \
the spec.
"""
return CLDFSpec(dir=self.cldf_dir)
@property
def cldf_specs_dict(self):
"""
Turn cldf_specs into a `dict` for simpler lookup.
:return: `dict` mapping lookup keys to `CLDFSpec` instances.
"""
specs = self.cldf_specs()
if isinstance(specs, CLDFSpec):
return {None: specs}
assert isinstance(specs, dict)
return specs
@lazyproperty
def cldf_dir(self):
return self.dir / 'cldf'
@lazyproperty
def raw_dir(self):
return self.dir / 'raw'
@lazyproperty
def etc_dir(self):
return self.dir / 'etc'
def cldf_writer(self, args, cldf_spec=None, clean=True):
"""
:param args:
:param cldf_spec: Key of the relevant `CLDFSpec` in `Dataset.cldf_specs`
:param clean: `bool` flag signaling whether to clean the CLDF dir before writing. \
Note that `False` must be passed for subsequent calls to `cldf_writer` in case the \
spec re-uses a directory.
:return: a `cldf_spec.writer_cls` instance, for write-access to CLDF data. \
This method should be used in a with-statement, and will then return a `CLDFWriter` with \
an empty working directory.
"""
if not isinstance(cldf_spec, CLDFSpec):
cldf_spec = self.cldf_specs_dict[cldf_spec]
return cldf_spec.get_writer(args=args, dataset=self, clean=clean)
def cldf_reader(self, cldf_spec=None):
"""
:param cldf_spec:
:return: a `pycldf.Dataset` instance, for read-access to the CLDF data.
"""
if not isinstance(cldf_spec, CLDFSpec):
cldf_spec = self.cldf_specs_dict[cldf_spec]
return cldf_spec.get_dataset()
@lazyproperty
def repo(self):
try:
return Repository(self.dir)
except ValueError: # pragma: no cover
return
#
# Workflow commands are implemented with two methods for each command:
# - cmd_<command>: The implementation of the command, typically overwritten by datasets.
# - _cmd_<command>: An (optional) wrapper providing setup and teardown functionality, calling
# cmd_<command> in between.
#
# Workflow commands must accept an `argparse.Namespace` as sole positional argument.
#
def _cmd_download(self, args):
self.raw_dir.mkdir(exist_ok=True)
self.cmd_download(args)
(self.raw_dir / 'README.md').write_text(
'Raw data downloaded {0}'.format(datetime.utcnow().isoformat()),
encoding='utf8')
def cmd_download(self, args):
args.log.warning('cmd_{0} not implemented for dataset {1}'.format(
'download', self.id))
return NOOP
def _cmd_readme(self, args):
if self.metadata:
self.dir.joinpath('README.md').write_text(self.cmd_readme(args),
encoding='utf8')
def cmd_readme(self, args):
return self.metadata.markdown() if self.metadata else ''
def _cmd_makecldf(self, args):
specs = list(self.cldf_specs_dict.values())
if len(specs) == 1:
# There's only one CLDF spec! We instantiate the writer now and inject it into `args`:
with self.cldf_writer(args, cldf_spec=specs[0]) as writer:
args.writer = writer
self.cmd_makecldf(args)
else:
self.cmd_makecldf(args)
if self.metadata and self.metadata.known_license:
legalcode = self.metadata.known_license.legalcode
if legalcode:
(self.dir / 'LICENSE').write_text(legalcode, encoding='utf8')
def cmd_makecldf(self, args):
"""
:param args: An `argparse.Namespace` including attributes:
- `writer`: `CLDFWriter` instance
"""
args.log.warning('cmd_{0} not implemented for dataset {1}'.format(
'makecldf', self.id))
return NOOP
def run(args):
ds = get_dataset(args)
ds_cldf = ds.cldf_reader()
release_dir = args.out / '{0}_{1}'.format(ds.id, MEDIA)
if ds_cldf.get('media.csv', None) is None: # pragma: no cover
args.log.error('Dataset has no media.csv')
raise ParserError
if args.parent_doi and not Zenodo.DOI_PATTERN.match(args.parent_doi):
args.log.error('Invalid passed DOI')
raise ParserError
if args.update_zenodo:
if not release_dir.exists():
args.log.error(
'"{0}" not found -- run --create-release first?'.format(
release_dir))
raise ParserError
if not (release_dir / ZENODO_FILE_NAME).exists():
args.log.error(
'"{0}" not found -- run --create-release first?'.format(
release_dir / ZENODO_FILE_NAME))
raise ParserError
if args.create_release:
args.log.error(
'You cannot create the release and update zenodo at the same time.'
)
raise ParserError
if args.create_release:
if not args.parent_doi:
args.log.error(
'The corresponding DOI is required (via --parent-doi).')
raise ParserError
mime_types = None
if args.mimetype:
mime_types = [m.strip() for m in nfilter(args.mimetype.split(','))]
if args.list:
size = collections.Counter()
number = collections.Counter()
else:
media_dir = args.out / MEDIA
media_dir.mkdir(exist_ok=True)
media = []
if not args.update_zenodo:
used_file_extensions = set()
with UnicodeWriter(media_dir /
INDEX_CSV if not args.list else None) as w:
for i, row in enumerate(
tqdm.tqdm([r for r in ds_cldf['media.csv']],
desc='Getting {0} items'.format(MEDIA))):
url = ds_cldf.get_row_url('media.csv', row)
if isinstance(url, rfc3986.URIReference):
url = url.normalize().unsplit()
row['URL'] = url
f_ext = url.split('.')[-1].lower()
if args.debug and i > 500:
break
if (mime_types is None) or f_ext in mime_types\
or any(row['mimetype'].startswith(x) for x in mime_types):
if args.list:
m = '{0} ({1})'.format(row['mimetype'], f_ext)
size[m] += int(row['size'])
number.update([m])
else:
used_file_extensions.add(f_ext.lower())
d = media_dir / row['ID'][:2]
d.mkdir(exist_ok=True)
fn = '.'.join([row['ID'], f_ext])
target = d / fn
row['local_path'] = pathlib.Path(row['ID'][:2]) / fn
if i == 0:
w.writerow(row)
w.writerow(row.values())
media.append(target)
if (not target.exists()) or md5(target) != row['ID']:
_create_download_thread(url, target)
if args.list:
for k, v in size.most_common():
print('\t'.join([k.ljust(20), str(number[k]), format_size(v)]))
return
# Waiting for the download threads to finish
if 'download_threads' in globals():
for t in download_threads:
t.join()
if args.create_release:
assert media_dir.exists(), 'No folder "{0}" found in {1}'.format(
MEDIA, media_dir.resolve())
release_dir.mkdir(exist_ok=True)
media.append(media_dir / INDEX_CSV)
try:
zipf = zipfile.ZipFile(str(release_dir / '{0}.zip'.format(MEDIA)),
'w', zipfile.ZIP_DEFLATED)
fp = args.out
for f in tqdm.tqdm(media, desc='Creating {0}.zip'.format(MEDIA)):
zipf.write(str(f), str(os.path.relpath(str(f), str(fp))))
zipf.close()
except Exception as e:
args.log.error(e)
raise
def _contrib(d):
return {
k: v
for k, v in d.items()
if k in {'name', 'affiliation', 'orcid', 'type'}
}
version_v = git_describe('.').split('-')[0]
version = version_v.replace('v', '')
git_url = [r for r in ds.repo.repo.remotes
if r.name == 'origin'][0].url.replace('.git', '')
with jsonlib.update(release_dir / ZENODO_FILE_NAME,
indent=4,
default=collections.OrderedDict()) as md:
contribs = ds.dir / 'CONTRIBUTORS.md'
creators, contributors = get_creators_and_contributors(
contribs.read_text(
encoding='utf8') if contribs.exists() else '',
strict=False)
if creators:
md['creators'] = [_contrib(p) for p in creators]
if contributors:
md['contributors'] = [_contrib(p) for p in contributors]
communities = [r["identifier"] for r in md.get("communities", [])] + \
[c.strip() for c in nfilter(args.communities.split(','))] + \
COMMUNITIES
if communities and not args.debug:
md['communities'] = [{
"identifier": community_id
} for community_id in sorted(set(communities))]
md.update({
'title':
'{0} {1} Files'.format(ds.metadata.title, MEDIA.title()),
'access_right':
'open',
'keywords':
sorted(set(md.get('keywords', []) + ['linguistics'])),
'upload_type':
'dataset',
'publication_date':
datetime.today().strftime('%Y-%m-%d'),
'version':
version,
'related_identifiers': [
{
'scheme': 'url',
'identifier':
'{0}/tree/{1}'.format(git_url, version_v),
'relation': 'isSupplementTo'
},
],
})
if args.parent_doi:
md['related_identifiers'].append({
'scheme': 'doi',
'identifier': args.parent_doi,
'relation': 'isPartOf'
})
supplement_to = " - Supplement to dataset " \
"<a href='https://doi.org/{0}'>{1}</a> ".format(
args.parent_doi, ds.metadata.title) # noqa: E122
if ds.metadata.url:
md['related_identifiers'].append({
'scheme':
'url',
'identifier':
ds.metadata.url,
'relation':
'isAlternateIdentifier'
})
formats = ', '.join(sorted(used_file_extensions))
descr = '<br /><br />' + ds.metadata.description if ds.metadata.description else ''
online_url, online = '', ''
if ds.metadata.url:
online_url = ds.metadata.url
online = "<br /><br />Available online at: <a href='{0}'>{0}</a>".format(
online_url)
md['description'] = html.escape(
DESCRIPTION.format(
url=online_url,
formats=' ({0})'.format(formats) if formats else '',
title=md['title'],
supplement_to=supplement_to,
descr=descr,
online=online))
license_md = ''
if ds.metadata.zenodo_license:
md['license'] = {'id': ds.metadata.zenodo_license}
license_md = LICENCE.format(ds.metadata.zenodo_license)
DataDir(release_dir).write(
'README.md',
README.format(
title=md['title'],
doi='https://doi.org/{0}'.format(args.parent_doi),
ds_title=ds.metadata.title,
license=license_md,
formats=' ({0})'.format(formats) if formats else '',
media=MEDIA,
index=INDEX_CSV))
if args.update_zenodo:
md = {}
md.update(jsonlib.load(release_dir / ZENODO_FILE_NAME))
if args.debug:
api_url = API_URL_SANDBOX
access_token = os.environ.get('ZENODO_SANDBOX_ACCESS_TOKEN')
else:
api_url = API_URL
access_token = ACCESS_TOKEN
zenodo_url = api_url.replace('api/', '')
args.log.info('Updating Deposit ID {0} on {1} with:'.format(
args.update_zenodo, zenodo_url))
api = Zenodo(api_url=api_url, access_token=access_token)
try:
rec = api.record_from_id('{0}record/{1}'.format(
zenodo_url, args.update_zenodo))
except Exception as e:
args.log.error(
'Check connection and credentials for accessing Zenodo.\n{0}'.
format(e))
return
latest_version = rec.links['latest'].split('/')[-1]
if latest_version != args.update_zenodo:
args.log.warn(
'Passed deposit ID does not refer to latest version {0}!'.
format(latest_version))
args.log.info(' DOI: ' + rec.metadata.doi)
args.log.info(' Title: ' + rec.metadata.title)
args.log.info(' Version: ' + rec.metadata.version)
args.log.info(' Date: ' + rec.metadata.publication_date)
args.log.info(' Files: ' + ', '.join([f.key for f in rec.files]))
p = input("Proceed? [y/N]: ")
if p.lower() == 'y':
dep = api.update_deposit(args.update_zenodo, **md)
if dep.state != PUBLISHED:
api.publish_deposit(dep)
args.log.info('Updated successfully')
def run(args):
ds = Dataset().cldf_reader()
release_dir = args.out / '{0}_audio'.format(Dataset().id)
zenodo_file_name = 'zenodo.json'
if args.list:
size = collections.Counter()
number = collections.Counter()
else:
f2c = {r['ID']: r['Parameter_ID'] for r in ds['FormTable']}
audio = args.out / 'audio'
audio.mkdir(exist_ok=True)
if not args.update_zenodo:
for row in tqdm.tqdm([r for r in ds['media.csv']]):
if args.list:
size[row['mimetype']] += int(row['size'])
number.update([row['mimetype']])
else:
d = audio / f2c[row['Form_ID']]
d.mkdir(exist_ok=True)
url = ds.get_row_url('media.csv', row)
target = d / '{}.{}'.format(row['ID'], url.split('.')[-1])
if (not target.exists()) or md5(target) != row['ID']:
if (args.mimetype is None) or target.suffix.endswith(
args.mimetype):
create_download_thread(url, target)
if args.list:
for k, v in size.most_common():
print('\t'.join([k, str(number[k]), format_size(v)]))
if args.create_release:
assert audio.exists(), 'No folder "audio" found in {0}'.format(
audio.resolve())
release_dir.mkdir(exist_ok=True)
args.log.info('creating audio ZIP archive per parameter folder ...')
try:
zipf = zipfile.ZipFile(str(release_dir / 'audio.zip'), 'w',
zipfile.ZIP_DEFLATED)
fp = args.out
for root, dirs, files in tqdm.tqdm(os.walk(audio)):
for f in files:
if not f.startswith('.') and not f.startswith('__')\
and ((args.mimetype is None) or f.endswith(args.mimetype)):
zipf.write(os.path.join(root, f),
os.path.relpath(os.path.join(root, f), fp))
zipf.close()
except Exception as e:
args.log.error(e)
raise
def contrib(d):
return {
k: v
for k, v in d.items()
if k in {'name', 'affiliation', 'orcid', 'type'}
}
with jsonlib.update(release_dir / zenodo_file_name,
indent=4,
default=collections.OrderedDict()) as md:
contribs = Dataset().dir / 'CONTRIBUTORS.md'
creators, contributors = get_creators_and_contributors(
contribs.read_text(
encoding='utf8') if contribs.exists() else '',
strict=False)
if creators:
md['creators'] = [contrib(p) for p in creators]
if contributors:
md['contributors'] = [contrib(p) for p in contributors]
if COMMUNITIES:
md['communities'] = [{
'id': community_id
} for community_id in COMMUNITIES]
md.update({
'title':
'{0} Audio Files'.format(Dataset().metadata.title),
'access_right':
'open',
'keywords':
sorted(set(md.get('keywords', []) + ['linguistics'])),
'upload_type':
'video',
'version':
VERSION,
'related_identifiers': [
{
'scheme': 'doi',
'identifier': '10.5281/zenodo.4309141',
'relation': 'isPartOf'
},
{
'scheme':
'url',
'identifier':
'{0}{1}/tree/v{2}'.format(GITHUB_PREFIX,
Dataset().id, VERSION),
'relation':
'isSupplementTo'
},
],
})
if Dataset().metadata.url:
md['related_identifiers'].append({
'scheme':
'url',
'identifier':
Dataset().metadata.url,
'relation':
'isAlternateIdentifier'
})
md['description'] = html.escape(
DESCRIPTION.format(
GITHUB_PREFIX,
Dataset().id,
Dataset().metadata.url if Dataset().metadata.url else '',
VERSION))
license_md = ''
if Dataset().metadata.zenodo_license:
md['license'] = {'id': Dataset().metadata.zenodo_license}
license_md = LISENCE.format(Dataset().metadata.zenodo_license)
DataDir(release_dir).write(
'README.md',
RELEASE_NOTE.format(md['title'], GITHUB_PREFIX,
Dataset().id,
Dataset().metadata.title, license_md))
if args.update_zenodo:
assert release_dir.exists()
assert (release_dir / zenodo_file_name).exists()
md = {}
md.update(jsonlib.load(release_dir / zenodo_file_name))
api_url = API_URL
zenodo_url = api_url.replace('api/', '')
args.log.info('Updating Deposit ID {0} on {1} with:'.format(
args.update_zenodo, zenodo_url))
api = Zenodo(api_url=api_url, access_token=ACCESS_TOKEN)
rec = api.record_from_id('{0}record/{1}'.format(
zenodo_url, args.update_zenodo))
args.log.info(' DOI: ' + rec.metadata.doi)
args.log.info(' Title: ' + rec.metadata.title)
args.log.info(' Date: ' + rec.metadata.publication_date)
args.log.info(' Files: ' + ', '.join([f.key for f in rec.files]))
p = input("Proceed? [y/N]: ")
if p.lower() == 'y':
dep = api.update_deposit(args.update_zenodo, **md)
if dep.state != zenodoclient.models.PUBLISHED:
api.publish_deposit(dep)
args.log.info('Updated successfully')
class Dataset(object):
"""
A cldfbench dataset ties together
- `raw` data, to be used as source for the
- `cldf` data, which is created using config data from
- `etc`.
To use the cldfbench infrastructure, one should sub-class `Dataset`.
cldfbench supports the following workflow:
- a `download` command populates a `Dataset`'s `raw` directory.
- a `makecldf` command (re)creates the CLDF dataset in `cldf`.
The following class attributes are supposed to be overwritten by subclasses:
:ivar dir: `pathlib.Path` pointing to the root directory of the dataset.
:ivar id: A `str` identifier for the dataset. No assumption about uniqueness properties of \
this identifier is made.
:ivar metadata_cls: Subclass of :class:`Metadata` (or :class:`Metadata` if not overwritten)
"""
dir = None
id = None
metadata_cls = Metadata
def __init__(self):
if not self.dir:
self.dir = pathlib.Path(inspect.getfile(self.__class__)).parent
self.dir = DataDir(self.dir)
md = self.dir / 'metadata.json'
self.metadata = self.metadata_cls.from_file(
md) if md.exists() else self.metadata_cls()
self.metadata.id = self.id
def __str__(self):
return '{0.__class__.__name__} "{0.id}" at {1}'.format(
self, self.dir.resolve())
@lazyproperty
def cldf_dir(self) -> DataDir:
"""
Directory where CLDF data generated from the Dataset will be stored (unless specified
differently by a :class:`CLDFSpec`).
"""
return self.dir / 'cldf'
@lazyproperty
def raw_dir(self) -> DataDir:
"""
Directory where cldfbench expects the raw or source data.
"""
return self.dir / 'raw'
@lazyproperty
def etc_dir(self) -> DataDir:
"""
Directory where cldfbench expects additional configuration or metadata.
"""
return self.dir / 'etc'
def cldf_specs(self) -> typing.Union[CLDFSpec, typing.Dict[str, CLDFSpec]]:
"""
A `Dataset` must declare all CLDF datasets that are derived from it.
:return: A single :class:`CLDFSpec` instance, or a `dict`, mapping names to `CLDFSpec` \
instances, where the name will be used by `cldf_reader`/`cldf_writer` to look up \
the spec.
"""
return CLDFSpec(dir=self.cldf_dir)
@property
def cldf_specs_dict(
self) -> typing.Dict[typing.Union[str, None], CLDFSpec]:
"""
Turn :meth:`cldf_specs` into a `dict` for simpler lookup.
:return: `dict` mapping lookup keys to `CLDFSpec` instances.
"""
specs = self.cldf_specs()
if isinstance(specs, CLDFSpec):
return {None: specs}
assert isinstance(specs, dict)
return specs
def update_submodules(self):
"""
Convenience method to be used in a `Dataset`'s `cmd_download` to update raw data curated
as git submodules.
"""
subprocess.check_call('git -C {} submodule update --remote'.format(
self.dir.resolve()),
shell=True)
def cldf_writer(self, args, cldf_spec=None, clean=True) -> CLDFWriter:
"""
:param args:
:param cldf_spec: Key of the relevant `CLDFSpec` in `Dataset.cldf_specs`
:param clean: `bool` flag signaling whether to clean the CLDF dir before writing. \
Note that `False` must be passed for subsequent calls to `cldf_writer` in case the \
spec re-uses a directory.
:return: a `cldf_spec.writer_cls` instance, for write-access to CLDF data. \
This method should be used in a with-statement, and will then return a `CLDFWriter` with \
an empty working directory.
"""
if not isinstance(cldf_spec, CLDFSpec):
cldf_spec = self.cldf_specs_dict[cldf_spec]
return cldf_spec.get_writer(args=args, dataset=self, clean=clean)
def cldf_reader(self,
cldf_spec: typing.Union[str,
None] = None) -> pycldf.Dataset:
"""
:param cldf_spec:
:return: a `pycldf.Dataset` instance, for read-access to the CLDF data.
"""
if not isinstance(cldf_spec, CLDFSpec):
cldf_spec = self.cldf_specs_dict[cldf_spec]
return cldf_spec.get_dataset()
@lazyproperty
def repo(self) -> typing.Union[Repository, None]:
"""
The git repository cloned to the dataset's directory (or `None`).
"""
try:
return Repository(self.dir)
except ValueError: # pragma: no cover
return
def _cmd_download(self, args):
self.raw_dir.mkdir(exist_ok=True)
self.cmd_download(args)
(self.raw_dir / 'README.md').write_text(
'Raw data downloaded {0}'.format(datetime.utcnow().isoformat()),
encoding='utf8')
def cmd_download(self, args: argparse.Namespace):
"""
Implementations of this methods should populate the dataset's `raw_dir` with the source
data.
"""
args.log.warning('cmd_{0} not implemented for dataset {1}'.format(
'download', self.id))
return NOOP
def _cmd_readme(self, args):
if self.metadata:
badge = build_status_badge(self)
md = self.cmd_readme(args)
if badge:
lines, title_found = [], False
for line in md.split('\n'):
lines.append(line)
if line.startswith('# ') and not title_found:
title_found = True
lines.extend(['', badge])
md = '\n'.join(lines)
section = [
'\n\n## CLDF Datasets\n',
'The following CLDF datasets are available in [{0}]({0}):\n'.
format(self.cldf_dir.resolve().relative_to(self.dir.resolve()))
]
for ds in self.cldf_specs_dict.values():
if ds.metadata_path.exists():
p = ds.metadata_path.resolve().relative_to(
self.dir.resolve())
section.append(
'- CLDF [{0}](https://github.com/cldf/cldf/tree/master/modules/{0}) '
'at [{1}]({1})'.format(ds.module, p))
self.dir.joinpath('README.md').write_text(md + '\n'.join(section),
encoding='utf8')
def cmd_readme(self, args: argparse.Namespace) -> str:
"""
Implementations of this method should create the content for the dataset's README.md
and return it as markdown formatted string.
"""
return self.metadata.markdown() if self.metadata else ''
def _cmd_makecldf(self, args):
specs = list(self.cldf_specs_dict.values())
if len(specs) == 1:
# There's only one CLDF spec! We instantiate the writer now and inject it into `args`:
with self.cldf_writer(args, cldf_spec=specs[0]) as writer:
args.writer = writer
self.cmd_makecldf(args)
else:
self.cmd_makecldf(args)
if self.metadata and self.metadata.known_license:
legalcode = self.metadata.known_license.legalcode
if legalcode:
(self.dir / 'LICENSE').write_text(legalcode, encoding='utf8')
def cmd_makecldf(self, args: argparse.Namespace):
"""
Implementations of this method should write the CLDF data curated by the dataset.
:param args: An `argparse.Namespace` including attributes: \
- `writer`: :class:`CLDFWriter` instance
"""
args.log.warning('cmd_{0} not implemented for dataset {1}'.format(
'makecldf', self.id))
return NOOP