class ProjectProvenance:
def __init__(self, database_helper, full_provenance=False):
"""
Initializes the provenance for the mjclawar_rarshad project
Parameters
----------
database_helper: DatabaseHelper
full_provenance: bool
Returns
-------
"""
assert isinstance(database_helper, DatabaseHelper)
self.database_helper = database_helper
if full_provenance:
self.prov_doc = ProvDocument.deserialize(dir_info.plan_json)
else:
self.prov_doc = ProvDocument()
self.prov_doc.add_namespace(mcras.BDP_NAMESPACE.name, mcras.BDP_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.ALG_NAMESPACE.name, mcras.ALG_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.DAT_NAMESPACE.name, mcras.DAT_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.LOG_NAMESPACE.name, mcras.LOG_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.ONT_NAMESPACE.name, mcras.ONT_NAMESPACE.link)
def write_provenance_json(self):
self.prov_doc.serialize(dir_info.plan_json)
class ProjectProvenance:
def __init__(self, database_helper, full_provenance=False):
"""
Initializes the provenance for the mjclawar_rarshad project
Parameters
----------
database_helper: DatabaseHelper
full_provenance: bool
Returns
-------
"""
assert isinstance(database_helper, DatabaseHelper)
self.database_helper = database_helper
if full_provenance:
self.prov_doc = ProvDocument.deserialize(dir_info.plan_json)
else:
self.prov_doc = ProvDocument()
self.prov_doc.add_namespace(mcras.BDP_NAMESPACE.name,
mcras.BDP_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.ALG_NAMESPACE.name,
mcras.ALG_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.DAT_NAMESPACE.name,
mcras.DAT_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.LOG_NAMESPACE.name,
mcras.LOG_NAMESPACE.link)
self.prov_doc.add_namespace(mcras.ONT_NAMESPACE.name,
mcras.ONT_NAMESPACE.link)
def write_provenance_json(self):
self.prov_doc.serialize(dir_info.plan_json)
def save_provenance(prov_doc: ProvDocument, filepath: Path):
logging.debug("Saving provenance files:")
logging.debug(" - %s", filepath)
with filepath.open("w") as f:
prov_doc.serialize(f)
provn_content = prov_doc.get_provn()
filepath = filepath.with_suffix(".provn")
logging.debug(" - %s", filepath)
with filepath.open("w") as f:
f.write(provn_content)
def create_document():
# Create a new provenance document
document = ProvDocument() # d1 is now an empty provenance document
# Before asserting provenance statements, we need to have a way to refer to the "things"
# we want to describe provenance (e.g. articles, data sets, people). For that purpose,
# PROV uses qualified names to identify things, which essentially a shortened representation
# of a URI in the form of prefix:localpart. Valid qualified names require their prefixes defined,
# which we is going to do next.
# Declaring namespaces for various prefixes used in the example
document.add_namespace('now', 'http://www.provbook.org/nownews/')
document.add_namespace('nowpeople',
'http://www.provbook.org/nownews/people/')
document.add_namespace('bk', 'http://www.provbook.org/ns/#')
# Entity: now:employment-article-v1.html
e1 = document.entity('now:employment-article-v1.html')
e1.add_attributes({'prov:value': 'Conteudo do HTML'})
document.agent('nowpeople:Filipe')
# Attributing the article to the agent
document.wasAttributedTo(
e1, 'nowpeople:Filipe_' + str(random.randint(1000, 1070000000)))
# add more namespace declarations
document.add_namespace('govftp',
'ftp://ftp.bls.gov/pub/special.requests/oes/')
document.add_namespace('void', 'http://vocab.deri.ie/void#')
# 'now:employment-article-v1.html' was derived from at dataset at govftp
document.entity('govftp:oesm11st.zip', {
'prov:label': 'employment-stats-2011',
'prov:type': 'void:Dataset'
})
document.wasDerivedFrom('now:employment-article-v1.html',
'govftp:oesm11st.zip')
# Adding an activity
document.add_namespace('is', 'http://www.provbook.org/nownews/is/#')
document.activity('is:writeArticle')
# Usage and Generation
document.used('is:writeArticle', 'govftp:oesm11st.zip')
document.wasGeneratedBy('now:employment-article-v1.html',
'is:writeArticle')
#print("Document prepared.")
# What we have so far (in PROV-N)
logging.debug(document.serialize(indent=2))
# d1.serialize('article-prov.json') # write to file
return document
class TrackedFile(object):
"""File with provenance tracking."""
def __init__(self, filename, attributes, ancestors=None):
"""Create an instance of a file with provenance tracking."""
self._filename = filename
self.attributes = copy.deepcopy(attributes)
self.provenance = None
self.entity = None
self.activity = None
self._ancestors = [] if ancestors is None else ancestors
def __str__(self):
"""Return summary string."""
return "{}: {}".format(self.__class__.__name__, self.filename)
def copy_provenance(self, target=None):
"""Create a copy with identical provenance information."""
if self.provenance is None:
raise ValueError("Provenance of {} not initialized".format(self))
if target is None:
new = TrackedFile(self.filename, self.attributes)
else:
if target.filename != self.filename:
raise ValueError(
"Attempt to copy provenance to incompatible file.")
new = target
new.attributes = copy.deepcopy(self.attributes)
new.provenance = copy.deepcopy(self.provenance)
new.entity = new.provenance.get_record(self.entity.identifier)[0]
new.activity = new.provenance.get_record(self.activity.identifier)[0]
return new
@property
def filename(self):
"""Filename."""
return self._filename
def initialize_provenance(self, activity):
"""Initialize the provenance document.
Note: this also copies the ancestor provenance. Therefore, changes
made to ancestor provenance after calling this function will not
propagate into the provenance of this file.
"""
if self.provenance is not None:
raise ValueError(
"Provenance of {} already initialized".format(self))
self.provenance = ProvDocument()
self._initialize_namespaces()
self._initialize_activity(activity)
self._initialize_entity()
self._initialize_ancestors(activity)
def _initialize_namespaces(self):
"""Inialize the namespaces."""
for namespace in ('file', 'attribute', 'preprocessor', 'task'):
create_namespace(self.provenance, namespace)
def _initialize_activity(self, activity):
"""Copy the preprocessor task activity."""
self.activity = activity
update_without_duplicating(self.provenance, activity.bundle)
def _initialize_entity(self):
"""Initialize the entity representing the file."""
attributes = {
'attribute:' + k: str(v)
for k, v in self.attributes.items()
if k not in ('authors', 'projects')
}
self.entity = self.provenance.entity('file:' + self.filename,
attributes)
attribute_to_authors(self.entity, self.attributes.get('authors', []))
attribute_to_projects(self.entity, self.attributes.get('projects', []))
def _initialize_ancestors(self, activity):
"""Register ancestor files for provenance tracking."""
for ancestor in self._ancestors:
if ancestor.provenance is None:
ancestor.initialize_provenance(activity)
update_without_duplicating(self.provenance, ancestor.provenance)
self.wasderivedfrom(ancestor)
def wasderivedfrom(self, other):
"""Let the file know that it was derived from other."""
if isinstance(other, TrackedFile):
other_entity = other.entity
else:
other_entity = other
update_without_duplicating(self.provenance, other_entity.bundle)
if not self.activity:
raise ValueError("Activity not initialized.")
self.entity.wasDerivedFrom(other_entity, self.activity)
def _select_for_include(self):
attributes = {
'provenance': self.provenance.serialize(format='xml'),
'software': "Created with ESMValTool v{}".format(__version__),
}
if 'caption' in self.attributes:
attributes['caption'] = self.attributes['caption']
return attributes
@staticmethod
def _include_provenance_nc(filename, attributes):
with Dataset(filename, 'a') as dataset:
for key, value in attributes.items():
setattr(dataset, key, value)
@staticmethod
def _include_provenance_png(filename, attributes):
pnginfo = PngInfo()
exif_tags = {
'provenance': 'ImageHistory',
'caption': 'ImageDescription',
'software': 'Software',
}
for key, value in attributes.items():
pnginfo.add_text(exif_tags.get(key, key), value, zip=True)
with Image.open(filename) as image:
image.save(filename, pnginfo=pnginfo)
def _include_provenance(self):
"""Include provenance information as metadata."""
attributes = self._select_for_include()
# List of files to attach provenance to
files = [self.filename]
if 'plot_file' in self.attributes:
files.append(self.attributes['plot_file'])
# Attach provenance to supported file types
for filename in files:
ext = os.path.splitext(filename)[1].lstrip('.').lower()
write = getattr(self, '_include_provenance_' + ext, None)
if write:
write(filename, attributes)
def save_provenance(self):
"""Export provenance information."""
self._include_provenance()
filename = os.path.splitext(self.filename)[0] + '_provenance'
self.provenance.serialize(filename + '.xml', format='xml')
# Only plot provenance if there are not too many records.
if len(self.provenance.records) > 100:
logger.debug("Not plotting large provenance tree of %s",
self.filename)
else:
figure = prov_to_dot(self.provenance)
figure.write_svg(filename + '.svg')
class Provenance(object):
def __init__(self, output_dir):
self.output_dir = output_dir
self.doc = None
self.workflow = None
def start(self, workflow=False):
from daops import __version__ as daops_version
from housemartin import __version__ as housemartin_version
self.doc = ProvDocument()
# Declaring namespaces for various prefixes
self.doc.set_default_namespace(uri="http://purl.org/roocs/prov#")
self.doc.add_namespace("prov", uri="http://www.w3.org/ns/prov#")
self.doc.add_namespace(
"provone", uri="http://purl.dataone.org/provone/2015/01/15/ontology#"
)
self.doc.add_namespace("dcterms", uri="http://purl.org/dc/terms/")
# Define entities
project_cds = self.doc.agent(
":copernicus_CDS",
{
"prov:type": "prov:Organization",
"dcterms:title": "Copernicus Climate Data Store",
},
)
self.sw_housemartin = self.doc.agent(
":housemartin",
{
"prov:type": "prov:SoftwareAgent",
"dcterms:source": f"https://github.com/cedadev/housemartin/releases/tag/v{housemartin_version}",
},
)
self.doc.wasAttributedTo(self.sw_housemartin, project_cds)
self.sw_daops = self.doc.agent(
":daops",
{
"prov:type": "prov:SoftwareAgent",
"dcterms:source": f"https://github.com/roocs/daops/releases/tag/v{daops_version}",
},
)
# workflow
if workflow is True:
self.workflow = self.doc.entity(
":workflow", {"prov:type": "provone:Workflow"}
)
orchestrate = self.doc.activity(
":orchestrate",
other_attributes={
"prov:startedAtTime": "2020-11-26T09:15:00",
"prov:endedAtTime": "2020-11-26T09:30:00",
},
)
self.doc.wasAssociatedWith(
orchestrate, agent=self.sw_housemartin, plan=self.workflow
)
def add_operator(self, operator, parameters, collection, output):
op = self.doc.activity(
f":{operator}",
other_attributes={
":time": parameters.get("time"),
":apply_fixes": parameters.get("apply_fixes"),
},
)
# input data
ds_in = os.path.basename(collection[0])
# ds_in_attrs = {
# 'prov:type': 'provone:Data',
# 'prov:value': f'{ds_in}',
# }
op_in = self.doc.entity(f":{ds_in}")
# operator started by daops
if self.workflow:
self.doc.wasAssociatedWith(op, agent=self.sw_daops, plan=self.workflow)
else:
self.doc.start(op, starter=self.sw_daops, trigger=self.sw_housemartin)
# Generated output file
ds_out = os.path.basename(output[0])
# ds_out_attrs = {
# 'prov:type': 'provone:Data',
# 'prov:value': f'{ds_out}',
# }
op_out = self.doc.entity(f":{ds_out}")
self.doc.wasDerivedFrom(op_out, op_in, activity=op)
def write_json(self):
outfile = os.path.join(self.output_dir, "provenance.json")
self.doc.serialize(outfile, format="json")
return outfile
def write_png(self):
outfile = os.path.join(self.output_dir, "provenance.png")
figure = prov_to_dot(self.doc)
figure.write_png(outfile)
return outfile
class NIDMExporter():
"""
Generic class to parse a result directory to extract the pieces of
information to be stored in NIDM-Results and to generate a NIDM-Results
export.
"""
def __init__(self, version, out_dir, zipped=True):
out_dirname = os.path.basename(out_dir)
out_path = os.path.dirname(out_dir)
# Create output path from output name
self.zipped = zipped
if not self.zipped:
out_dirname = out_dirname + ".nidm"
else:
out_dirname = out_dirname + ".nidm.zip"
out_dir = os.path.join(out_path, out_dirname)
# Quit if output path already exists and user doesn't want to overwrite
# it
if os.path.exists(out_dir):
msg = out_dir + " already exists, overwrite?"
if not input("%s (y/N) " % msg).lower() == 'y':
quit("Bye.")
if os.path.isdir(out_dir):
shutil.rmtree(out_dir)
else:
os.remove(out_dir)
self.out_dir = out_dir
if version == "dev":
self.version = {
'major': 10000,
'minor': 0,
'revision': 0,
'num': version
}
else:
major, minor, revision = version.split(".")
if "-rc" in revision:
revision, rc = revision.split("-rc")
else:
rc = -1
self.version = {
'major': int(major),
'minor': int(minor),
'revision': int(revision),
'rc': int(rc),
'num': version
}
# Initialise prov document
self.doc = ProvDocument()
self._add_namespaces()
# A temp directory that will contain the exported data
self.export_dir = tempfile.mkdtemp(prefix="nidm-", dir=out_path)
self.prepend_path = ''
def parse(self):
"""
Parse a result directory to extract the pieces information to be
stored in NIDM-Results.
"""
try:
# Methods: find_software, find_model_fitting, find_contrasts and
# find_inferences should be defined in the children classes and
# return a list of NIDM Objects as specified in the objects module
# Object of type Software describing the neuroimaging software
# package used for the analysis
self.software = self._find_software()
# List of objects of type ModelFitting describing the
# model fitting step in NIDM-Results (main activity: Model
# Parameters Estimation)
self.model_fittings = self._find_model_fitting()
# Dictionary of (key, value) pairs where where key is a tuple
# containing the identifier of a ModelParametersEstimation object
# and a tuple of identifiers of ParameterEstimateMap objects and
# value is an object of type Contrast describing the contrast
# estimation step in NIDM-Results (main activity: Contrast
# Estimation)
self.contrasts = self._find_contrasts()
# Inference activity and entities
# Dictionary of (key, value) pairs where key is the identifier of a
# ContrastEstimation object and value is an object of type
# Inference describing the inference step in NIDM-Results (main
# activity: Inference)
self.inferences = self._find_inferences()
except Exception:
self.cleanup()
raise
def cleanup(self):
if os.path.isdir(self.export_dir):
shutil.rmtree(self.export_dir)
def add_object(self, nidm_object, export_file=True):
"""
Add a NIDMObject to a NIDM-Results export.
"""
if not export_file:
export_dir = None
else:
export_dir = self.export_dir
if not isinstance(nidm_object, NIDMFile):
nidm_object.export(self.version, export_dir)
else:
nidm_object.export(self.version, export_dir, self.prepend_path)
# ProvDocument: add object to the bundle
if nidm_object.prov_type == PROV['Activity']:
self.bundle.activity(nidm_object.id,
other_attributes=nidm_object.attributes)
elif nidm_object.prov_type == PROV['Entity']:
self.bundle.entity(nidm_object.id,
other_attributes=nidm_object.attributes)
elif nidm_object.prov_type == PROV['Agent']:
self.bundle.agent(nidm_object.id,
other_attributes=nidm_object.attributes)
# self.bundle.update(nidm_object.p)
def export(self):
"""
Generate a NIDM-Results export.
"""
try:
if not os.path.isdir(self.export_dir):
os.mkdir(self.export_dir)
# Initialise main bundle
self._create_bundle(self.version)
self.add_object(self.software)
# Add model fitting steps
if not isinstance(self.model_fittings, list):
self.model_fittings = list(self.model_fittings.values())
for model_fitting in self.model_fittings:
# Design Matrix
# model_fitting.activity.used(model_fitting.design_matrix)
self.bundle.used(model_fitting.activity.id,
model_fitting.design_matrix.id)
self.add_object(model_fitting.design_matrix)
# *** Export visualisation of the design matrix
self.add_object(model_fitting.design_matrix.image)
if model_fitting.design_matrix.image.file is not None:
self.add_object(model_fitting.design_matrix.image.file)
if model_fitting.design_matrix.hrf_models is not None:
# drift model
self.add_object(model_fitting.design_matrix.drift_model)
if self.version['major'] > 1 or \
(self.version['major'] == 1 and
self.version['minor'] >= 3):
# Machine
# model_fitting.data.wasAttributedTo(model_fitting.machine)
self.bundle.wasAttributedTo(model_fitting.data.id,
model_fitting.machine.id)
self.add_object(model_fitting.machine)
# Imaged subject or group(s)
for sub in model_fitting.subjects:
self.add_object(sub)
# model_fitting.data.wasAttributedTo(sub)
self.bundle.wasAttributedTo(model_fitting.data.id,
sub.id)
# Data
# model_fitting.activity.used(model_fitting.data)
self.bundle.used(model_fitting.activity.id,
model_fitting.data.id)
self.add_object(model_fitting.data)
# Error Model
# model_fitting.activity.used(model_fitting.error_model)
self.bundle.used(model_fitting.activity.id,
model_fitting.error_model.id)
self.add_object(model_fitting.error_model)
# Parameter Estimate Maps
for param_estimate in model_fitting.param_estimates:
# param_estimate.wasGeneratedBy(model_fitting.activity)
self.bundle.wasGeneratedBy(param_estimate.id,
model_fitting.activity.id)
self.add_object(param_estimate)
self.add_object(param_estimate.coord_space)
self.add_object(param_estimate.file)
if param_estimate.derfrom is not None:
self.bundle.wasDerivedFrom(param_estimate.id,
param_estimate.derfrom.id)
self.add_object(param_estimate.derfrom)
self.add_object(param_estimate.derfrom.file,
export_file=False)
# Residual Mean Squares Map
# model_fitting.rms_map.wasGeneratedBy(model_fitting.activity)
self.add_object(model_fitting.rms_map)
self.bundle.wasGeneratedBy(model_fitting.rms_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.rms_map.coord_space)
self.add_object(model_fitting.rms_map.file)
if model_fitting.rms_map.derfrom is not None:
self.bundle.wasDerivedFrom(
model_fitting.rms_map.id,
model_fitting.rms_map.derfrom.id)
self.add_object(model_fitting.rms_map.derfrom)
self.add_object(model_fitting.rms_map.derfrom.file,
export_file=False)
# Resels per Voxel Map
if model_fitting.rpv_map is not None:
self.add_object(model_fitting.rpv_map)
self.bundle.wasGeneratedBy(model_fitting.rpv_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.rpv_map.coord_space)
self.add_object(model_fitting.rpv_map.file)
if model_fitting.rpv_map.inf_id is not None:
self.bundle.used(model_fitting.rpv_map.inf_id,
model_fitting.rpv_map.id)
if model_fitting.rpv_map.derfrom is not None:
self.bundle.wasDerivedFrom(
model_fitting.rpv_map.id,
model_fitting.rpv_map.derfrom.id)
self.add_object(model_fitting.rpv_map.derfrom)
self.add_object(model_fitting.rpv_map.derfrom.file,
export_file=False)
# Mask
# model_fitting.mask_map.wasGeneratedBy(model_fitting.activity)
self.bundle.wasGeneratedBy(model_fitting.mask_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.mask_map)
if model_fitting.mask_map.derfrom is not None:
self.bundle.wasDerivedFrom(
model_fitting.mask_map.id,
model_fitting.mask_map.derfrom.id)
self.add_object(model_fitting.mask_map.derfrom)
self.add_object(model_fitting.mask_map.derfrom.file,
export_file=False)
# Create coordinate space export
self.add_object(model_fitting.mask_map.coord_space)
# Create "Mask map" entity
self.add_object(model_fitting.mask_map.file)
# Grand Mean map
# model_fitting.grand_mean_map.wasGeneratedBy(model_fitting.activity)
self.bundle.wasGeneratedBy(model_fitting.grand_mean_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.grand_mean_map)
# Coordinate space entity
self.add_object(model_fitting.grand_mean_map.coord_space)
# Grand Mean Map entity
self.add_object(model_fitting.grand_mean_map.file)
# Model Parameters Estimation activity
self.add_object(model_fitting.activity)
self.bundle.wasAssociatedWith(model_fitting.activity.id,
self.software.id)
# model_fitting.activity.wasAssociatedWith(self.software)
# self.add_object(model_fitting)
# Add contrast estimation steps
analysis_masks = dict()
for (model_fitting_id,
pe_ids), contrasts in list(self.contrasts.items()):
for contrast in contrasts:
model_fitting = self._get_model_fitting(model_fitting_id)
# for contrast in contrasts:
# contrast.estimation.used(model_fitting.rms_map)
self.bundle.used(contrast.estimation.id,
model_fitting.rms_map.id)
# contrast.estimation.used(model_fitting.mask_map)
self.bundle.used(contrast.estimation.id,
model_fitting.mask_map.id)
analysis_masks[contrast.estimation.id] = \
model_fitting.mask_map.id
self.bundle.used(contrast.estimation.id,
contrast.weights.id)
self.bundle.used(contrast.estimation.id,
model_fitting.design_matrix.id)
# contrast.estimation.wasAssociatedWith(self.software)
self.bundle.wasAssociatedWith(contrast.estimation.id,
self.software.id)
for pe_id in pe_ids:
# contrast.estimation.used(pe_id)
self.bundle.used(contrast.estimation.id, pe_id)
# Create estimation activity
self.add_object(contrast.estimation)
# Create contrast weights
self.add_object(contrast.weights)
if contrast.contrast_map is not None:
# Create contrast Map
# contrast.contrast_map.wasGeneratedBy(contrast.estimation)
self.bundle.wasGeneratedBy(contrast.contrast_map.id,
contrast.estimation.id)
self.add_object(contrast.contrast_map)
self.add_object(contrast.contrast_map.coord_space)
# Copy contrast map in export directory
self.add_object(contrast.contrast_map.file)
if contrast.contrast_map.derfrom is not None:
self.bundle.wasDerivedFrom(
contrast.contrast_map.id,
contrast.contrast_map.derfrom.id)
self.add_object(contrast.contrast_map.derfrom)
self.add_object(contrast.contrast_map.derfrom.file,
export_file=False)
# Create Std Err. Map (T-tests) or Explained Mean Sq. Map
# (F-tests)
# contrast.stderr_or_expl_mean_sq_map.wasGeneratedBy
# (contrast.estimation)
stderr_explmeansq_map = (
contrast.stderr_or_expl_mean_sq_map)
self.bundle.wasGeneratedBy(stderr_explmeansq_map.id,
contrast.estimation.id)
self.add_object(stderr_explmeansq_map)
self.add_object(stderr_explmeansq_map.coord_space)
if isinstance(stderr_explmeansq_map,
ContrastStdErrMap) and \
stderr_explmeansq_map.contrast_var:
self.add_object(stderr_explmeansq_map.contrast_var)
if stderr_explmeansq_map.var_coord_space:
self.add_object(
stderr_explmeansq_map.var_coord_space)
if stderr_explmeansq_map.contrast_var.coord_space:
self.add_object(
stderr_explmeansq_map.contrast_var.coord_space)
self.add_object(
stderr_explmeansq_map.contrast_var.file,
export_file=False)
self.bundle.wasDerivedFrom(
stderr_explmeansq_map.id,
stderr_explmeansq_map.contrast_var.id)
self.add_object(stderr_explmeansq_map.file)
# Create Statistic Map
# contrast.stat_map.wasGeneratedBy(contrast.estimation)
self.bundle.wasGeneratedBy(contrast.stat_map.id,
contrast.estimation.id)
self.add_object(contrast.stat_map)
self.add_object(contrast.stat_map.coord_space)
# Copy Statistical map in export directory
self.add_object(contrast.stat_map.file)
if contrast.stat_map.derfrom is not None:
self.bundle.wasDerivedFrom(
contrast.stat_map.id, contrast.stat_map.derfrom.id)
self.add_object(contrast.stat_map.derfrom)
self.add_object(contrast.stat_map.derfrom.file,
export_file=False)
# Create Z Statistic Map
if contrast.z_stat_map:
# contrast.z_stat_map.wasGeneratedBy(contrast.estimation)
self.bundle.wasGeneratedBy(contrast.z_stat_map.id,
contrast.estimation.id)
self.add_object(contrast.z_stat_map)
self.add_object(contrast.z_stat_map.coord_space)
# Copy Statistical map in export directory
self.add_object(contrast.z_stat_map.file)
# self.add_object(contrast)
# Add inference steps
for contrast_id, inferences in list(self.inferences.items()):
contrast = self._get_contrast(contrast_id)
for inference in inferences:
if contrast.z_stat_map:
used_id = contrast.z_stat_map.id
else:
used_id = contrast.stat_map.id
# inference.inference_act.used(used_id)
self.bundle.used(inference.inference_act.id, used_id)
# inference.inference_act.wasAssociatedWith(self.software)
self.bundle.wasAssociatedWith(inference.inference_act.id,
self.software.id)
# self.add_object(inference)
# Excursion set
# inference.excursion_set.wasGeneratedBy(inference.inference_act)
self.bundle.wasGeneratedBy(inference.excursion_set.id,
inference.inference_act.id)
self.add_object(inference.excursion_set)
self.add_object(inference.excursion_set.coord_space)
if inference.excursion_set.visu is not None:
self.add_object(inference.excursion_set.visu)
if inference.excursion_set.visu.file is not None:
self.add_object(inference.excursion_set.visu.file)
# Copy "Excursion set map" file in export directory
self.add_object(inference.excursion_set.file)
if inference.excursion_set.clust_map is not None:
self.add_object(inference.excursion_set.clust_map)
self.add_object(inference.excursion_set.clust_map.file)
self.add_object(
inference.excursion_set.clust_map.coord_space)
if inference.excursion_set.mip is not None:
self.add_object(inference.excursion_set.mip)
self.add_object(inference.excursion_set.mip.file)
# Height threshold
if inference.height_thresh.equiv_thresh is not None:
for equiv in inference.height_thresh.equiv_thresh:
self.add_object(equiv)
self.add_object(inference.height_thresh)
# Extent threshold
if inference.extent_thresh.equiv_thresh is not None:
for equiv in inference.extent_thresh.equiv_thresh:
self.add_object(equiv)
self.add_object(inference.extent_thresh)
# Display Mask (potentially more than 1)
if inference.disp_mask:
for mask in inference.disp_mask:
# inference.inference_act.used(mask)
self.bundle.used(inference.inference_act.id,
mask.id)
self.add_object(mask)
# Create coordinate space entity
self.add_object(mask.coord_space)
# Create "Display Mask Map" entity
self.add_object(mask.file)
if mask.derfrom is not None:
self.bundle.wasDerivedFrom(
mask.id, mask.derfrom.id)
self.add_object(mask.derfrom)
self.add_object(mask.derfrom.file,
export_file=False)
# Search Space
self.bundle.wasGeneratedBy(inference.search_space.id,
inference.inference_act.id)
# inference.search_space.wasGeneratedBy(inference.inference_act)
self.add_object(inference.search_space)
self.add_object(inference.search_space.coord_space)
# Copy "Mask map" in export directory
self.add_object(inference.search_space.file)
# Peak Definition
if inference.peak_criteria:
# inference.inference_act.used(inference.peak_criteria)
self.bundle.used(inference.inference_act.id,
inference.peak_criteria.id)
self.add_object(inference.peak_criteria)
# Cluster Definition
if inference.cluster_criteria:
# inference.inference_act.used(inference.cluster_criteria)
self.bundle.used(inference.inference_act.id,
inference.cluster_criteria.id)
self.add_object(inference.cluster_criteria)
if inference.clusters:
# Clusters and peaks
for cluster in inference.clusters:
# cluster.wasDerivedFrom(inference.excursion_set)
self.bundle.wasDerivedFrom(
cluster.id, inference.excursion_set.id)
self.add_object(cluster)
for peak in cluster.peaks:
self.bundle.wasDerivedFrom(peak.id, cluster.id)
self.add_object(peak)
self.add_object(peak.coordinate)
if cluster.cog is not None:
self.bundle.wasDerivedFrom(
cluster.cog.id, cluster.id)
self.add_object(cluster.cog)
self.add_object(cluster.cog.coordinate)
# Inference activity
# inference.inference_act.wasAssociatedWith(inference.software_id)
# inference.inference_act.used(inference.height_thresh)
self.bundle.used(inference.inference_act.id,
inference.height_thresh.id)
# inference.inference_act.used(inference.extent_thresh)
self.bundle.used(inference.inference_act.id,
inference.extent_thresh.id)
self.bundle.used(inference.inference_act.id,
analysis_masks[contrast.estimation.id])
self.add_object(inference.inference_act)
# Write-out prov file
self.save_prov_to_files()
return self.out_dir
except Exception:
self.cleanup()
raise
def _get_model_fitting(self, mf_id):
"""
Retreive model fitting with identifier 'mf_id' from the list of model
fitting objects stored in self.model_fitting
"""
for model_fitting in self.model_fittings:
if model_fitting.activity.id == mf_id:
return model_fitting
raise Exception("Model fitting activity with id: " + str(mf_id) +
" not found.")
def _get_contrast(self, con_id):
"""
Retreive contrast with identifier 'con_id' from the list of contrast
objects stored in self.contrasts
"""
for contrasts in list(self.contrasts.values()):
for contrast in contrasts:
if contrast.estimation.id == con_id:
return contrast
raise Exception("Contrast activity with id: " + str(con_id) +
" not found.")
def _add_namespaces(self):
"""
Add namespaces to NIDM document.
"""
self.doc.add_namespace(NIDM)
self.doc.add_namespace(NIIRI)
self.doc.add_namespace(CRYPTO)
self.doc.add_namespace(DCT)
self.doc.add_namespace(DC)
self.doc.add_namespace(NFO)
self.doc.add_namespace(OBO)
self.doc.add_namespace(SCR)
self.doc.add_namespace(NIF)
def _create_bundle(self, version):
"""
Initialise NIDM-Results bundle.
"""
# *** Bundle entity
if not hasattr(self, 'bundle_ent'):
self.bundle_ent = NIDMResultsBundle(nidm_version=version['num'])
self.bundle = ProvBundle(identifier=self.bundle_ent.id)
self.bundle_ent.export(self.version, self.export_dir)
# # provn export
# self.bundle = ProvBundle(identifier=bundle_id)
self.doc.entity(self.bundle_ent.id,
other_attributes=self.bundle_ent.attributes)
# *** NIDM-Results Export Activity
if version['num'] not in ["1.0.0", "1.1.0"]:
if not hasattr(self, 'export_act'):
self.export_act = NIDMResultsExport()
self.export_act.export(self.version, self.export_dir)
# self.doc.update(self.export_act.p)
self.doc.activity(self.export_act.id,
other_attributes=self.export_act.attributes)
# *** bundle was Generated by NIDM-Results Export Activity
if not hasattr(self, 'export_time'):
self.export_time = str(datetime.datetime.now().time())
if version['num'] in ["1.0.0", "1.1.0"]:
self.doc.wasGeneratedBy(entity=self.bundle_ent.id,
time=self.export_time)
else:
# provn
self.doc.wasGeneratedBy(entity=self.bundle_ent.id,
activity=self.export_act.id,
time=self.export_time)
# *** NIDM-Results Exporter (Software Agent)
if version['num'] not in ["1.0.0", "1.1.0"]:
if not hasattr(self, 'exporter'):
self.exporter = self._get_exporter()
self.exporter.export(self.version, self.export_dir)
# self.doc.update(self.exporter.p)
self.doc.agent(self.exporter.id,
other_attributes=self.exporter.attributes)
self.doc.wasAssociatedWith(self.export_act.id, self.exporter.id)
def _get_model_parameters_estimations(self, error_model):
"""
Infer model estimation method from the 'error_model'. Return an object
of type ModelParametersEstimation.
"""
if error_model.dependance == NIDM_INDEPEDENT_ERROR:
if error_model.variance_homo:
estimation_method = STATO_OLS
else:
estimation_method = STATO_WLS
else:
estimation_method = STATO_GLS
mpe = ModelParametersEstimation(estimation_method, self.software.id)
return mpe
def use_prefixes(self, ttl):
prefix_file = os.path.join(os.path.dirname(__file__), 'prefixes.csv')
context = dict()
with open(prefix_file, encoding="ascii") as csvfile:
reader = csv.reader(csvfile)
next(reader, None) # skip the headers
for alphanum_id, prefix, uri in reader:
if alphanum_id in ttl:
context[prefix] = uri
ttl = "@prefix " + prefix + ": <" + uri + "> .\n" + ttl
ttl = ttl.replace(alphanum_id, prefix + ":")
if uri in ttl:
ttl = ttl.replace(alphanum_id, prefix + ":")
elif uri in ttl:
context[prefix] = uri
ttl = "@prefix " + prefix + ": <" + uri + "> .\n" + ttl
ttl = ttl.replace(alphanum_id, prefix + ":")
return (ttl, context)
def save_prov_to_files(self, showattributes=False):
"""
Write-out provn serialisation to nidm.provn.
"""
self.doc.add_bundle(self.bundle)
# provn_file = os.path.join(self.export_dir, 'nidm.provn')
# provn_fid = open(provn_file, 'w')
# # FIXME None
# # provn_fid.write(self.doc.get_provn(4).replace("None", "-"))
# provn_fid.close()
ttl_file = os.path.join(self.export_dir, 'nidm.ttl')
ttl_txt = self.doc.serialize(format='rdf', rdf_format='turtle')
ttl_txt, json_context = self.use_prefixes(ttl_txt)
# Add namespaces to json-ld context
for namespace in self.doc._namespaces.get_registered_namespaces():
json_context[namespace._prefix] = namespace._uri
for namespace in \
list(self.doc._namespaces._default_namespaces.values()):
json_context[namespace._prefix] = namespace._uri
json_context["xsd"] = "http://www.w3.org/2000/01/rdf-schema#"
# Work-around to issue with INF value in rdflib (reported in
# https://github.com/RDFLib/rdflib/pull/655)
ttl_txt = ttl_txt.replace(' inf ', ' "INF"^^xsd:float ')
with open(ttl_file, 'w') as ttl_fid:
ttl_fid.write(ttl_txt)
# print(json_context)
jsonld_file = os.path.join(self.export_dir, 'nidm.json')
jsonld_txt = self.doc.serialize(format='rdf',
rdf_format='json-ld',
context=json_context)
with open(jsonld_file, 'w') as jsonld_fid:
jsonld_fid.write(jsonld_txt)
# provjsonld_file = os.path.join(self.export_dir, 'nidm.provjsonld')
# provjsonld_txt = self.doc.serialize(format='jsonld')
# with open(provjsonld_file, 'w') as provjsonld_fid:
# provjsonld_fid.write(provjsonld_txt)
# provn_file = os.path.join(self.export_dir, 'nidm.provn')
# provn_txt = self.doc.serialize(format='provn')
# with open(provn_file, 'w') as provn_fid:
# provn_fid.write(provn_txt)
# Post-processing
if not self.zipped:
# Just rename temp directory to output_path
os.rename(self.export_dir, self.out_dir)
else:
# Create a zip file that contains the content of the temp directory
os.chdir(self.export_dir)
zf = zipfile.ZipFile(os.path.join("..", self.out_dir), mode='w')
try:
for root, dirnames, filenames in os.walk("."):
for filename in filenames:
zf.write(os.path.join(filename))
shutil.rmtree(os.path.join("..", self.export_dir))
finally:
zf.close()
os.chdir("..")
def get_provenance_history(uuid, normalized_provenance_dict):
prov_doc = ProvDocument()
# The 'prov' prefix is build-in namespace, no need to redefine here
prov_doc.add_namespace(HUBMAP_NAMESPACE, 'https://hubmapconsortium.org/')
# A bit validation
if 'relationships' not in normalized_provenance_dict:
raise LookupError(
f'Missing "relationships" key from the normalized_provenance_dict for Entity of uuid: {uuid}'
)
if 'nodes' not in normalized_provenance_dict:
raise LookupError(
f'Missing "nodes" key from the normalized_provenance_dict for Entity of uuid: {uuid}'
)
# Pack the nodes into a dictionary using the uuid as key
nodes_dict = {}
for node in normalized_provenance_dict['nodes']:
nodes_dict[node['uuid']] = node
# Loop through the relationships and build the provenance document
for rel_dict in normalized_provenance_dict['relationships']:
# (Activity) - [ACTIVITY_OUTPUT] -> (Entity)
if rel_dict['rel_data']['type'] == 'ACTIVITY_OUTPUT':
activity_uuid = rel_dict['fromNode']['uuid']
entity_uuid = rel_dict['toNode']['uuid']
# (Entity) - [ACTIVITY_INPUT] -> (Activity)
elif rel_dict['rel_data']['type'] == 'ACTIVITY_INPUT':
entity_uuid = rel_dict['fromNode']['uuid']
activity_uuid = rel_dict['toNode']['uuid']
activity_node = nodes_dict[activity_uuid]
entity_node = nodes_dict[entity_uuid]
activity_uri = None
entity_uri = None
# Skip Lab nodes for agent and organization
if entity_node['entity_type'] != 'Lab':
# Get the agent information from the entity node
agent_record = get_agent_record(entity_node)
# Use 'created_by_user_sub' as agent ID if presents
# Otherwise, fall back to use email by replacing @ and .
created_by_user_sub_prov_key = f'{HUBMAP_NAMESPACE}:userUUID'
created_by_user_email_prov_key = f'{HUBMAP_NAMESPACE}:userEmail'
if created_by_user_sub_prov_key in agent_record:
agent_id = agent_record[created_by_user_sub_prov_key]
elif created_by_user_email_prov_key in agent_record:
agent_id = str(
agent_record[created_by_user_email_prov_key]).replace(
'@', '-')
agent_id = str(agent_id).replace('.', '-')
else:
msg = f"Both 'created_by_user_sub' and 'created_by_user_email' are missing form entity of uuid: {entity_node['uuid']}"
logger.error(msg)
raise LookupError(msg)
# Build the agent uri
agent_uri = build_uri(HUBMAP_NAMESPACE, 'agent', agent_id)
# Only add the same agent once
# Multiple entities can be associated to the same agent
agent = prov_doc.get_record(agent_uri)
if len(agent) == 0:
doc_agent = prov_doc.agent(agent_uri, agent_record)
else:
doc_agent = agent[0]
# Organization
# Get the organization information from the entity node
org_record = get_organization_record(entity_node)
# Build the organization uri
group_uuid_prov_key = f'{HUBMAP_NAMESPACE}:groupUUID'
org_uri = build_uri(HUBMAP_NAMESPACE, 'organization',
org_record[group_uuid_prov_key])
# Only add the same organization once
# Multiple entities can be associated to different agents who are from the same organization
org = prov_doc.get_record(org_uri)
if len(org) == 0:
doc_org = prov_doc.agent(org_uri, org_record)
else:
doc_org = org[0]
# Build the activity uri
activity_uri = build_uri(HUBMAP_NAMESPACE, 'activities',
activity_node['uuid'])
# Register activity if not already registered
activity = prov_doc.get_record(activity_uri)
if len(activity) == 0:
# Shared attributes to be added to the PROV document
activity_attributes = {'prov:type': 'Activity'}
# Convert the timestampt integer to datetime string
# Note: in our case, prov:startTime is the same as prov:endTime
activity_time = timestamp_to_datetime(
activity_node['created_timestamp'])
# Add prefix to all other attributes
for key in activity_node:
prov_key = f'{HUBMAP_NAMESPACE}:{key}'
# Use datetime string instead of timestamp integer
if key == 'created_timestamp':
activity_attributes[prov_key] = activity_time
else:
activity_attributes[prov_key] = activity_node[key]
# Register activity
doc_activity = prov_doc.activity(activity_uri, activity_time,
activity_time,
activity_attributes)
# Relationship: the agent actedOnBehalfOf the org
prov_doc.actedOnBehalfOf(doc_agent, doc_org, doc_activity)
else:
doc_activity = activity[0]
# Build the entity uri
entity_uri = build_uri(HUBMAP_NAMESPACE, 'entities',
entity_node['uuid'])
# Register entity is not already registered
if len(prov_doc.get_record(entity_uri)) == 0:
# Shared attributes to be added to the PROV document
entity_attributes = {'prov:type': 'Entity'}
# Add prefix to all other attributes
for key in entity_node:
# Entity property values can be list or dict, skip
# And list and dict are unhashable types when calling `prov_doc.entity()`
if not isinstance(entity_node[key], (list, dict)):
prov_key = f'{HUBMAP_NAMESPACE}:{key}'
# Use datetime string instead of timestamp integer
if key in [
'created_timestamp', 'last_modified_timestamp',
'published_timestamp'
]:
entity_attributes[prov_key] = activity_time
else:
entity_attributes[prov_key] = entity_node[key]
# Register entity
prov_doc.entity(entity_uri, entity_attributes)
# Build activity uri and entity uri if not already built
# For the Lab nodes
if activity_uri is None:
activity_uri = build_uri(HUBMAP_NAMESPACE, 'activities',
activity_node['uuid'])
if entity_uri is None:
entity_uri = build_uri(HUBMAP_NAMESPACE, 'entities',
entity_node['uuid'])
# The following relationships apply to all node including Lab entity nodes
# (Activity) - [ACTIVITY_OUTPUT] -> (Entity)
if rel_dict['rel_data']['type'] == 'ACTIVITY_OUTPUT':
# Relationship: the entity wasGeneratedBy the activity
prov_doc.wasGeneratedBy(entity_uri, activity_uri)
# (Entity) - [ACTIVITY_INPUT] -> (Activity)
elif rel_dict['rel_data']['type'] == 'ACTIVITY_INPUT':
# Relationship: the activity used the entity
prov_doc.used(activity_uri, entity_uri)
# Format into json string based on the PROV-JSON Serialization
# https://www.w3.org/Submission/prov-json/
serialized_json = prov_doc.serialize()
return serialized_json
def example():
g = ProvDocument()
# Local namespace
# Doesnt exist yet so we are creating it
ap = Namespace('aip', 'https://araport.org/provenance/')
# Dublin Core
g.add_namespace("dcterms", "http://purl.org/dc/terms/")
# FOAF
g.add_namespace("foaf", "http://xmlns.com/foaf/0.1/")
# Add sponsors and contributors as Agents
# ap['matthew_vaughn']
# aip:matthew_vaughn
# https://araport.org/provenance/:matthew_vaughn
# Learn this from a call to profiles service? Adds a dependency on Agave so I am open to figuring out another way
me = g.agent(
ap['matthew_vaughn'], {
'prov:type': PROV["Person"],
'foaf:givenName': "Matthew Vaughn",
'foaf:mbox': "<mailto:[email protected]>"
})
# Hard coded for now
walter = g.agent(
ap['walter_moreira'], {
'prov:type': PROV["Person"],
'foaf:givenName': "Walter Moreira",
'foaf:mbox': "<mailto:[email protected]>"
})
utexas = g.agent(
ap['university_of_texas'], {
'prov:type': PROV["Organization"],
'foaf:givenName': "University of Texas at Austin"
})
# Set delegation to our host University
# We may have trouble doing this for other users since we don't always capture their host instituion
g.actedOnBehalfOf(walter, utexas)
g.actedOnBehalfOf(me, utexas)
# Include the ADAMA platform as an Agent and set attribution
# dcterms:title and dcterms:description are hardcoded
# dcterms:language is hard-coded
# dcterms:source is the URI of the public git source repository for ADAMA
# "dcterms:updated": "2015-04-17T09:44:56" - this would actually be the date ADAMA was updated
adama_platform = g.agent(
ap['adama_platform'], {
'dcterms:title': "ADAMA",
'dcterms:description': "Araport Data and Microservices API",
'dcterms:language': "en-US",
'dcterms:identifier': "https://api.araport.org/community/v0.3/",
'dcterms:updated': "2015-04-17T09:44:56"
})
g.wasGeneratedBy(adama_platform, walter)
# Include the ADAMA microservice as an Agent and set attribution+delegation
# dcterms:title and dcterms:description are inherited from the service's metadata
# dcterms:language is hard-coded
# dcterms:identifier is the deployment URI for the service
# dcterms:source is the URI of the public git source repository. The URL in this example is just a dummy
#
# The name for each microservice should be unique. We've decided to
# use the combination of namespace, service name, and version
microservice_name = 'mwvaughn/bar_annotation_v1.0.0'
adama_microservice = g.agent(
ap[microservice_name], {
'dcterms:title':
"BAR Annotation Service",
'dcterms:description':
"Returns annotation from locus ID",
'dcterms:language':
"en-US",
'dcterms:identifier':
"https://api.araport.org/community/v0.3/mwvaughn/bar_annotation_v1.0.0",
'dcterms:source':
"https://github.com/Arabidopsis-Information-Portal/prov-enabled-api-sample"
})
# the microservice was generated by me on date X (don't use now, use when the service was updated)
g.wasGeneratedBy(adama_microservice, me, datetime.datetime.now())
# The microservice used the platform now
g.used(adama_microservice, adama_platform, datetime.datetime.now())
# Sources
#
# Define BAR
# Agents
nick = g.agent(
ap['nicholas_provart'], {
'prov:type': PROV["Person"],
'foaf:givenName': "Nicholas Provart",
'foaf:mbox': "*****@*****.**"
})
utoronto = g.agent(
ap['university_of_toronto'], {
'prov:type': PROV["Organization"],
'foaf:givenName': "University of Toronto",
'dcterms:identifier': "http://www.utoronto.ca/"
})
g.actedOnBehalfOf(nick, utoronto)
# Entity
# All fields derived from Sources.yml
# dcterms:title and dcterms:description come straight from the YAML
# dcterms:identifier - URI pointing to the source's canonical URI representation
# optional - dcterms:language: Recommended best practice is to use a controlled vocabulary such as RFC 4646
# optional - dcterms:updated: date the source was published or last updated
# optional - dcterms:license: Simple string or URI to license. Validate URI if provided?
datasource1 = g.entity(
ap['datasource1'], {
'dcterms:title': "BAR Arabidopsis AGI -> Annotation",
'dcterms:description': "Most recent annotation for given AGI",
'dcterms:language': "en-US",
'dcterms:identifier':
"http://bar.utoronto.ca/webservices/agiToAnnot.php",
'dcterms:updated': "2015-04-17T09:44:56",
'dcterms:license': "Creative Commons 3.0"
})
# Set up attribution to Nick
g.wasAttributedTo(datasource1, nick)
# Define TAIR
# Agents
# dcterms:language: Recommended best practice is to use a controlled vocabulary such as RFC 4646
eva = g.agent(ap['eva_huala'], {
'prov:type': PROV["Person"],
'foaf:givenName': "Eva Huala"
})
phoenix = g.agent(
ap['phoenix_bioinformatics'], {
'prov:type': PROV["Organization"],
'foaf:givenName': "Phoenix Bioinformatics"
})
g.actedOnBehalfOf(eva, phoenix)
# Entity
# All fields derived from Sources.yml
# optional - dcterms:citation: Plain text bibliographic citation. If only provided as doi, should we try to validate it?
datasource2 = g.entity(
ap['datasource2'], {
'dcterms:title':
"TAIR",
'dcterms:description':
"The Arabidopsis Information Resource",
'dcterms:language':
"en-US",
'dcterms:identifier':
"https://www.arabidopsis.org/",
'dcterms:citation':
"The Arabidopsis Information Resource (TAIR): improved gene annotation and new tools. Nucleic Acids Research 2011 doi: 10.1093/nar/gkr1090"
})
g.wasAttributedTo(datasource2, eva)
# In Sources.yml, these two sources are nested. Define that relationship here
# There are other types of relationships but we will just use derived from for simplicity in this prototype
g.wasDerivedFrom(ap['datasource1'], ap['datasource2'])
# Depending on which ADAMA microservice type we are using, define an activity
# Eventually, break these into more atomic actions in a chain
action1 = g.activity(ap['do_query'], datetime.datetime.now())
# action1 = g.activity(ap['do_map'], datetime.datetime.now())
# action1 = g.activity(ap['do_generic'], datetime.datetime.now())
# action1 = g.activity(ap['do_passthrough'], datetime.datetime.now())
# Future... Support for ADAMA-native microservices
# action1 = g.activity(ap['generate'], datetime.datetime.now())
# Define current ADAMA response as an Entity
# This is what's being returned to the user and is thus the subject of the PROV record
# May be able to add more attributes to it but this is the minimum
response = g.entity(ap['adama_response'])
# Response is generated by the process_query action
# Time-stamp it!
g.wasGeneratedBy(response, ap['do_query'], datetime.datetime.now())
# The process_query used the microservice
g.used(ap['do_query'], adama_microservice, datetime.datetime.now())
# The microservice used datasource1
g.used(adama_microservice, datasource1, datetime.datetime.now())
# Print prov_n
print(g.get_provn())
# Print prov-json
print(g.serialize())
# Write out as a pretty picture
graph = prov.dot.prov_to_dot(g)
graph.write_png('Sources.png')
def write_targets_prov(self, tlist, C, bundle_id):
#Initialisation
# cs = b.agent('CrowdScanner')
if self.document_id == -1:
d = ProvDocument()
d.add_namespace(AO)
d.set_default_namespace(self.defaultns % self.game_id)
if uploadprov:
provstore_document = self.api.document.create(d, name="Operation%s CrowdScanner" % self.game_id, public=True)
document_uri = provstore_document.url
logging.info("prov doc URI: " + str(document_uri))
self.provfilelist.append(provstore_document.id)
self.savelocalrecord()
self.document_id = provstore_document.id
b = ProvDocument() # Create a new document for this update
b.add_namespace(AO)
b.set_default_namespace(self.defaultns % self.game_id)
# cs to be used with all targets
cs = b.agent('agent/CrowdScanner', (('prov:type', AO['IBCCAlgo']), ('prov:type', PROV['SoftwareAgent'])))
timestamp = time.time() # Record the timestamp at each update to generate unique identifiers
startTime = datetime.datetime.fromtimestamp(timestamp)
endTime = startTime
activity = b.activity('activity/cs/update_report_%s' % timestamp, startTime, endTime)
activity.wasAssociatedWith(cs)
#Add target and report entities
for i, tdata in enumerate(tlist):
if self.changedtargets[i]==0:
continue
#Target entity for target i
tid = int(tdata[0])
x = tdata[1]
y = tdata[2]
# targettype = tdata[3] #don't record here, it will be revealed and recorded by UAVs
v = int(tdata[4])
agentids = tdata[7]
targetattributes = {'ao:longitude': x, 'ao:latitude': y, }
#'ao:asset_type':str(targettype)}
target_v0 = b.entity('cs/target/'+str(tid)+'.'+str(v), targetattributes)
#Post the root report if this is the first version
if v==0:
self.targets[tid] = b.entity('cs/target/'+str(tid))
else:
try:
target_v0.wasDerivedFrom(self.targetversions[tid])
except KeyError:
logging.error("Got a key error for key " + str(tid) + ', which is supposed to be version' + str(v))
self.targetversions[tid] = target_v0
target_v0.specializationOf(self.targets[tid])
target_v0.wasAttributedTo(cs)
#Report entities for origins of target i
for j, r in enumerate(self.target_rep_ids[i]):
if r not in self.postedreports:
Crow = C[r,:]
x = Crow[1]
y = Crow[2]
reptext = tdata[5][j].decode('utf8')
# Try to replace unusual characters
reptext = reptext.encode('ascii', 'replace')
agentid = agentids[j]
reporter_name = 'agent/crowdreporter%s' % agentid
b.agent(reporter_name, (('prov:type', AO['CrowdReporter']), ('prov:type', PROV['Person'])))
reportattributes = {'ao:longitude': x, 'ao:latitude': y, 'ao:report': reptext}
self.postedreports[r] = b.entity('cs/report/'+str(r), reportattributes)
self.postedreports[r].wasAttributedTo(reporter_name)
activity.used(self.postedreports[r])
target_v0.wasDerivedFrom(self.postedreports[r])
if uploadprov:
#Invalidate old targets no longer in use
for i,tid in enumerate(self.targets_to_invalidate):
target_v = self.targetversions[tid]
b.wasInvalidatedBy(target_v, activity)
#Post the document to the server
#bundle = b.bundle('crowd_scanner')
bundle_id = 'bundle/csupdate/%s' % timestamp
self.api.add_bundle(self.document_id, b.serialize(), bundle_id)
def get_blank_prov_document():
return ProvDocument(namespaces=all_namespaces)
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser(
description=
"Process and generate provenance for a MIMIC patient admission")
parser.add_argument("admission_id",
type=int,
help="The ID of admission to process")
args = parser.parse_args()
prov_doc1 = ProvDocument(namespaces=all_namespaces)
admission = Admission(prov_doc1, args.admission_id)
admission.process()
filepath = output_path / f"{args.admission_id}.json"
with filepath.open("w") as f:
prov_doc1.serialize(f)
provn_content = prov_doc1.get_provn()
print(provn_content)
with filepath.with_suffix(".provn").open("w") as f:
f.write(provn_content)
dot = prov_to_dot(prov_doc1)
dot.write_pdf(filepath.with_suffix(".pdf"))
db.close_session()
def example():
g = ProvDocument()
# Local namespace
# Doesnt exist yet so we are creating it
ap = Namespace('aip', 'https://araport.org/provenance/')
# Dublin Core
g.add_namespace("dcterms", "http://purl.org/dc/terms/")
# FOAF
g.add_namespace("foaf", "http://xmlns.com/foaf/0.1/")
# Add sponsors and contributors as Agents
# ap['matthew_vaughn']
# aip:matthew_vaughn
# https://araport.org/provenance/:matthew_vaughn
# Learn this from a call to profiles service? Adds a dependency on Agave so I am open to figuring out another way
me = g.agent(ap['matthew_vaughn'], {
'prov:type': PROV["Person"], 'foaf:givenName': "Matthew Vaughn", 'foaf:mbox': "<mailto:[email protected]>"
})
# Hard coded for now
walter = g.agent(ap['walter_moreira'], {
'prov:type': PROV["Person"], 'foaf:givenName': "Walter Moreira", 'foaf:mbox': "<mailto:[email protected]>"
})
utexas = g.agent(ap['university_of_texas'], {
'prov:type': PROV["Organization"], 'foaf:givenName': "University of Texas at Austin"
})
# Set delegation to our host University
# We may have trouble doing this for other users since we don't always capture their host instituion
g.actedOnBehalfOf(walter, utexas)
g.actedOnBehalfOf(me, utexas)
# Include the ADAMA platform as an Agent and set attribution
# dcterms:title and dcterms:description are hardcoded
# dcterms:language is hard-coded
# dcterms:source is the URI of the public git source repository for ADAMA
# "dcterms:updated": "2015-04-17T09:44:56" - this would actually be the date ADAMA was updated
adama_platform = g.agent(ap['adama_platform'], {'dcterms:title': "ADAMA", 'dcterms:description': "Araport Data and Microservices API", 'dcterms:language':"en-US", 'dcterms:identifier':"https://api.araport.org/community/v0.3/", 'dcterms:updated': "2015-04-17T09:44:56" })
g.wasGeneratedBy(adama_platform, walter)
# Include the ADAMA microservice as an Agent and set attribution+delegation
# dcterms:title and dcterms:description are inherited from the service's metadata
# dcterms:language is hard-coded
# dcterms:identifier is the deployment URI for the service
# dcterms:source is the URI of the public git source repository. The URL in this example is just a dummy
#
# The name for each microservice should be unique. We've decided to
# use the combination of namespace, service name, and version
microservice_name = 'mwvaughn/bar_annotation_v1.0.0'
adama_microservice = g.agent(ap[microservice_name], {'dcterms:title': "BAR Annotation Service", 'dcterms:description': "Returns annotation from locus ID", 'dcterms:language':"en-US", 'dcterms:identifier':"https://api.araport.org/community/v0.3/mwvaughn/bar_annotation_v1.0.0", 'dcterms:source':"https://github.com/Arabidopsis-Information-Portal/prov-enabled-api-sample" })
# the microservice was generated by me on date X (don't use now, use when the service was updated)
g.wasGeneratedBy(adama_microservice, me, datetime.datetime.now())
# The microservice used the platform now
g.used(adama_microservice, adama_platform, datetime.datetime.now())
# Sources
#
# Define BAR
# Agents
nick = g.agent(ap['nicholas_provart'], {
'prov:type': PROV["Person"], 'foaf:givenName': "Nicholas Provart", 'foaf:mbox': "*****@*****.**"
})
utoronto = g.agent(ap['university_of_toronto'], {
'prov:type': PROV["Organization"], 'foaf:givenName': "University of Toronto", 'dcterms:identifier':"http://www.utoronto.ca/"
})
g.actedOnBehalfOf(nick, utoronto)
# Entity
# All fields derived from Sources.yml
# dcterms:title and dcterms:description come straight from the YAML
# dcterms:identifier - URI pointing to the source's canonical URI representation
# optional - dcterms:language: Recommended best practice is to use a controlled vocabulary such as RFC 4646
# optional - dcterms:updated: date the source was published or last updated
# optional - dcterms:license: Simple string or URI to license. Validate URI if provided?
datasource1 = g.entity(ap['datasource1'], {'dcterms:title': "BAR Arabidopsis AGI -> Annotation", 'dcterms:description': "Most recent annotation for given AGI", 'dcterms:language':"en-US", 'dcterms:identifier':"http://bar.utoronto.ca/webservices/agiToAnnot.php", 'dcterms:updated':"2015-04-17T09:44:56", 'dcterms:license':"Creative Commons 3.0" })
# Set up attribution to Nick
g.wasAttributedTo(datasource1, nick)
# Define TAIR
# Agents
# dcterms:language: Recommended best practice is to use a controlled vocabulary such as RFC 4646
eva = g.agent(ap['eva_huala'], {
'prov:type': PROV["Person"], 'foaf:givenName': "Eva Huala"
})
phoenix = g.agent(ap['phoenix_bioinformatics'], {
'prov:type': PROV["Organization"], 'foaf:givenName': "Phoenix Bioinformatics"
})
g.actedOnBehalfOf(eva, phoenix)
# Entity
# All fields derived from Sources.yml
# optional - dcterms:citation: Plain text bibliographic citation. If only provided as doi, should we try to validate it?
datasource2 = g.entity(ap['datasource2'], {'dcterms:title': "TAIR", 'dcterms:description': "The Arabidopsis Information Resource", 'dcterms:language':"en-US", 'dcterms:identifier':"https://www.arabidopsis.org/", 'dcterms:citation':"The Arabidopsis Information Resource (TAIR): improved gene annotation and new tools. Nucleic Acids Research 2011 doi: 10.1093/nar/gkr1090"})
g.wasAttributedTo(datasource2, eva)
# In Sources.yml, these two sources are nested. Define that relationship here
# There are other types of relationships but we will just use derived from for simplicity in this prototype
g.wasDerivedFrom(ap['datasource1'], ap['datasource2'])
# Depending on which ADAMA microservice type we are using, define an activity
# Eventually, break these into more atomic actions in a chain
action1 = g.activity(ap['do_query'], datetime.datetime.now())
# action1 = g.activity(ap['do_map'], datetime.datetime.now())
# action1 = g.activity(ap['do_generic'], datetime.datetime.now())
# action1 = g.activity(ap['do_passthrough'], datetime.datetime.now())
# Future... Support for ADAMA-native microservices
# action1 = g.activity(ap['generate'], datetime.datetime.now())
# Define current ADAMA response as an Entity
# This is what's being returned to the user and is thus the subject of the PROV record
# May be able to add more attributes to it but this is the minimum
response = g.entity(ap['adama_response'])
# Response is generated by the process_query action
# Time-stamp it!
g.wasGeneratedBy(response, ap['do_query'], datetime.datetime.now())
# The process_query used the microservice
g.used(ap['do_query'], adama_microservice, datetime.datetime.now())
# The microservice used datasource1
g.used(adama_microservice, datasource1, datetime.datetime.now())
# Print prov_n
print(g.get_provn())
# Print prov-json
print(g.serialize())
# Write out as a pretty picture
graph = prov.dot.prov_to_dot(g)
graph.write_png('Sources.png')
def get_provenance_history(self, driver, uuid, depth=None):
prov_doc = ProvDocument()
#prov_doc.
#NOTE!! There is a bug with the JSON serializer. I can't add the prov prefix using this mechanism
prov_doc.add_namespace('ex', 'http://example.org/')
prov_doc.add_namespace('hubmap', 'https://hubmapconsortium.org/')
#prov_doc.add_namespace('dct', 'http://purl.org/dc/terms/')
#prov_doc.add_namespace('foaf','http://xmlns.com/foaf/0.1/')
relation_list = []
with driver.session() as session:
try:
# max_level_str is the string used to put a limit on the number of levels to traverse
max_level_str = ''
if depth is not None and len(str(depth)) > 0:
max_level_str = """maxLevel: {depth},""".format(
depth=depth)
"""
Basically this Cypher query returns a collection of nodes and relationships. The relationships include ACTIVITY_INPUT, ACTIVITY_OUTPUT and
HAS_METADATA. First, we build a dictionary of the nodes using uuid as a key. Next, we loop through the relationships looking for HAS_METADATA
relationships. The HAS_METADATA relationships connect the Entity nodes with their metadata. The data from the Metadata node
becomes the 'metadata' attribute for the Entity node.
"""
"""Possible replacement:
THIS WORKS...NEEDS LOTS of COMMENTS!!
MATCH (entity_metadata)<-[r1:HAS_METADATA]-(e)<-[r2:ACTIVITY_OUTPUT]-(a:Activity)-[r3:HAS_METADATA]->(activity_metadata)
WHERE e.hubmap_identifier = 'TEST0010-LK-1-1'
WITH [e,a, entity_metadata, activity_metadata] AS entities, COLLECT(r1) + COLLECT(r2) + COLLECT(r3) AS relationships
WITH [node in entities | node {.*, label:labels(node)}] AS nodes, [rel in relationships | rel { .*, fromNode: { label:labels(startNode(rel))[0], uuid:startNode(rel).uuid } , toNode: { label:labels(endNode(rel))[0], uuid:endNode(rel).uuid }, rel_data: { type: type(rel) } } ] as rels
RETURN nodes, rels
UNION OPTIONAL MATCH (activity_metadata)<-[r1:HAS_METADATA]-(a:Activity)<-[r2:ACTIVITY_INPUT|:ACTIVITY_OUTPUT*]-(parent)-[r3:HAS_METADATA]->(parent_metadata),
(e)<-[r4:ACTIVITY_OUTPUT]-(a:Activity)
WHERE e.hubmap_identifier = 'TEST0010-LK-1-1'
WITH [parent,parent_metadata, a, activity_metadata] AS nodes, [rel in COLLECT(r1) + COLLECT(r3) + COLLECT(r4)+COLLECT(apoc.convert.toRelationship(r2)) | rel { .*, fromNode: { label:labels(startNode(rel))[0], uuid:startNode(rel).uuid } , toNode: { label:labels(endNode(rel))[0], uuid:endNode(rel).uuid }, rel_data: { type: type(rel) } } ] as rels
RETURN DISTINCT nodes, rels
uuid for TEST0010-LK-1-1 for testing: eda3916db4695d834eb6c51a893d06f1
"""
stmt = """MATCH (n:Entity {{ uuid: '{uuid}' }})
CALL apoc.path.subgraphAll(n, {{ {max_level_str} relationshipFilter:'<ACTIVITY_INPUT|<ACTIVITY_OUTPUT|HAS_METADATA>' }}) YIELD nodes, relationships
WITH [node in nodes | node {{ .*, label:labels(node)[0] }} ] as nodes,
[rel in relationships | rel {{ .*, fromNode: {{ label:labels(startNode(rel))[0], uuid:startNode(rel).uuid }} , toNode: {{ label:labels(endNode(rel))[0], uuid:endNode(rel).uuid }}, rel_data: {{ type: type(rel) }} }} ] as rels
WITH {{ nodes:nodes, relationships:rels }} as json
RETURN json""".format(uuid=uuid, max_level_str=max_level_str)
result = session.run(stmt)
#there should only be one record
for jsonData in result:
try:
record = dict(jsonData)['json']
if 'relationships' not in record:
raise LookupError(
'Error, unable to find relationships for uuid:'
+ uuid)
if 'nodes' not in record:
raise LookupError(
'Error, unable to find nodes for uuid:' + uuid)
node_dict = {}
# pack the nodes into a dictionary using the uuid as a key
for node_record in record['nodes']:
node_dict[node_record['uuid']] = node_record
# TODO: clean up nodes
# remove nodes that lack metadata
# need to devise a methodology for this
# try preprocessing the record['relationships'] here:
# make a copy of the node_dict called unreferenced_node_dict
# loop through the relationships and find all the has_metadata relationships
# for each node pair in the has_metadata relationship, delete it from the unreferenced_node_dict
# once the loop is finished, continue as before
# add some logic when generating the wasGenerated and used relationships. If either node is in the
# unreferenced_node_dict, then ignore the relationship
# now, connect the nodes
for rel_record in record['relationships']:
from_uuid = rel_record['fromNode']['uuid']
to_uuid = rel_record['toNode']['uuid']
from_node = node_dict[from_uuid]
to_node = node_dict[to_uuid]
if rel_record['rel_data'][
'type'] == HubmapConst.HAS_METADATA_REL:
# assign the metadata node as the metadata attribute
# just extract the provenance information from the metadata node
entity_timestamp_json = Provenance.get_json_timestamp(
int(to_node[
HubmapConst.
PROVENANCE_CREATE_TIMESTAMP_ATTRIBUTE])
)
provenance_data = {
ProvConst.PROV_GENERATED_TIME_ATTRIBUTE:
entity_timestamp_json
}
type_code = None
isEntity = True
if HubmapConst.ENTITY_TYPE_ATTRIBUTE in from_node:
type_code = from_node[
HubmapConst.ENTITY_TYPE_ATTRIBUTE]
elif HubmapConst.ACTIVITY_TYPE_ATTRIBUTE in from_node:
type_code = from_node[
HubmapConst.ACTIVITY_TYPE_ATTRIBUTE]
isEntity = False
label_text = None
if HubmapConst.LAB_IDENTIFIER_ATTRIBUTE in from_node:
label_text = from_node[
HubmapConst.LAB_IDENTIFIER_ATTRIBUTE]
else:
label_text = from_node[
HubmapConst.UUID_ATTRIBUTE]
# build metadata attribute from the Metadata node
metadata_attribute = {}
for attribute_key in to_node:
if attribute_key not in self.metadata_ignore_attributes:
if attribute_key in self.known_attribute_map:
# special case: timestamps
if attribute_key == HubmapConst.PROVENANCE_MODIFIED_TIMESTAMP_ATTRIBUTE:
provenance_data[
self.known_attribute_map[
attribute_key]] = Provenance.get_json_timestamp(
int(to_node[
attribute_key])
)
else: #add any extraneous data to the metadata attribute
metadata_attribute[
attribute_key] = to_node[
attribute_key]
# Need to add the agent and organization here, plus the appropriate relationships (between the entity and the agent plus orgainzation)
agent_record = self.get_agent_record(to_node)
agent_unique_id = str(agent_record[
ProvConst.HUBMAP_PROV_USER_EMAIL]).replace(
'@', '-')
agent_unique_id = str(agent_unique_id).replace(
'.', '-')
if ProvConst.HUBMAP_PROV_USER_UUID in agent_record:
agent_unique_id = agent_record[
ProvConst.HUBMAP_PROV_USER_UUID]
agent_uri = Provenance.build_uri(
'hubmap', 'agent', agent_unique_id)
organization_record = self.get_organization_record(
to_node)
organization_uri = Provenance.build_uri(
'hubmap', 'organization',
organization_record[
ProvConst.HUBMAP_PROV_GROUP_UUID])
doc_agent = None
doc_org = None
get_agent = prov_doc.get_record(agent_uri)
# only add this once
if len(get_agent) == 0:
doc_agent = prov_doc.agent(
agent_uri, agent_record)
else:
doc_agent = get_agent[0]
get_org = prov_doc.get_record(organization_uri)
# only add this once
if len(get_org) == 0:
doc_org = prov_doc.agent(
organization_uri, organization_record)
else:
doc_org = get_org[0]
other_attributes = {
ProvConst.PROV_LABEL_ATTRIBUTE:
label_text,
ProvConst.PROV_TYPE_ATTRIBUTE:
type_code,
ProvConst.HUBMAP_DOI_ATTRIBUTE:
from_node[HubmapConst.DOI_ATTRIBUTE],
ProvConst.HUBMAP_DISPLAY_DOI_ATTRIBUTE:
from_node[
HubmapConst.DISPLAY_DOI_ATTRIBUTE],
ProvConst.HUBMAP_DISPLAY_IDENTIFIER_ATTRIBUTE:
label_text,
ProvConst.HUBMAP_UUID_ATTRIBUTE:
from_node[HubmapConst.UUID_ATTRIBUTE]
}
# only add metadata if it contains data
if len(metadata_attribute) > 0:
other_attributes[
ProvConst.
HUBMAP_METADATA_ATTRIBUTE] = json.dumps(
metadata_attribute)
# add the provenance data to the other_attributes
other_attributes.update(provenance_data)
if isEntity == True:
prov_doc.entity(
Provenance.build_uri(
'hubmap', 'entities',
from_node['uuid']),
other_attributes)
else:
activity_timestamp_json = Provenance.get_json_timestamp(
int(to_node[
HubmapConst.
PROVENANCE_CREATE_TIMESTAMP_ATTRIBUTE]
))
doc_activity = prov_doc.activity(
Provenance.build_uri(
'hubmap', 'activities',
from_node['uuid']),
activity_timestamp_json,
activity_timestamp_json,
other_attributes)
prov_doc.actedOnBehalfOf(
doc_agent, doc_org, doc_activity)
elif rel_record['rel_data']['type'] in [
HubmapConst.ACTIVITY_OUTPUT_REL,
HubmapConst.ACTIVITY_INPUT_REL
]:
to_node_uri = None
from_node_uri = None
if HubmapConst.ENTITY_TYPE_ATTRIBUTE in to_node:
to_node_uri = Provenance.build_uri(
'hubmap', 'entities', to_node['uuid'])
else:
to_node_uri = Provenance.build_uri(
'hubmap', 'activities',
to_node['uuid'])
if HubmapConst.ENTITY_TYPE_ATTRIBUTE in from_node:
from_node_uri = Provenance.build_uri(
'hubmap', 'entities',
from_node['uuid'])
else:
from_node_uri = Provenance.build_uri(
'hubmap', 'activities',
from_node['uuid'])
if rel_record['rel_data'][
'type'] == 'ACTIVITY_OUTPUT':
#prov_doc.wasGeneratedBy(entity, activity, time, identifier, other_attributes)
prov_doc.wasGeneratedBy(
to_node_uri, from_node_uri)
if rel_record['rel_data'][
'type'] == 'ACTIVITY_INPUT':
#prov_doc.used(activity, entity, time, identifier, other_attributes)
prov_doc.used(to_node_uri, from_node_uri)
# for now, simply create a "relation" where the fromNode's uuid is connected to a toNode's uuid via a relationship:
# ex: {'fromNodeUUID': '42e10053358328c9079f1c8181287b6d', 'relationship': 'ACTIVITY_OUTPUT', 'toNodeUUID': '398400024fda58e293cdb435db3c777e'}
rel_data_record = {
'fromNodeUUID': from_node['uuid'],
'relationship':
rel_record['rel_data']['type'],
'toNodeUUID': to_node['uuid']
}
relation_list.append(rel_data_record)
return_data = {
'nodes': node_dict,
'relations': relation_list
}
except Exception as e:
raise e
# there is a bug in the JSON serializer. So manually insert the prov prefix
output_doc = prov_doc.serialize(indent=2)
output_doc = output_doc.replace(
'"prefix": {',
'"prefix": {\n "prov" : "http://www.w3.org/ns/prov#", ')
#output_doc = prov_doc.serialize(format='rdf', rdf_format='trig')
#output_doc = prov_doc.serialize(format='provn')
return output_doc
except ConnectionError as ce:
print('A connection error occurred: ', str(ce.args[0]))
raise ce
except ValueError as ve:
print('A value error occurred: ', ve.value)
raise ve
except Exception as e:
print('An exception occurred in get_provenance_history: ' +
str(e))
traceback.print_exc()
logpage_ident = get_logpage(str(logpage), prov_doc)
return plate_ident
# Create a new provenance document
d1 = ProvDocument()
declare_namespaces(d1)
# get V468Cyg
# get_plate
# process = get_process('2180', d1)
try:
# scan = get_entity('2462','scan', d1)
id = '2180'
prov_type = 'lightcurve'
# plate_name = get_entity(id,prov_type, d1)
# process_name = get_process('9804',d1)
# logpage_name = get_logpage('10085',d1)
# source_id = get_source('40000001', d1)
# plate_ident = get_plate_prov(id, d1)
id = get_lightcurve('614-089373', d1)
except TypeError:
print('the job is still executing...')
print(d1.get_provn())
filename = 'prov_' + prov_type + id
d1.serialize(filename + '.xml', format='xml')
dot = prov_to_dot(d1)
dot.write_png(filename + '.png')
def declare_directory(self, value: CWLObjectType) -> ProvEntity:
"""Register any nested files/directories."""
# FIXME: Calculate a hash-like identifier for directory
# so we get same value if it's the same filenames/hashes
# in a different location.
# For now, mint a new UUID to identify this directory, but
# attempt to keep it inside the value dictionary
dir_id = cast(str, value.setdefault("@id", uuid.uuid4().urn))
# New annotation file to keep the ORE Folder listing
ore_doc_fn = dir_id.replace("urn:uuid:", "directory-") + ".ttl"
dir_bundle = self.document.bundle(self.metadata_ns[ore_doc_fn])
coll = self.document.entity(
dir_id,
[
(PROV_TYPE, WFPROV["Artifact"]),
(PROV_TYPE, PROV["Collection"]),
(PROV_TYPE, PROV["Dictionary"]),
(PROV_TYPE, RO["Folder"]),
],
)
# ORE description of ro:Folder, saved separately
coll_b = dir_bundle.entity(
dir_id,
[(PROV_TYPE, RO["Folder"]), (PROV_TYPE, ORE["Aggregation"])],
)
self.document.mentionOf(dir_id + "#ore", dir_id, dir_bundle.identifier)
# dir_manifest = dir_bundle.entity(
# dir_bundle.identifier, {PROV["type"]: ORE["ResourceMap"],
# ORE["describes"]: coll_b.identifier})
coll_attribs = [(ORE["isDescribedBy"], dir_bundle.identifier)]
coll_b_attribs = [] # type: List[Tuple[Identifier, ProvEntity]]
# FIXME: .listing might not be populated yet - hopefully
# a later call to this method will sort that
is_empty = True
if "listing" not in value:
get_listing(self.fsaccess, value)
for entry in cast(MutableSequence[CWLObjectType],
value.get("listing", [])):
is_empty = False
# Declare child-artifacts
entity = self.declare_artefact(entry)
self.document.membership(coll, entity)
# Membership relation aka our ORE Proxy
m_id = uuid.uuid4().urn
m_entity = self.document.entity(m_id)
m_b = dir_bundle.entity(m_id)
# PROV-O style Dictionary
# https://www.w3.org/TR/prov-dictionary/#dictionary-ontological-definition
# ..as prov.py do not currently allow PROV-N extensions
# like hadDictionaryMember(..)
m_entity.add_asserted_type(PROV["KeyEntityPair"])
m_entity.add_attributes({
PROV["pairKey"]: entry["basename"],
PROV["pairEntity"]: entity,
})
# As well as a being a
# http://wf4ever.github.io/ro/2016-01-28/ro/#FolderEntry
m_b.add_asserted_type(RO["FolderEntry"])
m_b.add_asserted_type(ORE["Proxy"])
m_b.add_attributes({
RO["entryName"]: entry["basename"],
ORE["proxyIn"]: coll,
ORE["proxyFor"]: entity,
})
coll_attribs.append((PROV["hadDictionaryMember"], m_entity))
coll_b_attribs.append((ORE["aggregates"], m_b))
coll.add_attributes(coll_attribs)
coll_b.add_attributes(coll_b_attribs)
# Also Save ORE Folder as annotation metadata
ore_doc = ProvDocument()
ore_doc.add_namespace(ORE)
ore_doc.add_namespace(RO)
ore_doc.add_namespace(UUID)
ore_doc.add_bundle(dir_bundle)
ore_doc = ore_doc.flattened()
ore_doc_path = str(PurePosixPath(METADATA, ore_doc_fn))
with self.research_object.write_bag_file(
ore_doc_path) as provenance_file:
ore_doc.serialize(provenance_file,
format="rdf",
rdf_format="turtle")
self.research_object.add_annotation(dir_id, [ore_doc_fn],
ORE["isDescribedBy"].uri)
if is_empty:
# Empty directory
coll.add_asserted_type(PROV["EmptyCollection"])
coll.add_asserted_type(PROV["EmptyDictionary"])
self.research_object.add_uri(coll.identifier.uri)
return coll
class ProvenanceProfile:
"""
Provenance profile.
Populated as the workflow runs.
"""
def __init__(
self,
research_object: "ResearchObject",
full_name: str,
host_provenance: bool,
user_provenance: bool,
orcid: str,
fsaccess: StdFsAccess,
run_uuid: Optional[uuid.UUID] = None,
) -> None:
"""Initialize the provenance profile."""
self.fsaccess = fsaccess
self.orcid = orcid
self.research_object = research_object
self.folder = self.research_object.folder
self.document = ProvDocument()
self.host_provenance = host_provenance
self.user_provenance = user_provenance
self.engine_uuid = research_object.engine_uuid # type: str
self.add_to_manifest = self.research_object.add_to_manifest
if self.orcid:
_logger.debug("[provenance] Creator ORCID: %s", self.orcid)
self.full_name = full_name
if self.full_name:
_logger.debug("[provenance] Creator Full name: %s", self.full_name)
self.workflow_run_uuid = run_uuid or uuid.uuid4()
self.workflow_run_uri = self.workflow_run_uuid.urn # type: str
self.generate_prov_doc()
def __str__(self) -> str:
"""Represent this Provenvance profile as a string."""
return "ProvenanceProfile <{}> in <{}>".format(
self.workflow_run_uri,
self.research_object,
)
def generate_prov_doc(self) -> Tuple[str, ProvDocument]:
"""Add basic namespaces."""
def host_provenance(document: ProvDocument) -> None:
"""Record host provenance."""
document.add_namespace(CWLPROV)
document.add_namespace(UUID)
document.add_namespace(FOAF)
hostname = getfqdn()
# won't have a foaf:accountServiceHomepage for unix hosts, but
# we can at least provide hostname
document.agent(
ACCOUNT_UUID,
{
PROV_TYPE: FOAF["OnlineAccount"],
"prov:location": hostname,
CWLPROV["hostname"]: hostname,
},
)
self.cwltool_version = "cwltool %s" % versionstring().split()[-1]
self.document.add_namespace("wfprov",
"http://purl.org/wf4ever/wfprov#")
# document.add_namespace('prov', 'http://www.w3.org/ns/prov#')
self.document.add_namespace("wfdesc",
"http://purl.org/wf4ever/wfdesc#")
# TODO: Make this ontology. For now only has cwlprov:image
self.document.add_namespace("cwlprov", "https://w3id.org/cwl/prov#")
self.document.add_namespace("foaf", "http://xmlns.com/foaf/0.1/")
self.document.add_namespace("schema", "http://schema.org/")
self.document.add_namespace("orcid", "https://orcid.org/")
self.document.add_namespace("id", "urn:uuid:")
# NOTE: Internet draft expired 2004-03-04 (!)
# https://tools.ietf.org/html/draft-thiemann-hash-urn-01
# TODO: Change to nih:sha-256; hashes
# https://tools.ietf.org/html/rfc6920#section-7
self.document.add_namespace("data", "urn:hash::sha1:")
# Also needed for docker images
self.document.add_namespace(SHA256, "nih:sha-256;")
# info only, won't really be used by prov as sub-resources use /
self.document.add_namespace("researchobject",
self.research_object.base_uri)
# annotations
self.metadata_ns = self.document.add_namespace(
"metadata", self.research_object.base_uri + METADATA + "/")
# Pre-register provenance directory so we can refer to its files
self.provenance_ns = self.document.add_namespace(
"provenance",
self.research_object.base_uri + posix_path(PROVENANCE) + "/")
ro_identifier_workflow = self.research_object.base_uri + "workflow/packed.cwl#"
self.wf_ns = self.document.add_namespace("wf", ro_identifier_workflow)
ro_identifier_input = (self.research_object.base_uri +
"workflow/primary-job.json#")
self.document.add_namespace("input", ro_identifier_input)
# More info about the account (e.g. username, fullname)
# may or may not have been previously logged by user_provenance()
# .. but we always know cwltool was launched (directly or indirectly)
# by a user account, as cwltool is a command line tool
account = self.document.agent(ACCOUNT_UUID)
if self.orcid or self.full_name:
person = {PROV_TYPE: PROV["Person"], "prov:type": SCHEMA["Person"]}
if self.full_name:
person["prov:label"] = self.full_name
person["foaf:name"] = self.full_name
person["schema:name"] = self.full_name
else:
# TODO: Look up name from ORCID API?
pass
agent = self.document.agent(self.orcid or uuid.uuid4().urn, person)
self.document.actedOnBehalfOf(account, agent)
else:
if self.host_provenance:
host_provenance(self.document)
if self.user_provenance:
self.research_object.user_provenance(self.document)
# The execution of cwltool
wfengine = self.document.agent(
self.engine_uuid,
{
PROV_TYPE: PROV["SoftwareAgent"],
"prov:type": WFPROV["WorkflowEngine"],
"prov:label": self.cwltool_version,
},
)
# FIXME: This datetime will be a bit too delayed, we should
# capture when cwltool.py earliest started?
self.document.wasStartedBy(wfengine, None, account,
datetime.datetime.now())
# define workflow run level activity
self.document.activity(
self.workflow_run_uri,
datetime.datetime.now(),
None,
{
PROV_TYPE: WFPROV["WorkflowRun"],
"prov:label": "Run of workflow/packed.cwl#main",
},
)
# association between SoftwareAgent and WorkflowRun
main_workflow = "wf:main"
self.document.wasAssociatedWith(self.workflow_run_uri,
self.engine_uuid, main_workflow)
self.document.wasStartedBy(self.workflow_run_uri, None,
self.engine_uuid, datetime.datetime.now())
return (self.workflow_run_uri, self.document)
def evaluate(
self,
process: Process,
job: JobsType,
job_order_object: CWLObjectType,
research_obj: "ResearchObject",
) -> None:
"""Evaluate the nature of job."""
if not hasattr(process, "steps"):
# record provenance of independent commandline tool executions
self.prospective_prov(job)
customised_job = copy_job_order(job, job_order_object)
self.used_artefacts(customised_job, self.workflow_run_uri)
research_obj.create_job(customised_job)
elif hasattr(job, "workflow"):
# record provenance of workflow executions
self.prospective_prov(job)
customised_job = copy_job_order(job, job_order_object)
self.used_artefacts(customised_job, self.workflow_run_uri)
def record_process_start(
self,
process: Process,
job: JobsType,
process_run_id: Optional[str] = None) -> Optional[str]:
if not hasattr(process, "steps"):
process_run_id = self.workflow_run_uri
elif not hasattr(job, "workflow"):
# commandline tool execution as part of workflow
name = ""
if isinstance(job, (CommandLineJob, JobBase, WorkflowJob)):
name = job.name
process_name = urllib.parse.quote(name, safe=":/,#")
process_run_id = self.start_process(process_name,
datetime.datetime.now())
return process_run_id
def start_process(
self,
process_name: str,
when: datetime.datetime,
process_run_id: Optional[str] = None,
) -> str:
"""Record the start of each Process."""
if process_run_id is None:
process_run_id = uuid.uuid4().urn
prov_label = "Run of workflow/packed.cwl#main/" + process_name
self.document.activity(
process_run_id,
None,
None,
{
PROV_TYPE: WFPROV["ProcessRun"],
PROV_LABEL: prov_label
},
)
self.document.wasAssociatedWith(process_run_id, self.engine_uuid,
str("wf:main/" + process_name))
self.document.wasStartedBy(process_run_id, None, self.workflow_run_uri,
when, None, None)
return process_run_id
def record_process_end(
self,
process_name: str,
process_run_id: str,
outputs: Union[CWLObjectType, MutableSequence[CWLObjectType], None],
when: datetime.datetime,
) -> None:
self.generate_output_prov(outputs, process_run_id, process_name)
self.document.wasEndedBy(process_run_id, None, self.workflow_run_uri,
when)
def declare_file(
self, value: CWLObjectType) -> Tuple[ProvEntity, ProvEntity, str]:
if value["class"] != "File":
raise ValueError("Must have class:File: %s" % value)
# Need to determine file hash aka RO filename
entity = None # type: Optional[ProvEntity]
checksum = None
if "checksum" in value:
csum = cast(str, value["checksum"])
(method, checksum) = csum.split("$", 1)
if method == SHA1 and self.research_object.has_data_file(checksum):
entity = self.document.entity("data:" + checksum)
if not entity and "location" in value:
location = str(value["location"])
# If we made it here, we'll have to add it to the RO
with self.fsaccess.open(location, "rb") as fhandle:
relative_path = self.research_object.add_data_file(fhandle)
# FIXME: This naively relies on add_data_file setting hash as filename
checksum = PurePath(relative_path).name
entity = self.document.entity("data:" + checksum,
{PROV_TYPE: WFPROV["Artifact"]})
if "checksum" not in value:
value["checksum"] = f"{SHA1}${checksum}"
if not entity and "contents" in value:
# Anonymous file, add content as string
entity, checksum = self.declare_string(cast(
str, value["contents"]))
# By here one of them should have worked!
if not entity or not checksum:
raise ValueError(
"class:File but missing checksum/location/content: %r" % value)
# Track filename and extension, this is generally useful only for
# secondaryFiles. Note that multiple uses of a file might thus record
# different names for the same entity, so we'll
# make/track a specialized entity by UUID
file_id = value.setdefault("@id", uuid.uuid4().urn)
# A specialized entity that has just these names
file_entity = self.document.entity(
file_id,
[(PROV_TYPE, WFPROV["Artifact"]), (PROV_TYPE, WF4EVER["File"])],
) # type: ProvEntity
if "basename" in value:
file_entity.add_attributes(
{CWLPROV["basename"]: value["basename"]})
if "nameroot" in value:
file_entity.add_attributes(
{CWLPROV["nameroot"]: value["nameroot"]})
if "nameext" in value:
file_entity.add_attributes({CWLPROV["nameext"]: value["nameext"]})
self.document.specializationOf(file_entity, entity)
# Check for secondaries
for sec in cast(MutableSequence[CWLObjectType],
value.get("secondaryFiles", [])):
# TODO: Record these in a specializationOf entity with UUID?
if sec["class"] == "File":
(sec_entity, _, _) = self.declare_file(sec)
elif sec["class"] == "Directory":
sec_entity = self.declare_directory(sec)
else:
raise ValueError(f"Got unexpected secondaryFiles value: {sec}")
# We don't know how/when/where the secondary file was generated,
# but CWL convention is a kind of summary/index derived
# from the original file. As its generally in a different format
# then prov:Quotation is not appropriate.
self.document.derivation(
sec_entity,
file_entity,
other_attributes={PROV["type"]: CWLPROV["SecondaryFile"]},
)
return file_entity, entity, checksum
def declare_directory(self, value: CWLObjectType) -> ProvEntity:
"""Register any nested files/directories."""
# FIXME: Calculate a hash-like identifier for directory
# so we get same value if it's the same filenames/hashes
# in a different location.
# For now, mint a new UUID to identify this directory, but
# attempt to keep it inside the value dictionary
dir_id = cast(str, value.setdefault("@id", uuid.uuid4().urn))
# New annotation file to keep the ORE Folder listing
ore_doc_fn = dir_id.replace("urn:uuid:", "directory-") + ".ttl"
dir_bundle = self.document.bundle(self.metadata_ns[ore_doc_fn])
coll = self.document.entity(
dir_id,
[
(PROV_TYPE, WFPROV["Artifact"]),
(PROV_TYPE, PROV["Collection"]),
(PROV_TYPE, PROV["Dictionary"]),
(PROV_TYPE, RO["Folder"]),
],
)
# ORE description of ro:Folder, saved separately
coll_b = dir_bundle.entity(
dir_id,
[(PROV_TYPE, RO["Folder"]), (PROV_TYPE, ORE["Aggregation"])],
)
self.document.mentionOf(dir_id + "#ore", dir_id, dir_bundle.identifier)
# dir_manifest = dir_bundle.entity(
# dir_bundle.identifier, {PROV["type"]: ORE["ResourceMap"],
# ORE["describes"]: coll_b.identifier})
coll_attribs = [(ORE["isDescribedBy"], dir_bundle.identifier)]
coll_b_attribs = [] # type: List[Tuple[Identifier, ProvEntity]]
# FIXME: .listing might not be populated yet - hopefully
# a later call to this method will sort that
is_empty = True
if "listing" not in value:
get_listing(self.fsaccess, value)
for entry in cast(MutableSequence[CWLObjectType],
value.get("listing", [])):
is_empty = False
# Declare child-artifacts
entity = self.declare_artefact(entry)
self.document.membership(coll, entity)
# Membership relation aka our ORE Proxy
m_id = uuid.uuid4().urn
m_entity = self.document.entity(m_id)
m_b = dir_bundle.entity(m_id)
# PROV-O style Dictionary
# https://www.w3.org/TR/prov-dictionary/#dictionary-ontological-definition
# ..as prov.py do not currently allow PROV-N extensions
# like hadDictionaryMember(..)
m_entity.add_asserted_type(PROV["KeyEntityPair"])
m_entity.add_attributes({
PROV["pairKey"]: entry["basename"],
PROV["pairEntity"]: entity,
})
# As well as a being a
# http://wf4ever.github.io/ro/2016-01-28/ro/#FolderEntry
m_b.add_asserted_type(RO["FolderEntry"])
m_b.add_asserted_type(ORE["Proxy"])
m_b.add_attributes({
RO["entryName"]: entry["basename"],
ORE["proxyIn"]: coll,
ORE["proxyFor"]: entity,
})
coll_attribs.append((PROV["hadDictionaryMember"], m_entity))
coll_b_attribs.append((ORE["aggregates"], m_b))
coll.add_attributes(coll_attribs)
coll_b.add_attributes(coll_b_attribs)
# Also Save ORE Folder as annotation metadata
ore_doc = ProvDocument()
ore_doc.add_namespace(ORE)
ore_doc.add_namespace(RO)
ore_doc.add_namespace(UUID)
ore_doc.add_bundle(dir_bundle)
ore_doc = ore_doc.flattened()
ore_doc_path = str(PurePosixPath(METADATA, ore_doc_fn))
with self.research_object.write_bag_file(
ore_doc_path) as provenance_file:
ore_doc.serialize(provenance_file,
format="rdf",
rdf_format="turtle")
self.research_object.add_annotation(dir_id, [ore_doc_fn],
ORE["isDescribedBy"].uri)
if is_empty:
# Empty directory
coll.add_asserted_type(PROV["EmptyCollection"])
coll.add_asserted_type(PROV["EmptyDictionary"])
self.research_object.add_uri(coll.identifier.uri)
return coll
def declare_string(self, value: str) -> Tuple[ProvEntity, str]:
"""Save as string in UTF-8."""
byte_s = BytesIO(str(value).encode(ENCODING))
data_file = self.research_object.add_data_file(byte_s,
content_type=TEXT_PLAIN)
checksum = PurePosixPath(data_file).name
# FIXME: Don't naively assume add_data_file uses hash in filename!
data_id = "data:%s" % PurePosixPath(data_file).stem
entity = self.document.entity(data_id, {
PROV_TYPE: WFPROV["Artifact"],
PROV_VALUE: str(value)
}) # type: ProvEntity
return entity, checksum
def declare_artefact(self, value: Optional[CWLOutputType]) -> ProvEntity:
"""Create data artefact entities for all file objects."""
if value is None:
# FIXME: If this can happen in CWL, we'll
# need a better way to represent this in PROV
return self.document.entity(CWLPROV["None"], {PROV_LABEL: "None"})
if isinstance(value, (bool, int, float)):
# Typically used in job documents for flags
# FIXME: Make consistent hash URIs for these
# that somehow include the type
# (so "1" != 1 != "1.0" != true)
entity = self.document.entity(uuid.uuid4().urn,
{PROV_VALUE: value})
self.research_object.add_uri(entity.identifier.uri)
return entity
if isinstance(value, (str, str)):
(entity, _) = self.declare_string(value)
return entity
if isinstance(value, bytes):
# If we got here then we must be in Python 3
byte_s = BytesIO(value)
data_file = self.research_object.add_data_file(byte_s)
# FIXME: Don't naively assume add_data_file uses hash in filename!
data_id = "data:%s" % PurePosixPath(data_file).stem
return self.document.entity(
data_id,
{
PROV_TYPE: WFPROV["Artifact"],
PROV_VALUE: str(value)
},
)
if isinstance(value, MutableMapping):
if "@id" in value:
# Already processed this value, but it might not be in this PROV
entities = self.document.get_record(value["@id"])
if entities:
return entities[0]
# else, unknown in PROV, re-add below as if it's fresh
# Base case - we found a File we need to update
if value.get("class") == "File":
(entity, _, _) = self.declare_file(value)
value["@id"] = entity.identifier.uri
return entity
if value.get("class") == "Directory":
entity = self.declare_directory(value)
value["@id"] = entity.identifier.uri
return entity
coll_id = value.setdefault("@id", uuid.uuid4().urn)
# some other kind of dictionary?
# TODO: also Save as JSON
coll = self.document.entity(
coll_id,
[
(PROV_TYPE, WFPROV["Artifact"]),
(PROV_TYPE, PROV["Collection"]),
(PROV_TYPE, PROV["Dictionary"]),
],
)
if value.get("class"):
_logger.warning("Unknown data class %s.", value["class"])
# FIXME: The class might be "http://example.com/somethingelse"
coll.add_asserted_type(CWLPROV[value["class"]])
# Let's iterate and recurse
coll_attribs = [] # type: List[Tuple[Identifier, ProvEntity]]
for (key, val) in value.items():
v_ent = self.declare_artefact(val)
self.document.membership(coll, v_ent)
m_entity = self.document.entity(uuid.uuid4().urn)
# Note: only support PROV-O style dictionary
# https://www.w3.org/TR/prov-dictionary/#dictionary-ontological-definition
# as prov.py do not easily allow PROV-N extensions
m_entity.add_asserted_type(PROV["KeyEntityPair"])
m_entity.add_attributes({
PROV["pairKey"]: str(key),
PROV["pairEntity"]: v_ent
})
coll_attribs.append((PROV["hadDictionaryMember"], m_entity))
coll.add_attributes(coll_attribs)
self.research_object.add_uri(coll.identifier.uri)
return coll
# some other kind of Collection?
# TODO: also save as JSON
try:
members = []
for each_input_obj in iter(value):
# Recurse and register any nested objects
e = self.declare_artefact(each_input_obj)
members.append(e)
# If we reached this, then we were allowed to iterate
coll = self.document.entity(
uuid.uuid4().urn,
[
(PROV_TYPE, WFPROV["Artifact"]),
(PROV_TYPE, PROV["Collection"]),
],
)
if not members:
coll.add_asserted_type(PROV["EmptyCollection"])
else:
for member in members:
# FIXME: This won't preserve order, for that
# we would need to use PROV.Dictionary
# with numeric keys
self.document.membership(coll, member)
self.research_object.add_uri(coll.identifier.uri)
# FIXME: list value does not support adding "@id"
return coll
except TypeError:
_logger.warning("Unrecognized type %s of %r", type(value), value)
# Let's just fall back to Python repr()
entity = self.document.entity(uuid.uuid4().urn,
{PROV_LABEL: repr(value)})
self.research_object.add_uri(entity.identifier.uri)
return entity
def used_artefacts(
self,
job_order: Union[CWLObjectType, List[CWLObjectType]],
process_run_id: str,
name: Optional[str] = None,
) -> None:
"""Add used() for each data artefact."""
if isinstance(job_order, list):
for entry in job_order:
self.used_artefacts(entry, process_run_id, name)
else:
# FIXME: Use workflow name in packed.cwl, "main" is wrong for nested workflows
base = "main"
if name is not None:
base += "/" + name
for key, value in job_order.items():
prov_role = self.wf_ns[f"{base}/{key}"]
try:
entity = self.declare_artefact(value)
self.document.used(
process_run_id,
entity,
datetime.datetime.now(),
None,
{"prov:role": prov_role},
)
except OSError:
pass
def generate_output_prov(
self,
final_output: Union[CWLObjectType, MutableSequence[CWLObjectType],
None],
process_run_id: Optional[str],
name: Optional[str],
) -> None:
"""Call wasGeneratedBy() for each output,copy the files into the RO."""
if isinstance(final_output, MutableSequence):
for entry in final_output:
self.generate_output_prov(entry, process_run_id, name)
elif final_output is not None:
# Timestamp should be created at the earliest
timestamp = datetime.datetime.now()
# For each output, find/register the corresponding
# entity (UUID) and document it as generated in
# a role corresponding to the output
for output, value in final_output.items():
entity = self.declare_artefact(value)
if name is not None:
name = urllib.parse.quote(str(name), safe=":/,#")
# FIXME: Probably not "main" in nested workflows
role = self.wf_ns[f"main/{name}/{output}"]
else:
role = self.wf_ns["main/%s" % output]
if not process_run_id:
process_run_id = self.workflow_run_uri
self.document.wasGeneratedBy(entity, process_run_id, timestamp,
None, {"prov:role": role})
def prospective_prov(self, job: JobsType) -> None:
"""Create prospective prov recording as wfdesc prov:Plan."""
if not isinstance(job, WorkflowJob):
# direct command line tool execution
self.document.entity(
"wf:main",
{
PROV_TYPE: WFDESC["Process"],
"prov:type": PROV["Plan"],
"prov:label": "Prospective provenance",
},
)
return
self.document.entity(
"wf:main",
{
PROV_TYPE: WFDESC["Workflow"],
"prov:type": PROV["Plan"],
"prov:label": "Prospective provenance",
},
)
for step in job.steps:
stepnametemp = "wf:main/" + str(step.name)[5:]
stepname = urllib.parse.quote(stepnametemp, safe=":/,#")
provstep = self.document.entity(
stepname,
{
PROV_TYPE: WFDESC["Process"],
"prov:type": PROV["Plan"]
},
)
self.document.entity(
"wf:main",
{
"wfdesc:hasSubProcess": provstep,
"prov:label": "Prospective provenance",
},
)
# TODO: Declare roles/parameters as well
def activity_has_provenance(self, activity, prov_ids):
# type: (str, List[Identifier]) -> None
"""Add http://www.w3.org/TR/prov-aq/ relations to nested PROV files."""
# NOTE: The below will only work if the corresponding metadata/provenance arcp URI
# is a pre-registered namespace in the PROV Document
attribs = [(PROV["has_provenance"], prov_id) for prov_id in prov_ids]
self.document.activity(activity, other_attributes=attribs)
# Tip: we can't use https://www.w3.org/TR/prov-links/#term-mention
# as prov:mentionOf() is only for entities, not activities
uris = [i.uri for i in prov_ids]
self.research_object.add_annotation(activity, uris,
PROV["has_provenance"].uri)
def finalize_prov_profile(self, name):
# type: (Optional[str]) -> List[Identifier]
"""Transfer the provenance related files to the RO."""
# NOTE: Relative posix path
if name is None:
# main workflow, fixed filenames
filename = "primary.cwlprov"
else:
# ASCII-friendly filename, avoiding % as we don't want %2520 in manifest.json
wf_name = urllib.parse.quote(str(name), safe="").replace("%", "_")
# Note that the above could cause overlaps for similarly named
# workflows, but that's OK as we'll also include run uuid
# which also covers thhe case of this step being run in
# multiple places or iterations
filename = f"{wf_name}.{self.workflow_run_uuid}.cwlprov"
basename = str(PurePosixPath(PROVENANCE) / filename)
# TODO: Also support other profiles than CWLProv, e.g. ProvOne
# list of prov identifiers of provenance files
prov_ids = []
# https://www.w3.org/TR/prov-xml/
with self.research_object.write_bag_file(basename +
".xml") as provenance_file:
self.document.serialize(provenance_file, format="xml", indent=4)
prov_ids.append(self.provenance_ns[filename + ".xml"])
# https://www.w3.org/TR/prov-n/
with self.research_object.write_bag_file(basename +
".provn") as provenance_file:
self.document.serialize(provenance_file, format="provn", indent=2)
prov_ids.append(self.provenance_ns[filename + ".provn"])
# https://www.w3.org/Submission/prov-json/
with self.research_object.write_bag_file(basename +
".json") as provenance_file:
self.document.serialize(provenance_file, format="json", indent=2)
prov_ids.append(self.provenance_ns[filename + ".json"])
# "rdf" aka https://www.w3.org/TR/prov-o/
# which can be serialized to ttl/nt/jsonld (and more!)
# https://www.w3.org/TR/turtle/
with self.research_object.write_bag_file(basename +
".ttl") as provenance_file:
self.document.serialize(provenance_file,
format="rdf",
rdf_format="turtle")
prov_ids.append(self.provenance_ns[filename + ".ttl"])
# https://www.w3.org/TR/n-triples/
with self.research_object.write_bag_file(basename +
".nt") as provenance_file:
self.document.serialize(provenance_file,
format="rdf",
rdf_format="ntriples")
prov_ids.append(self.provenance_ns[filename + ".nt"])
# https://www.w3.org/TR/json-ld/
# TODO: Use a nice JSON-LD context
# see also https://eprints.soton.ac.uk/395985/
# 404 Not Found on https://provenance.ecs.soton.ac.uk/prov.jsonld :(
with self.research_object.write_bag_file(basename +
".jsonld") as provenance_file:
self.document.serialize(provenance_file,
format="rdf",
rdf_format="json-ld")
prov_ids.append(self.provenance_ns[filename + ".jsonld"])
_logger.debug("[provenance] added provenance: %s", prov_ids)
return prov_ids
class NIDMExporter():
"""
Generic class to parse a result directory to extract the pieces of
information to be stored in NIDM-Results and to generate a NIDM-Results
export.
"""
def __init__(self, version, out_dir, zipped=True):
out_dirname = os.path.basename(out_dir)
out_path = os.path.dirname(out_dir)
# Create output path from output name
self.zipped = zipped
if not self.zipped:
out_dirname = out_dirname+".nidm"
else:
out_dirname = out_dirname+".nidm.zip"
out_dir = os.path.join(out_path, out_dirname)
# Quit if output path already exists and user doesn't want to overwrite
# it
if os.path.exists(out_dir):
msg = out_dir+" already exists, overwrite?"
if not input("%s (y/N) " % msg).lower() == 'y':
quit("Bye.")
if os.path.isdir(out_dir):
shutil.rmtree(out_dir)
else:
os.remove(out_dir)
self.out_dir = out_dir
if version == "dev":
self.version = {'major': 10000, 'minor': 0, 'revision': 0,
'num': version}
else:
major, minor, revision = version.split(".")
if "-rc" in revision:
revision, rc = revision.split("-rc")
else:
rc = -1
self.version = {'major': int(major), 'minor': int(minor),
'revision': int(revision), 'rc': int(rc),
'num': version}
# Initialise prov document
self.doc = ProvDocument()
self._add_namespaces()
# A temp directory that will contain the exported data
self.export_dir = tempfile.mkdtemp(prefix="nidm-", dir=out_path)
self.prepend_path = ''
def parse(self):
"""
Parse a result directory to extract the pieces information to be
stored in NIDM-Results.
"""
try:
# Methods: find_software, find_model_fitting, find_contrasts and
# find_inferences should be defined in the children classes and
# return a list of NIDM Objects as specified in the objects module
# Object of type Software describing the neuroimaging software
# package used for the analysis
self.software = self._find_software()
# List of objects of type ModelFitting describing the
# model fitting step in NIDM-Results (main activity: Model
# Parameters Estimation)
self.model_fittings = self._find_model_fitting()
# Dictionary of (key, value) pairs where where key is a tuple
# containing the identifier of a ModelParametersEstimation object
# and a tuple of identifiers of ParameterEstimateMap objects and
# value is an object of type Contrast describing the contrast
# estimation step in NIDM-Results (main activity: Contrast
# Estimation)
self.contrasts = self._find_contrasts()
# Inference activity and entities
# Dictionary of (key, value) pairs where key is the identifier of a
# ContrastEstimation object and value is an object of type
# Inference describing the inference step in NIDM-Results (main
# activity: Inference)
self.inferences = self._find_inferences()
except Exception:
self.cleanup()
raise
def cleanup(self):
if os.path.isdir(self.export_dir):
shutil.rmtree(self.export_dir)
def add_object(self, nidm_object, export_file=True):
"""
Add a NIDMObject to a NIDM-Results export.
"""
if not export_file:
export_dir = None
else:
export_dir = self.export_dir
if not isinstance(nidm_object, NIDMFile):
nidm_object.export(self.version, export_dir)
else:
nidm_object.export(self.version, export_dir, self.prepend_path)
# ProvDocument: add object to the bundle
if nidm_object.prov_type == PROV['Activity']:
self.bundle.activity(nidm_object.id,
other_attributes=nidm_object.attributes)
elif nidm_object.prov_type == PROV['Entity']:
self.bundle.entity(nidm_object.id,
other_attributes=nidm_object.attributes)
elif nidm_object.prov_type == PROV['Agent']:
self.bundle.agent(nidm_object.id,
other_attributes=nidm_object.attributes)
# self.bundle.update(nidm_object.p)
def export(self):
"""
Generate a NIDM-Results export.
"""
try:
if not os.path.isdir(self.export_dir):
os.mkdir(self.export_dir)
# Initialise main bundle
self._create_bundle(self.version)
self.add_object(self.software)
# Add model fitting steps
if not isinstance(self.model_fittings, list):
self.model_fittings = list(self.model_fittings.values())
for model_fitting in self.model_fittings:
# Design Matrix
# model_fitting.activity.used(model_fitting.design_matrix)
self.bundle.used(model_fitting.activity.id,
model_fitting.design_matrix.id)
self.add_object(model_fitting.design_matrix)
# *** Export visualisation of the design matrix
self.add_object(model_fitting.design_matrix.image)
if model_fitting.design_matrix.image.file is not None:
self.add_object(model_fitting.design_matrix.image.file)
if model_fitting.design_matrix.hrf_models is not None:
# drift model
self.add_object(model_fitting.design_matrix.drift_model)
if self.version['major'] > 1 or \
(self.version['major'] == 1 and
self.version['minor'] >= 3):
# Machine
# model_fitting.data.wasAttributedTo(model_fitting.machine)
self.bundle.wasAttributedTo(model_fitting.data.id,
model_fitting.machine.id)
self.add_object(model_fitting.machine)
# Imaged subject or group(s)
for sub in model_fitting.subjects:
self.add_object(sub)
# model_fitting.data.wasAttributedTo(sub)
self.bundle.wasAttributedTo(model_fitting.data.id,
sub.id)
# Data
# model_fitting.activity.used(model_fitting.data)
self.bundle.used(model_fitting.activity.id,
model_fitting.data.id)
self.add_object(model_fitting.data)
# Error Model
# model_fitting.activity.used(model_fitting.error_model)
self.bundle.used(model_fitting.activity.id,
model_fitting.error_model.id)
self.add_object(model_fitting.error_model)
# Parameter Estimate Maps
for param_estimate in model_fitting.param_estimates:
# param_estimate.wasGeneratedBy(model_fitting.activity)
self.bundle.wasGeneratedBy(param_estimate.id,
model_fitting.activity.id)
self.add_object(param_estimate)
self.add_object(param_estimate.coord_space)
self.add_object(param_estimate.file)
if param_estimate.derfrom is not None:
self.bundle.wasDerivedFrom(param_estimate.id,
param_estimate.derfrom.id)
self.add_object(param_estimate.derfrom)
self.add_object(param_estimate.derfrom.file,
export_file=False)
# Residual Mean Squares Map
# model_fitting.rms_map.wasGeneratedBy(model_fitting.activity)
self.add_object(model_fitting.rms_map)
self.bundle.wasGeneratedBy(model_fitting.rms_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.rms_map.coord_space)
self.add_object(model_fitting.rms_map.file)
if model_fitting.rms_map.derfrom is not None:
self.bundle.wasDerivedFrom(
model_fitting.rms_map.id,
model_fitting.rms_map.derfrom.id)
self.add_object(model_fitting.rms_map.derfrom)
self.add_object(model_fitting.rms_map.derfrom.file,
export_file=False)
# Resels per Voxel Map
if model_fitting.rpv_map is not None:
self.add_object(model_fitting.rpv_map)
self.bundle.wasGeneratedBy(model_fitting.rpv_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.rpv_map.coord_space)
self.add_object(model_fitting.rpv_map.file)
if model_fitting.rpv_map.inf_id is not None:
self.bundle.used(model_fitting.rpv_map.inf_id,
model_fitting.rpv_map.id)
if model_fitting.rpv_map.derfrom is not None:
self.bundle.wasDerivedFrom(
model_fitting.rpv_map.id,
model_fitting.rpv_map.derfrom.id)
self.add_object(model_fitting.rpv_map.derfrom)
self.add_object(model_fitting.rpv_map.derfrom.file,
export_file=False)
# Mask
# model_fitting.mask_map.wasGeneratedBy(model_fitting.activity)
self.bundle.wasGeneratedBy(model_fitting.mask_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.mask_map)
if model_fitting.mask_map.derfrom is not None:
self.bundle.wasDerivedFrom(
model_fitting.mask_map.id,
model_fitting.mask_map.derfrom.id)
self.add_object(model_fitting.mask_map.derfrom)
self.add_object(model_fitting.mask_map.derfrom.file,
export_file=False)
# Create coordinate space export
self.add_object(model_fitting.mask_map.coord_space)
# Create "Mask map" entity
self.add_object(model_fitting.mask_map.file)
# Grand Mean map
# model_fitting.grand_mean_map.wasGeneratedBy(model_fitting.activity)
self.bundle.wasGeneratedBy(model_fitting.grand_mean_map.id,
model_fitting.activity.id)
self.add_object(model_fitting.grand_mean_map)
# Coordinate space entity
self.add_object(model_fitting.grand_mean_map.coord_space)
# Grand Mean Map entity
self.add_object(model_fitting.grand_mean_map.file)
# Model Parameters Estimation activity
self.add_object(model_fitting.activity)
self.bundle.wasAssociatedWith(model_fitting.activity.id,
self.software.id)
# model_fitting.activity.wasAssociatedWith(self.software)
# self.add_object(model_fitting)
# Add contrast estimation steps
analysis_masks = dict()
for (model_fitting_id, pe_ids), contrasts in list(
self.contrasts.items()):
for contrast in contrasts:
model_fitting = self._get_model_fitting(model_fitting_id)
# for contrast in contrasts:
# contrast.estimation.used(model_fitting.rms_map)
self.bundle.used(contrast.estimation.id,
model_fitting.rms_map.id)
# contrast.estimation.used(model_fitting.mask_map)
self.bundle.used(contrast.estimation.id,
model_fitting.mask_map.id)
analysis_masks[contrast.estimation.id] = \
model_fitting.mask_map.id
self.bundle.used(contrast.estimation.id,
contrast.weights.id)
self.bundle.used(contrast.estimation.id,
model_fitting.design_matrix.id)
# contrast.estimation.wasAssociatedWith(self.software)
self.bundle.wasAssociatedWith(contrast.estimation.id,
self.software.id)
for pe_id in pe_ids:
# contrast.estimation.used(pe_id)
self.bundle.used(contrast.estimation.id, pe_id)
# Create estimation activity
self.add_object(contrast.estimation)
# Create contrast weights
self.add_object(contrast.weights)
if contrast.contrast_map is not None:
# Create contrast Map
# contrast.contrast_map.wasGeneratedBy(contrast.estimation)
self.bundle.wasGeneratedBy(contrast.contrast_map.id,
contrast.estimation.id)
self.add_object(contrast.contrast_map)
self.add_object(contrast.contrast_map.coord_space)
# Copy contrast map in export directory
self.add_object(contrast.contrast_map.file)
if contrast.contrast_map.derfrom is not None:
self.bundle.wasDerivedFrom(
contrast.contrast_map.id,
contrast.contrast_map.derfrom.id)
self.add_object(contrast.contrast_map.derfrom)
self.add_object(contrast.contrast_map.derfrom.file,
export_file=False)
# Create Std Err. Map (T-tests) or Explained Mean Sq. Map
# (F-tests)
# contrast.stderr_or_expl_mean_sq_map.wasGeneratedBy
# (contrast.estimation)
stderr_explmeansq_map = (
contrast.stderr_or_expl_mean_sq_map)
self.bundle.wasGeneratedBy(
stderr_explmeansq_map.id,
contrast.estimation.id)
self.add_object(stderr_explmeansq_map)
self.add_object(
stderr_explmeansq_map.coord_space)
if isinstance(stderr_explmeansq_map,
ContrastStdErrMap) and \
stderr_explmeansq_map.contrast_var:
self.add_object(
stderr_explmeansq_map.contrast_var)
if stderr_explmeansq_map.var_coord_space:
self.add_object(
stderr_explmeansq_map.var_coord_space)
if stderr_explmeansq_map.contrast_var.coord_space:
self.add_object(
stderr_explmeansq_map.contrast_var.coord_space)
self.add_object(
stderr_explmeansq_map.contrast_var.file,
export_file=False)
self.bundle.wasDerivedFrom(
stderr_explmeansq_map.id,
stderr_explmeansq_map.contrast_var.id)
self.add_object(stderr_explmeansq_map.file)
# Create Statistic Map
# contrast.stat_map.wasGeneratedBy(contrast.estimation)
self.bundle.wasGeneratedBy(contrast.stat_map.id,
contrast.estimation.id)
self.add_object(contrast.stat_map)
self.add_object(contrast.stat_map.coord_space)
# Copy Statistical map in export directory
self.add_object(contrast.stat_map.file)
if contrast.stat_map.derfrom is not None:
self.bundle.wasDerivedFrom(
contrast.stat_map.id,
contrast.stat_map.derfrom.id)
self.add_object(contrast.stat_map.derfrom)
self.add_object(contrast.stat_map.derfrom.file,
export_file=False)
# Create Z Statistic Map
if contrast.z_stat_map:
# contrast.z_stat_map.wasGeneratedBy(contrast.estimation)
self.bundle.wasGeneratedBy(contrast.z_stat_map.id,
contrast.estimation.id)
self.add_object(contrast.z_stat_map)
self.add_object(contrast.z_stat_map.coord_space)
# Copy Statistical map in export directory
self.add_object(contrast.z_stat_map.file)
# self.add_object(contrast)
# Add inference steps
for contrast_id, inferences in list(self.inferences.items()):
contrast = self._get_contrast(contrast_id)
for inference in inferences:
if contrast.z_stat_map:
used_id = contrast.z_stat_map.id
else:
used_id = contrast.stat_map.id
# inference.inference_act.used(used_id)
self.bundle.used(inference.inference_act.id, used_id)
# inference.inference_act.wasAssociatedWith(self.software)
self.bundle.wasAssociatedWith(inference.inference_act.id,
self.software.id)
# self.add_object(inference)
# Excursion set
# inference.excursion_set.wasGeneratedBy(inference.inference_act)
self.bundle.wasGeneratedBy(inference.excursion_set.id,
inference.inference_act.id)
self.add_object(inference.excursion_set)
self.add_object(inference.excursion_set.coord_space)
if inference.excursion_set.visu is not None:
self.add_object(inference.excursion_set.visu)
if inference.excursion_set.visu.file is not None:
self.add_object(inference.excursion_set.visu.file)
# Copy "Excursion set map" file in export directory
self.add_object(inference.excursion_set.file)
if inference.excursion_set.clust_map is not None:
self.add_object(inference.excursion_set.clust_map)
self.add_object(inference.excursion_set.clust_map.file)
self.add_object(
inference.excursion_set.clust_map.coord_space)
if inference.excursion_set.mip is not None:
self.add_object(inference.excursion_set.mip)
self.add_object(inference.excursion_set.mip.file)
# Height threshold
if inference.height_thresh.equiv_thresh is not None:
for equiv in inference.height_thresh.equiv_thresh:
self.add_object(equiv)
self.add_object(inference.height_thresh)
# Extent threshold
if inference.extent_thresh.equiv_thresh is not None:
for equiv in inference.extent_thresh.equiv_thresh:
self.add_object(equiv)
self.add_object(inference.extent_thresh)
# Display Mask (potentially more than 1)
if inference.disp_mask:
for mask in inference.disp_mask:
# inference.inference_act.used(mask)
self.bundle.used(inference.inference_act.id,
mask.id)
self.add_object(mask)
# Create coordinate space entity
self.add_object(mask.coord_space)
# Create "Display Mask Map" entity
self.add_object(mask.file)
if mask.derfrom is not None:
self.bundle.wasDerivedFrom(mask.id,
mask.derfrom.id)
self.add_object(mask.derfrom)
self.add_object(mask.derfrom.file,
export_file=False)
# Search Space
self.bundle.wasGeneratedBy(inference.search_space.id,
inference.inference_act.id)
# inference.search_space.wasGeneratedBy(inference.inference_act)
self.add_object(inference.search_space)
self.add_object(inference.search_space.coord_space)
# Copy "Mask map" in export directory
self.add_object(inference.search_space.file)
# Peak Definition
if inference.peak_criteria:
# inference.inference_act.used(inference.peak_criteria)
self.bundle.used(inference.inference_act.id,
inference.peak_criteria.id)
self.add_object(inference.peak_criteria)
# Cluster Definition
if inference.cluster_criteria:
# inference.inference_act.used(inference.cluster_criteria)
self.bundle.used(inference.inference_act.id,
inference.cluster_criteria.id)
self.add_object(inference.cluster_criteria)
if inference.clusters:
# Clusters and peaks
for cluster in inference.clusters:
# cluster.wasDerivedFrom(inference.excursion_set)
self.bundle.wasDerivedFrom(
cluster.id, inference.excursion_set.id)
self.add_object(cluster)
for peak in cluster.peaks:
self.bundle.wasDerivedFrom(peak.id, cluster.id)
self.add_object(peak)
self.add_object(peak.coordinate)
if cluster.cog is not None:
self.bundle.wasDerivedFrom(cluster.cog.id,
cluster.id)
self.add_object(cluster.cog)
self.add_object(cluster.cog.coordinate)
# Inference activity
# inference.inference_act.wasAssociatedWith(inference.software_id)
# inference.inference_act.used(inference.height_thresh)
self.bundle.used(inference.inference_act.id,
inference.height_thresh.id)
# inference.inference_act.used(inference.extent_thresh)
self.bundle.used(inference.inference_act.id,
inference.extent_thresh.id)
self.bundle.used(inference.inference_act.id,
analysis_masks[contrast.estimation.id])
self.add_object(inference.inference_act)
# Write-out prov file
self.save_prov_to_files()
return self.out_dir
except Exception:
self.cleanup()
raise
def _get_model_fitting(self, mf_id):
"""
Retreive model fitting with identifier 'mf_id' from the list of model
fitting objects stored in self.model_fitting
"""
for model_fitting in self.model_fittings:
if model_fitting.activity.id == mf_id:
return model_fitting
raise Exception("Model fitting activity with id: " + str(mf_id) +
" not found.")
def _get_contrast(self, con_id):
"""
Retreive contrast with identifier 'con_id' from the list of contrast
objects stored in self.contrasts
"""
for contrasts in list(self.contrasts.values()):
for contrast in contrasts:
if contrast.estimation.id == con_id:
return contrast
raise Exception("Contrast activity with id: " + str(con_id) +
" not found.")
def _add_namespaces(self):
"""
Add namespaces to NIDM document.
"""
self.doc.add_namespace(NIDM)
self.doc.add_namespace(NIIRI)
self.doc.add_namespace(CRYPTO)
self.doc.add_namespace(DCT)
self.doc.add_namespace(DC)
self.doc.add_namespace(NFO)
self.doc.add_namespace(OBO)
self.doc.add_namespace(SCR)
self.doc.add_namespace(NIF)
def _create_bundle(self, version):
"""
Initialise NIDM-Results bundle.
"""
# *** Bundle entity
if not hasattr(self, 'bundle_ent'):
self.bundle_ent = NIDMResultsBundle(nidm_version=version['num'])
self.bundle = ProvBundle(identifier=self.bundle_ent.id)
self.bundle_ent.export(self.version, self.export_dir)
# # provn export
# self.bundle = ProvBundle(identifier=bundle_id)
self.doc.entity(self.bundle_ent.id,
other_attributes=self.bundle_ent.attributes)
# *** NIDM-Results Export Activity
if version['num'] not in ["1.0.0", "1.1.0"]:
if not hasattr(self, 'export_act'):
self.export_act = NIDMResultsExport()
self.export_act.export(self.version, self.export_dir)
# self.doc.update(self.export_act.p)
self.doc.activity(self.export_act.id,
other_attributes=self.export_act.attributes)
# *** bundle was Generated by NIDM-Results Export Activity
if not hasattr(self, 'export_time'):
self.export_time = str(datetime.datetime.now().time())
if version['num'] in ["1.0.0", "1.1.0"]:
self.doc.wasGeneratedBy(entity=self.bundle_ent.id,
time=self.export_time)
else:
# provn
self.doc.wasGeneratedBy(
entity=self.bundle_ent.id, activity=self.export_act.id,
time=self.export_time)
# *** NIDM-Results Exporter (Software Agent)
if version['num'] not in ["1.0.0", "1.1.0"]:
if not hasattr(self, 'exporter'):
self.exporter = self._get_exporter()
self.exporter.export(self.version, self.export_dir)
# self.doc.update(self.exporter.p)
self.doc.agent(self.exporter.id,
other_attributes=self.exporter.attributes)
self.doc.wasAssociatedWith(self.export_act.id, self.exporter.id)
def _get_model_parameters_estimations(self, error_model):
"""
Infer model estimation method from the 'error_model'. Return an object
of type ModelParametersEstimation.
"""
if error_model.dependance == NIDM_INDEPEDENT_ERROR:
if error_model.variance_homo:
estimation_method = STATO_OLS
else:
estimation_method = STATO_WLS
else:
estimation_method = STATO_GLS
mpe = ModelParametersEstimation(estimation_method, self.software.id)
return mpe
def use_prefixes(self, ttl):
prefix_file = os.path.join(os.path.dirname(__file__), 'prefixes.csv')
context = dict()
with open(prefix_file, encoding="ascii") as csvfile:
reader = csv.reader(csvfile)
next(reader, None) # skip the headers
for alphanum_id, prefix, uri in reader:
if alphanum_id in ttl:
context[prefix] = uri
ttl = "@prefix " + prefix + ": <" + uri + "> .\n" + ttl
ttl = ttl.replace(alphanum_id, prefix + ":")
if uri in ttl:
ttl = ttl.replace(alphanum_id, prefix + ":")
elif uri in ttl:
context[prefix] = uri
ttl = "@prefix " + prefix + ": <" + uri + "> .\n" + ttl
ttl = ttl.replace(alphanum_id, prefix + ":")
return (ttl, context)
def save_prov_to_files(self, showattributes=False):
"""
Write-out provn serialisation to nidm.provn.
"""
self.doc.add_bundle(self.bundle)
# provn_file = os.path.join(self.export_dir, 'nidm.provn')
# provn_fid = open(provn_file, 'w')
# # FIXME None
# # provn_fid.write(self.doc.get_provn(4).replace("None", "-"))
# provn_fid.close()
ttl_file = os.path.join(self.export_dir, 'nidm.ttl')
ttl_txt = self.doc.serialize(format='rdf', rdf_format='turtle')
ttl_txt, json_context = self.use_prefixes(ttl_txt)
# Add namespaces to json-ld context
for namespace in self.doc._namespaces.get_registered_namespaces():
json_context[namespace._prefix] = namespace._uri
for namespace in \
list(self.doc._namespaces._default_namespaces.values()):
json_context[namespace._prefix] = namespace._uri
json_context["xsd"] = "http://www.w3.org/2000/01/rdf-schema#"
# Work-around to issue with INF value in rdflib (reported in
# https://github.com/RDFLib/rdflib/pull/655)
ttl_txt = ttl_txt.replace(' inf ', ' "INF"^^xsd:float ')
with open(ttl_file, 'w') as ttl_fid:
ttl_fid.write(ttl_txt)
# print(json_context)
jsonld_file = os.path.join(self.export_dir, 'nidm.json')
jsonld_txt = self.doc.serialize(format='rdf', rdf_format='json-ld',
context=json_context)
with open(jsonld_file, 'w') as jsonld_fid:
jsonld_fid.write(jsonld_txt)
# provjsonld_file = os.path.join(self.export_dir, 'nidm.provjsonld')
# provjsonld_txt = self.doc.serialize(format='jsonld')
# with open(provjsonld_file, 'w') as provjsonld_fid:
# provjsonld_fid.write(provjsonld_txt)
# provn_file = os.path.join(self.export_dir, 'nidm.provn')
# provn_txt = self.doc.serialize(format='provn')
# with open(provn_file, 'w') as provn_fid:
# provn_fid.write(provn_txt)
# Post-processing
if not self.zipped:
# Just rename temp directory to output_path
os.rename(self.export_dir, self.out_dir)
else:
# Create a zip file that contains the content of the temp directory
os.chdir(self.export_dir)
zf = zipfile.ZipFile(os.path.join("..", self.out_dir), mode='w')
try:
for root, dirnames, filenames in os.walk("."):
for filename in filenames:
zf.write(os.path.join(filename))
finally:
zf.close()
# Need to move up before deleting the folder
os.chdir("..")
shutil.rmtree(os.path.join("..", self.export_dir))
