inm-icf-utilities/bin/dataladify_studyvisit
2023-05-25 14:03:58 +02:00

220 lines
7.4 KiB
Python
Executable file

#!/usr/bin/env python3
"""
"""
import logging
import os
from pathlib import Path
import tempfile
# this implementation works with pydicom 2x
from pydicom import dcmread
from pydicom.errors import InvalidDicomError
import datalad.api as dl
from datalad.utils import md5sum
lgr = logging.getLogger('inm-icf-utilities')
# this points to the top of the ICF data store.
# internally it will be amended with the missing components
# for study and visit deposit locations
icfstore_baseurl = 'https://data.inm-icf.de'
# which DICOM tags to extract from DICOM files and store as
# git-annex metadata (e.g., to enable metadata-driven views
# of visit datasets)
dicom_metadata_keys = [
"SeriesDescription",
"SeriesNumber",
"Modality",
"MRAcquisitionType",
"ProtocolName",
"PulseSequenceName",
]
def main(store_dir: str,
study_id: str,
visit_id: str):
store_base_dir = Path(store_dir)
# where to deposit the final datalad dataset
repo_base_path = store_base_dir / study_id / f'{visit_id}_'
deposit_conflicts = [str(p) for p in repo_base_path.parent.glob(
f'{repo_base_path.name}XDLRA*')]
if deposit_conflicts:
# be safe
raise ValueError(
f'existing dataset deposit {deposit_conflicts}, '
'refusing to overwrite')
# locate input tarball
tar_path = store_base_dir / study_id / f'{visit_id}_dicom.tar'
if not tar_path.exists():
raise ValueError(f'no tarball at {tar_path}')
with tempfile.TemporaryDirectory(prefix='dataladify_visit_') as wdir:
runshit(
# workdir
wdir,
# source visit tarball
tar_path.resolve(),
# source visit tarball URL
f'{icfstore_baseurl}/{study_id}/{tar_path.name}',
# path to deposit the repo at
repo_base_path.absolute(),
)
def runshit(wdir, tarpath, tarurl, repobasepath):
ds = dl.create(wdir)
# alias for speed, `.repo` is really expensive
repo = ds.repo
# enable uncurl remote to have the tarball URL be claimed by it
# and future-proof access (via its reconfiguration possibilities
# without having to touch the annex record
repo.call_annex([
'initremote',
'uncurl',
'type=external',
'externaltype=uncurl',
'encryption=none',
# auto-enabling is cheap (makes no connection attempts), and convenient
'autoenable=true',
])
# we need its UUID later
uncurl_uuid = repo.call_annex_records(['info', 'uncurl'])[0]['uuid']
assert uncurl_uuid
# register the URL of the tarball
res = ds.addurls(
[{
'size': tarpath.stat().st_size,
'md5': md5sum(tarpath),
'path': str(Path('icf', tarpath.name)),
'url': tarurl,
}],
'{url}',
'{path}',
key='et:MD5-s{size}--{md5}',
)
# fish out annex key of tarball.
# we could also construct that, but let's not duplicate the setup above
tarkey = [r.get('annexkey') for r in res
if r.get('action') == 'fromkey'
and r.get('path', '').endswith(tarpath.name)]
assert len(tarkey) == 1
tarkey = tarkey[0]
assert tarkey
# assure tar key availability
repo.call_annex(['setpresentkey', tarkey, uncurl_uuid, '1'])
# here we register the datalad-archives special remote, to claim
# the dl+archives URLs registered below.
# this really should be the archivist remote, but it is not yet implemented.
# this will box ourselves into a corner, and force us to replace the
# implementation of datalad-archives with the archivist code, rather than
# have them coexist -- otherwise we would need to fix all datasets
repo.call_annex([
'initremote',
'datalad-archives',
'type=external',
'externaltype=datalad-archives',
'encryption=none',
# auto-enabling is cheap (makes no connection attempts), and convenient
'autoenable=true',
])
dlarchives_uuid = repo.call_annex_records(
['info', 'datalad-archives'])[0]['uuid']
assert dlarchives_uuid
dicoms = []
for r in dl.ls_file_collection(
'tarfile',
tarpath,
hash=['md5'],
result_renderer='disabled',
return_type='generator'):
fp = r.get('fp')
if fp is None:
# not a file
continue
d = dict(path=str(r['item']), md5=r['hash-md5'], size=r['size'])
# rewind file pointer to ready for dcmread()
fp.seek(0)
try:
with dcmread(
fp,
stop_before_pixels=True,
specific_tags=dicom_metadata_keys,
) as dcm:
# extract target DICOM tags for metadata
d.update({
dmk: getattr(dcm, dmk, '')
for dmk in dicom_metadata_keys
})
dicoms.append(d)
except InvalidDicomError:
lgr.info('skipping non-DICOM file: %s', r['item'])
dicom_recs = ds.addurls(
dicoms,
f'dl+archive:{tarkey}#path={{path}}&size={{size}}',
'{path}',
key='et:MD5-s{size}--{md5}',
# field names are limited to alphanumerics (and [_-.]),
# and are case insensitive
meta=[
f'{dmk.lower()}={{{dmk}}}'
for dmk in dicom_metadata_keys
],
)
# assure availability for each DICOM
dicomkeys = [
r['annexkey']
for r in dicom_recs if r.get('action') == 'fromkey'
]
for dicomkey in dicomkeys:
repo.call_annex(['setpresentkey', dicomkey, dlarchives_uuid, '1'])
repo.call_git([
'remote', 'add', 'icfstore',
# this is a little twisted:
# the first line is an f-string, because we need to get the base URL
# pointing to the study directory into the remote URL
f'datalad-annex::?type=external&externaltype=uncurl&url=file://{repobasepath}'
# this second line is NOT an f-string, and braces are quoted!!
# this is because datalad-annex:: will pass this URL to uncurl
# (removing the quoting; it can do placeholders too!), and uncurl
# will then fill in the annex key of the deposit in order to get
# the final upload URL
'{{annex_key}}&encryption=none'
])
# probe the availability metadata. This seems to be necessary at times to
# get git-annex to commit the metadata operations performed above
# to be able to actually push everything
repo.call_annex(['whereis', '--key', dicomkeys[0]])
ds.push(
to='icfstore',
# under no circumstances do we want to push annexed content.
# and there also should be none
data='nothing',
)
if __name__ == '__main__':
import argparse
p = argparse.ArgumentParser(description=__doc__)
p.add_argument(
"-o", "--store-dir", metavar='PATH', default=os.getcwd(),
help="Root directory of the ICF data store. "
"Visit data will be read from it, and the DataLad dataset will be "
"deposited into it."
)
p.add_argument(
'--id', nargs=2, metavar=('STUDY-ID', 'VISIT-ID'), required=True,
help="The study and visit identifiers, used to "
"locate the visit archive in the storage organization. "
)
args = p.parse_args()
main(store_dir=args.store_dir,
study_id=args.id[0],
visit_id=args.id[1],
)