示例#1
0
def main():
    """Initializes main script from command-line call to generate single-subject or multi-subject workflow(s)"""
    import os
    import gc
    import sys
    import json
    import ast
    import yaml
    import itertools
    from types import SimpleNamespace
    from pathlib import Path
    import pkg_resources
    from pynets.core.utils import flatten
    from pynets.cli.pynets_run import build_workflow
    from multiprocessing import set_start_method, Process, Manager
    try:
        import pynets
    except ImportError:
        print('PyNets not installed! Ensure that you are referencing the correct site-packages and using Python3.6+')

    if len(sys.argv) < 1:
        print("\nMissing command-line inputs! See help options with the -h flag.\n")
        sys.exit()

    print('Obtaining Derivatives Layout...')

    modalities = ['func', 'dwi']

    bids_args = get_bids_parser().parse_args()
    participant_label = bids_args.participant_label
    session_label = bids_args.session_label
    modality = bids_args.modality
    bids_config = bids_args.config
    analysis_level = bids_args.analysis_level
    clean = bids_args.clean

    if analysis_level == 'group' and participant_label is not None:
        raise ValueError('Error: You have indicated a group analysis level run, but specified a participant label!')

    if analysis_level == 'participant' and participant_label is None:
        raise ValueError('Error: You have indicated a participant analysis level run, but not specified a participant '
                         'label!')

    if bids_config:
        with open(bids_config, 'r') as stream:
            arg_dict = json.load(stream)
    else:
        with open(pkg_resources.resource_filename("pynets", "config/bids_config_test.json"), 'r') as stream:
            arg_dict = json.load(stream)
        stream.close()

    # Available functional and structural connectivity models
    with open(pkg_resources.resource_filename("pynets", "runconfig.yaml"), 'r') as stream:
        hardcoded_params = yaml.load(stream)
        try:
            func_models = hardcoded_params['available_models']['func_models']
        except KeyError:
            print('ERROR: available functional models not successfully extracted from runconfig.yaml')
            sys.exit()
        try:
            struct_models = hardcoded_params['available_models']['struct_models']
        except KeyError:
            print('ERROR: available structural models not successfully extracted from runconfig.yaml')
            sys.exit()

        space = hardcoded_params['bids_defaults']['space'][0]
        func_desc = hardcoded_params['bids_defaults']['desc'][0]
    stream.close()

    # S3
    # Primary inputs

    s3 = bids_args.bids_dir.startswith("s3://")

    if not s3:
        bids_dir = bids_args.bids_dir

    # secondary inputs
    sec_s3_objs = []
    if isinstance(bids_args.ua, list):
        for i in bids_args.ua:
            if i.startswith("s3://"):
                print('Downloading user atlas: ', i, ' from S3...')
                sec_s3_objs.append(i)
    if isinstance(bids_args.cm, list):
        for i in bids_args.cm:
            if i.startswith("s3://"):
                print('Downloading clustering mask: ', i, ' from S3...')
                sec_s3_objs.append(i)
    if isinstance(bids_args.roi, list):
        for i in bids_args.roi:
            if i.startswith("s3://"):
                print('Downloading ROI mask: ', i, ' from S3...')
                sec_s3_objs.append(i)
    if isinstance(bids_args.way, list):
        for i in bids_args.way:
            if i.startswith("s3://"):
                print('Downloading tractography waymask: ', i, ' from S3...')
                sec_s3_objs.append(i)

    if bids_args.ref:
        if bids_args.ref.startswith("s3://"):
            print('Downloading atlas labeling reference file: ', bids_args.ref, ' from S3...')
            sec_s3_objs.append(bids_args.ref)

    if s3 or len(sec_s3_objs) > 0:
        from boto3.session import Session
        from pynets.core import cloud_utils
        from pynets.core.utils import as_directory

        home = os.path.expanduser("~")
        creds = bool(cloud_utils.get_credentials())

        if s3:
            buck, remo = cloud_utils.parse_path(bids_args.bids_dir)
            os.makedirs(f"{home}/.pynets", exist_ok=True)
            os.makedirs(f"{home}/.pynets/input", exist_ok=True)
            os.makedirs(f"{home}/.pynets/output", exist_ok=True)
            bids_dir = as_directory(f"{home}/.pynets/input", remove=False)
            if (not creds) and bids_args.push_location:
                raise AttributeError("""No AWS credentials found, but `--push_location` flag called. 
                Pushing will most likely fail.""")
            else:
                output_dir = as_directory(f"{home}/.pynets/output", remove=False)

            # Get S3 input data if needed
            if analysis_level == 'participant':
                for partic, ses in list(itertools.product(participant_label, session_label)):
                    if ses is not None:
                        info = "sub-" + partic + '/ses-' + ses
                    elif ses is None:
                        info = "sub-" + partic
                    cloud_utils.s3_get_data(buck, remo, bids_dir, modality, info=info)
            elif analysis_level == 'group':
                if len(session_label) > 1 and session_label[0] != 'None':
                    for ses in session_label:
                        info = 'ses-' + ses
                        cloud_utils.s3_get_data(buck, remo, bids_dir, modality, info=info)
                else:
                    cloud_utils.s3_get_data(buck, remo, bids_dir, modality)

        if len(sec_s3_objs) > 0:
            [access_key, secret_key] = cloud_utils.get_credentials()

            session = Session(
                aws_access_key_id=access_key,
                aws_secret_access_key=secret_key
            )

            s3_r = session.resource('s3')
            s3_c = cloud_utils.s3_client(service="s3")
            sec_dir = as_directory(home + "/.pynets/secondary_files", remove=False)
            for s3_obj in [i for i in sec_s3_objs if i is not None]:
                buck, remo = cloud_utils.parse_path(s3_obj)
                s3_c.download_file(buck, remo, f"{sec_dir}/{os.path.basename(s3_obj)}")

            if isinstance(bids_args.ua, list):
                local_ua = bids_args.ua.copy()
                for i in local_ua:
                    if i.startswith("s3://"):
                        local_ua[local_ua.index(i)] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.ua = local_ua
            if isinstance(bids_args.cm, list):
                local_cm = bids_args.cm.copy()
                for i in bids_args.cm:
                    if i.startswith("s3://"):
                        local_cm[local_cm.index(i)] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.cm = local_cm
            if isinstance(bids_args.roi, list):
                local_roi = bids_args.roi.copy()
                for i in bids_args.roi:
                    if i.startswith("s3://"):
                        local_roi[local_roi.index(i)] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.roi = local_roi
            if isinstance(bids_args.way, list):
                local_way = bids_args.way.copy()
                for i in bids_args.way:
                    if i.startswith("s3://"):
                        local_way[local_way.index(i)] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.way = local_way

            if bids_args.ref:
                if bids_args.ref.startswith("s3://"):
                    bids_args.ref = f"{sec_dir}/{os.path.basename(bids_args.ref)}"
    else:
        output_dir = bids_args.output_dir
        if output_dir is None:
            raise ValueError('Must specify an output directory')

    intermodal_dict = {k: [] for k in ['funcs', 'confs', 'dwis', 'bvals', 'bvecs', 'anats', 'masks',
                                       'subjs', 'seshs']}
    if analysis_level == 'group':
        if len(modality) > 1:
            i = 0
            for mod in modality:
                outs = sweep_directory(bids_dir, modality=mod, space=space, func_desc=func_desc, sesh=session_label)
                if mod == 'func':
                    if i == 0:
                        funcs, confs, _, _, _, anats, masks, subjs, seshs = outs
                    else:
                        funcs, confs, _, _, _, _, _, _, _ = outs
                    intermodal_dict['funcs'].append(funcs)
                    intermodal_dict['confs'].append(confs)
                elif mod == 'dwi':
                    if i == 0:
                        _, _, dwis, bvals, bvecs, anats, masks, subjs, seshs = outs
                    else:
                        _, _, dwis, bvals, bvecs, _, _, _, _ = outs
                    intermodal_dict['dwis'].append(dwis)
                    intermodal_dict['bvals'].append(bvals)
                    intermodal_dict['bvecs'].append(bvecs)
                intermodal_dict['anats'].append(anats)
                intermodal_dict['masks'].append(masks)
                intermodal_dict['subjs'].append(subjs)
                intermodal_dict['seshs'].append(seshs)
                i += 1
        else:
            intermodal_dict = None
            outs = sweep_directory(bids_dir, modality=modality[0], space=space, func_desc=func_desc,
                                   sesh=session_label)
            funcs, confs, dwis, bvals, bvecs, anats, masks, subjs, seshs = outs
    elif analysis_level == 'participant':
        if len(modality) > 1:
            i = 0
            for mod in modality:
                outs = sweep_directory(bids_dir, modality=mod, space=space, func_desc=func_desc,
                                       subj=participant_label, sesh=session_label)
                if mod == 'func':
                    if i == 0:
                        funcs, confs, _, _, _, anats, masks, subjs, seshs = outs
                    else:
                        funcs, confs, _, _, _, _, _, _, _ = outs
                    intermodal_dict['funcs'].append(funcs)
                    intermodal_dict['confs'].append(confs)
                elif mod == 'dwi':
                    if i == 0:
                        _, _, dwis, bvals, bvecs, anats, masks, subjs, seshs = outs
                    else:
                        _, _, dwis, bvals, bvecs, _, _, _, _ = outs
                    intermodal_dict['dwis'].append(dwis)
                    intermodal_dict['bvals'].append(bvals)
                    intermodal_dict['bvecs'].append(bvecs)
                intermodal_dict['anats'].append(anats)
                intermodal_dict['masks'].append(masks)
                intermodal_dict['subjs'].append(subjs)
                intermodal_dict['seshs'].append(seshs)
                i += 1
        else:
            intermodal_dict = None
            outs = sweep_directory(bids_dir, modality=modality[0], space=space, func_desc=func_desc,
                                   subj=participant_label, sesh=session_label)
            funcs, confs, dwis, bvals, bvecs, anats, masks, subjs, seshs = outs
    else:
        raise ValueError('Analysis level invalid. Must be `participant` or `group`. See --help.')

    if intermodal_dict:
        funcs, confs, dwis, bvals, bvecs, anats, masks, subjs, seshs = [list(set(list(flatten(i)))) for i in
                                                                        intermodal_dict.values()]

    arg_list = []
    for mod in modalities:
        arg_list.append(arg_dict[mod])

    arg_list.append(arg_dict['gen'])

    args_dict_all = {}
    models = []
    for d in arg_list:
        if 'mod' in d.keys():
            if len(modality) == 1:
                if any(x in d['mod'] for x in func_models):
                    if 'dwi' in modality:
                        del d['mod']
                elif any(x in d['mod'] for x in struct_models):
                    if 'func' in modality:
                        del d['mod']
            else:
                if any(x in d['mod'] for x in func_models) or any(x in d['mod'] for x in struct_models):
                    models.append(ast.literal_eval(d['mod']))
        args_dict_all.update(d)

    if len(modality) > 1:
        args_dict_all['mod'] = str(list(set(flatten(models))))

    print('Arguments parsed from bids_config.json:\n')
    print(args_dict_all)

    for key, val in args_dict_all.items():
        if isinstance(val, str):
            args_dict_all[key] = ast.literal_eval(val)

    id_list = []
    for i in sorted(list(set(subjs))):
        for ses in sorted(list(set(seshs))):
            id_list.append(i + '_' + ses)

    args_dict_all['work'] = bids_args.work
    args_dict_all['output_dir'] = output_dir
    args_dict_all['plug'] = bids_args.plug
    args_dict_all['pm'] = bids_args.pm
    args_dict_all['v'] = bids_args.v
    args_dict_all['clean'] = bids_args.clean
    if funcs is not None:
        args_dict_all['func'] = sorted(funcs)
    else:
        args_dict_all['func'] = None
    if confs is not None:
        args_dict_all['conf'] = sorted(confs)
    else:
        args_dict_all['conf'] = None
    if dwis is not None:
        args_dict_all['dwi'] = sorted(dwis)
        args_dict_all['bval'] = sorted(bvals)
        args_dict_all['bvec'] = sorted(bvecs)
    else:
        args_dict_all['dwi'] = None
        args_dict_all['bval'] = None
        args_dict_all['bvec'] = None
    if anats is not None:
        args_dict_all['anat'] = sorted(anats)
    else:
        args_dict_all['anat'] = None
    if masks is not None:
        args_dict_all['m'] = sorted(masks)
    else:
        args_dict_all['m'] = None
    args_dict_all['g'] = None
    if ('dwi' in modality) and (bids_args.way is not None):
        args_dict_all['way'] = bids_args.way
    else:
        args_dict_all['way'] = None
    args_dict_all['id'] = id_list
    args_dict_all['ua'] = bids_args.ua
    args_dict_all['ref'] = bids_args.ref
    args_dict_all['roi'] = bids_args.roi
    if ('func' in modality) and (bids_args.cm is not None):
        args_dict_all['cm'] = bids_args.cm
    else:
        args_dict_all['cm'] = None

    # Mimic argparse with SimpleNamespace object
    args = SimpleNamespace(**args_dict_all)
    print(args)

    set_start_method('forkserver')
    with Manager() as mgr:
        retval = mgr.dict()
        p = Process(target=build_workflow, args=(args, retval))
        p.start()
        p.join()
        if p.is_alive():
            p.terminate()

        retcode = p.exitcode or retval.get('return_code', 0)

        pynets_wf = retval.get('workflow', None)
        work_dir = retval.get('work_dir')
        plugin_settings = retval.get('plugin_settings', None)
        plugin_settings = retval.get('plugin_settings', None)
        execution_dict = retval.get('execution_dict', None)
        run_uuid = retval.get('run_uuid', None)

        retcode = retcode or int(pynets_wf is None)
        if retcode != 0:
            sys.exit(retcode)

        # Clean up master process before running workflow, which may create forks
        gc.collect()

    mgr.shutdown()

    if bids_args.push_location:
        print(f"Pushing to s3 at {bids_args.push_location}.")
        push_buck, push_remo = cloud_utils.parse_path(bids_args.push_location)
        for id in id_list:
            cloud_utils.s3_push_data(
                push_buck,
                push_remo,
                output_dir,
                modality,
                subject=id.split('_')[0],
                session=id.split('_')[1],
                creds=creds,
            )

    sys.exit(0)

    return
示例#2
0
def create_json(
    bucket,
    dataset,
    push_dir,
    modality,
    seshs,
    user_atlas,
    cluster_mask,
    roi,
    ref,
    way,
    plugin,
    resources,
    working_dir,
    verbose,
    jobdir,
    credentials,
):
    """Creates the json files for each of the jobs"""
    from pathlib import Path
    jobsjson = f"{jobdir}/jobs.json"

    # set up infrastructure
    out = subprocess.check_output(f"mkdir -p {jobdir}", shell=True)
    out = subprocess.check_output(f"mkdir -p {jobdir}/jobs/", shell=True)
    out = subprocess.check_output(f"mkdir -p {jobdir}/ids/", shell=True)

    with open("%s%s" % (str(Path(__file__).parent.parent), '/config/cloud_config.json'), 'r') as inf:
    #with open('/Users/derekpisner/Applications/PyNets/pynets/cloud_config.json') as inf:
        template = json.load(inf)

    co = template["containerOverrides"]
    cmd = co["command"]
    env = co["environment"]

    # modify template
    if credentials is not None:
        env[0]["value"], env[1]["value"] = get_credentials()
    else:
        env = []
    co["environment"] = env

    # edit non-defaults
    procmem = list(eval(str(resources)))
    jobs = []
    cmd[cmd.index("<INPUT>")] = f's3://{bucket}/{dataset}'
    cmd[cmd.index("<PUSH>")] = f's3://{bucket}/{push_dir}'
    cmd[cmd.index("<MODALITY>")] = modality[0]
    co["vcpus"] = int(procmem[0])
    co["memory"] = int(1000*float(procmem[1]))

    if user_atlas is not None:
        cmd.append('-ua')
        for i in user_atlas:
            cmd.append(i)
    if cluster_mask is not None:
        cmd.append('-cm')
        for i in cluster_mask:
            cmd.append(i)
    if roi is not None:
        cmd.append('-roi')
        for i in roi:
            cmd.append(i)
    if ref is not None:
        cmd.append('-ref')
        for i in ref:
            cmd.append(i)
    if way is not None:
        cmd.append('-way')
        for i in way:
            cmd.append(i)
    if verbose is True:
        cmd.append('-v')
    if plugin is not None:
        cmd.append('-plug')
        cmd.append(plugin)
    if plugin is not None:
        cmd.append('-pm')
        cmd.append(resources)
    if working_dir is not None:
        cmd.append('-work')
        cmd.append(working_dir)

    # edit participant-specific values ()
    # loop over every session of every participant
    for subj in seshs.keys():
        print(f"Generating job for sub-{subj}")
        # and for each subject number in each participant number,
        for sesh in seshs[subj]:
            # add format-specific commands,
            job_cmd = deepcopy(cmd)
            job_cmd[job_cmd.index("<SUBJ>")] = subj
            if sesh is not None:
                job_cmd[job_cmd.index("<SESH>")] = sesh

            # then, grab the template,
            # add additional parameters,
            # make the json file for this iteration,
            # and add the path to its json file to `jobs`.
            job_json = deepcopy(template)
            ver = pynets.__version__.replace(".", "-")
            if dataset:
                name = f"pynets_{ver}_{dataset}_sub-{subj}"
            else:
                name = f"pynets_{ver}_sub-{subj}"
            if sesh is not None:
                name = f"{name}_ses-{sesh}"
            print(job_cmd)
            job_json["jobName"] = name
            job_json["containerOverrides"]["command"] = job_cmd
            job = os.path.join(jobdir, "jobs", name + ".json")
            with open(job, "w") as outfile:
                json.dump(job_json, outfile)
            jobs += [job]

    # return list of job jsons
    with open(jobsjson, "w") as f:
        json.dump(jobs, f)
    return jobs
示例#3
0
def main():
    """Initializes main script from command-line call to generate
    single-subject or multi-subject workflow(s)"""
    import os
    import gc
    import sys
    import json
    from pynets.core.utils import build_args_from_config
    import itertools
    from types import SimpleNamespace
    import pkg_resources
    from pynets.core.utils import flatten
    from pynets.cli.pynets_run import build_workflow
    from multiprocessing import set_start_method, Process, Manager
    from colorama import Fore, Style

    try:
        import pynets
    except ImportError:
        print(
            "PyNets not installed! Ensure that you are referencing the correct"
            " site-packages and using Python3.6+"
        )

    if len(sys.argv) < 1:
        print("\nMissing command-line inputs! See help options with the -h"
              " flag.\n")
        sys.exit()

    print(f"{Fore.LIGHTBLUE_EX}\nBIDS API\n")

    print(Style.RESET_ALL)

    print(f"{Fore.LIGHTGREEN_EX}Obtaining Derivatives Layout...")

    print(Style.RESET_ALL)

    modalities = ["func", "dwi"]
    space = 'T1w'

    bids_args = get_bids_parser().parse_args()
    participant_label = bids_args.participant_label
    session_label = bids_args.session_label
    run = bids_args.run_label
    if isinstance(run, list):
        run = str(run[0]).zfill(2)
    modality = bids_args.modality
    bids_config = bids_args.config
    analysis_level = bids_args.analysis_level
    clean = bids_args.clean

    if analysis_level == "group" and participant_label is not None:
        raise ValueError(
            "Error: You have indicated a group analysis level run, but"
            " specified a participant label!"
        )

    if analysis_level == "participant" and participant_label is None:
        raise ValueError(
            "Error: You have indicated a participant analysis level run, but"
            " not specified a participant "
            "label!")

    if bids_config:
        with open(bids_config, "r") as stream:
            arg_dict = json.load(stream)
    else:
        with open(
            pkg_resources.resource_filename("pynets",
                                            "config/bids_config.json"),
            "r",
        ) as stream:
            arg_dict = json.load(stream)
        stream.close()

    # S3
    # Primary inputs
    s3 = bids_args.bids_dir.startswith("s3://")

    if not s3:
        bids_dir = bids_args.bids_dir

    # secondary inputs
    sec_s3_objs = []
    if isinstance(bids_args.ua, list):
        for i in bids_args.ua:
            if i.startswith("s3://"):
                print("Downloading user atlas: ", i, " from S3...")
                sec_s3_objs.append(i)
    if isinstance(bids_args.cm, list):
        for i in bids_args.cm:
            if i.startswith("s3://"):
                print("Downloading clustering mask: ", i, " from S3...")
                sec_s3_objs.append(i)
    if isinstance(bids_args.roi, list):
        for i in bids_args.roi:
            if i.startswith("s3://"):
                print("Downloading ROI mask: ", i, " from S3...")
                sec_s3_objs.append(i)
    if isinstance(bids_args.way, list):
        for i in bids_args.way:
            if i.startswith("s3://"):
                print("Downloading tractography waymask: ", i, " from S3...")
                sec_s3_objs.append(i)

    if bids_args.ref:
        if bids_args.ref.startswith("s3://"):
            print(
                "Downloading atlas labeling reference file: ",
                bids_args.ref,
                " from S3...",
            )
            sec_s3_objs.append(bids_args.ref)

    if s3 or len(sec_s3_objs) > 0:
        from boto3.session import Session
        from pynets.core import cloud_utils
        from pynets.core.utils import as_directory

        home = os.path.expanduser("~")
        creds = bool(cloud_utils.get_credentials())

        if s3:
            buck, remo = cloud_utils.parse_path(bids_args.bids_dir)
            os.makedirs(f"{home}/.pynets", exist_ok=True)
            os.makedirs(f"{home}/.pynets/input", exist_ok=True)
            os.makedirs(f"{home}/.pynets/output", exist_ok=True)
            bids_dir = as_directory(f"{home}/.pynets/input", remove=False)
            if (not creds) and bids_args.push_location:
                raise AttributeError(
                    """No AWS credentials found, but `--push_location` flag
                     called. Pushing will most likely fail.""")
            else:
                output_dir = as_directory(
                    f"{home}/.pynets/output", remove=False)

            # Get S3 input data if needed
            if analysis_level == "participant":
                for partic, ses in list(
                    itertools.product(participant_label, session_label)
                ):
                    if ses is not None:
                        info = "sub-" + partic + "/ses-" + ses
                    elif ses is None:
                        info = "sub-" + partic
                    cloud_utils.s3_get_data(
                        buck, remo, bids_dir, modality, info=info)
            elif analysis_level == "group":
                if len(session_label) > 1 and session_label[0] != "None":
                    for ses in session_label:
                        info = "ses-" + ses
                        cloud_utils.s3_get_data(
                            buck, remo, bids_dir, modality, info=info
                        )
                else:
                    cloud_utils.s3_get_data(buck, remo, bids_dir, modality)

        if len(sec_s3_objs) > 0:
            [access_key, secret_key] = cloud_utils.get_credentials()

            session = Session(
                aws_access_key_id=access_key, aws_secret_access_key=secret_key
            )

            s3_r = session.resource("s3")
            s3_c = cloud_utils.s3_client(service="s3")
            sec_dir = as_directory(
                home + "/.pynets/secondary_files", remove=False)
            for s3_obj in [i for i in sec_s3_objs if i is not None]:
                buck, remo = cloud_utils.parse_path(s3_obj)
                s3_c.download_file(
                    buck, remo, f"{sec_dir}/{os.path.basename(s3_obj)}")

            if isinstance(bids_args.ua, list):
                local_ua = bids_args.ua.copy()
                for i in local_ua:
                    if i.startswith("s3://"):
                        local_ua[local_ua.index(
                            i)] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.ua = local_ua
            if isinstance(bids_args.cm, list):
                local_cm = bids_args.cm.copy()
                for i in bids_args.cm:
                    if i.startswith("s3://"):
                        local_cm[local_cm.index(
                            i)] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.cm = local_cm
            if isinstance(bids_args.roi, list):
                local_roi = bids_args.roi.copy()
                for i in bids_args.roi:
                    if i.startswith("s3://"):
                        local_roi[
                            local_roi.index(i)
                        ] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.roi = local_roi
            if isinstance(bids_args.way, list):
                local_way = bids_args.way.copy()
                for i in bids_args.way:
                    if i.startswith("s3://"):
                        local_way[
                            local_way.index(i)
                        ] = f"{sec_dir}/{os.path.basename(i)}"
                bids_args.way = local_way

            if bids_args.ref:
                if bids_args.ref.startswith("s3://"):
                    bids_args.ref = f"{sec_dir}/" \
                                    f"{os.path.basename(bids_args.ref)}"
    else:
        output_dir = bids_args.output_dir
        if output_dir is None:
            raise ValueError("Must specify an output directory")

    intermodal_dict = {
        k: []
        for k in [
            "funcs",
            "confs",
            "dwis",
            "bvals",
            "bvecs",
            "anats",
            "masks",
            "subjs",
            "seshs",
        ]
    }
    if analysis_level == "group":
        if len(modality) > 1:
            i = 0
            for mod_ in modality:
                outs = sweep_directory(
                    bids_dir,
                    modality=mod_,
                    space=space,
                    sesh=session_label,
                    run=run
                )
                if mod_ == "func":
                    if i == 0:
                        funcs, confs, _, _, _, anats, masks, subjs, seshs =\
                            outs
                    else:
                        funcs, confs, _, _, _, _, _, _, _ = outs
                    intermodal_dict["funcs"].append(funcs)
                    intermodal_dict["confs"].append(confs)
                elif mod_ == "dwi":
                    if i == 0:
                        _, _, dwis, bvals, bvecs, anats, masks, subjs, seshs =\
                            outs
                    else:
                        _, _, dwis, bvals, bvecs, _, _, _, _ = outs
                    intermodal_dict["dwis"].append(dwis)
                    intermodal_dict["bvals"].append(bvals)
                    intermodal_dict["bvecs"].append(bvecs)
                intermodal_dict["anats"].append(anats)
                intermodal_dict["masks"].append(masks)
                intermodal_dict["subjs"].append(subjs)
                intermodal_dict["seshs"].append(seshs)
                i += 1
        else:
            intermodal_dict = None
            outs = sweep_directory(
                bids_dir,
                modality=modality[0],
                space=space,
                sesh=session_label,
                run=run
            )
            funcs, confs, dwis, bvals, bvecs, anats, masks, subjs, seshs = outs
    elif analysis_level == "participant":
        if len(modality) > 1:
            i = 0
            for mod_ in modality:
                outs = sweep_directory(
                    bids_dir,
                    modality=mod_,
                    space=space,
                    subj=participant_label,
                    sesh=session_label,
                    run=run
                )
                if mod_ == "func":
                    if i == 0:
                        funcs, confs, _, _, _, anats, masks, subjs, seshs =\
                            outs
                    else:
                        funcs, confs, _, _, _, _, _, _, _ = outs
                    intermodal_dict["funcs"].append(funcs)
                    intermodal_dict["confs"].append(confs)
                elif mod_ == "dwi":
                    if i == 0:
                        _, _, dwis, bvals, bvecs, anats, masks, subjs, seshs =\
                            outs
                    else:
                        _, _, dwis, bvals, bvecs, _, _, _, _ = outs
                    intermodal_dict["dwis"].append(dwis)
                    intermodal_dict["bvals"].append(bvals)
                    intermodal_dict["bvecs"].append(bvecs)
                intermodal_dict["anats"].append(anats)
                intermodal_dict["masks"].append(masks)
                intermodal_dict["subjs"].append(subjs)
                intermodal_dict["seshs"].append(seshs)
                i += 1
        else:
            intermodal_dict = None
            outs = sweep_directory(
                bids_dir,
                modality=modality[0],
                space=space,
                subj=participant_label,
                sesh=session_label,
                run=run
            )
            funcs, confs, dwis, bvals, bvecs, anats, masks, subjs, seshs = outs
    else:
        raise ValueError(
            "Analysis level invalid. Must be `participant` or `group`. See"
            " --help."
        )

    if intermodal_dict:
        funcs, confs, dwis, bvals, bvecs, anats, masks, subjs, seshs = [
            list(set(list(flatten(i)))) for i in intermodal_dict.values()
        ]

    args_dict_all = build_args_from_config(modality, arg_dict)

    id_list = []
    for i in sorted(list(set(subjs))):
        for ses in sorted(list(set(seshs))):
            id_list.append(i + "_" + ses)

    args_dict_all["work"] = bids_args.work
    args_dict_all["output_dir"] = output_dir
    args_dict_all["plug"] = bids_args.plug
    args_dict_all["pm"] = bids_args.pm
    args_dict_all["v"] = bids_args.v
    args_dict_all["clean"] = bids_args.clean
    if funcs is not None:
        args_dict_all["func"] = sorted(funcs)
    else:
        args_dict_all["func"] = None
    if confs is not None:
        args_dict_all["conf"] = sorted(confs)
    else:
        args_dict_all["conf"] = None
    if dwis is not None:
        args_dict_all["dwi"] = sorted(dwis)
        args_dict_all["bval"] = sorted(bvals)
        args_dict_all["bvec"] = sorted(bvecs)
    else:
        args_dict_all["dwi"] = None
        args_dict_all["bval"] = None
        args_dict_all["bvec"] = None
    if anats is not None:
        args_dict_all["anat"] = sorted(anats)
    else:
        args_dict_all["anat"] = None
    if masks is not None:
        args_dict_all["m"] = sorted(masks)
    else:
        args_dict_all["m"] = None
    args_dict_all["g"] = None
    if ("dwi" in modality) and (bids_args.way is not None):
        args_dict_all["way"] = bids_args.way
    else:
        args_dict_all["way"] = None
    args_dict_all["id"] = id_list
    args_dict_all["ua"] = bids_args.ua
    args_dict_all["ref"] = bids_args.ref
    args_dict_all["roi"] = bids_args.roi
    if ("func" in modality) and (bids_args.cm is not None):
        args_dict_all["cm"] = bids_args.cm
    else:
        args_dict_all["cm"] = None

    # Mimic argparse with SimpleNamespace object
    args = SimpleNamespace(**args_dict_all)
    print(args)

    set_start_method("forkserver")
    with Manager() as mgr:
        retval = mgr.dict()
        p = Process(target=build_workflow, args=(args, retval))
        p.start()
        p.join()
        if p.is_alive():
            p.terminate()

        retcode = p.exitcode or retval.get("return_code", 0)

        pynets_wf = retval.get("workflow", None)
        work_dir = retval.get("work_dir")
        plugin_settings = retval.get("plugin_settings", None)
        plugin_settings = retval.get("plugin_settings", None)
        execution_dict = retval.get("execution_dict", None)
        run_uuid = retval.get("run_uuid", None)

        retcode = retcode or int(pynets_wf is None)
        if retcode != 0:
            sys.exit(retcode)

        # Clean up master process before running workflow, which may create
        # forks
        gc.collect()

    mgr.shutdown()

    if bids_args.push_location:
        print(f"Pushing to s3 at {bids_args.push_location}.")
        push_buck, push_remo = cloud_utils.parse_path(bids_args.push_location)
        for id in id_list:
            cloud_utils.s3_push_data(
                push_buck,
                push_remo,
                output_dir,
                modality,
                subject=id.split("_")[0],
                session=id.split("_")[1],
                creds=creds,
            )

    sys.exit(0)

    return
示例#4
0
def create_json(bucket,
                path,
                threads,
                jobdir,
                credentials=None,
                dataset=None,
                modif=""):
    """Creates the json files for each of the jobs

    Parameters
    ----------
    bucket : str
        The S3 bucket with the input dataset formatted according to the BIDS standard.
    path : str
        The directory where the dataset is stored on the S3 bucket
    threads : OrderedDict
        dictionary containing all subjects and sessions from the path location
    jobdir : str
        Directory of batch jobs to generate/check up on
    credentials : [type], optional
        AWS formatted csv of credentials, by default None
    dataset : [type], optional
        Name added to the generated json job files "pynets_<version>_<dataset>_sub-<sub>_ses-<ses>", by default None
    modif : str, optional
        Name of folder on s3 to push to. If empty, push to a folder with pynets's version number, by default ""

    Returns
    -------
    list
        list of job jsons
    """
    from pathlib import Path
    jobsjson = f"{jobdir}/jobs.json"
    if os.path.isfile(jobsjson):
        with open(jobsjson, "r") as f:
            jobs = json.load(f)
        return jobs

    # set up infrastructure
    out = subprocess.check_output(f"mkdir -p {jobdir}", shell=True)
    out = subprocess.check_output(f"mkdir -p {jobdir}/jobs/", shell=True)
    out = subprocess.check_output(f"mkdir -p {jobdir}/ids/", shell=True)
    seshs = threads

    templ = os.path.dirname(__file__)

    with open(
            "%s%s" % (str(Path(__file__).parent.parent), '/cloud_config.json'),
            'r') as inf:
        template = json.load(inf)

    cmd = template["containerOverrides"]["command"]
    env = template["containerOverrides"]["environment"]

    # modify template
    if credentials is not None:
        env[0]["value"], env[1]["value"] = get_credentials()
    else:
        env = []
    template["containerOverrides"]["environment"] = env

    # edit non-defaults
    jobs = []
    cmd[cmd.index("<INPUT>")] = f's3://{bucket}/{path}'
    cmd[cmd.index("<PUSH>")] = f's3://{bucket}/{path}/{modif}'

    # edit participant-specific values ()
    # loop over every session of every participant
    for subj in seshs.keys():
        print(f"... Generating job for sub-{subj}")
        # and for each subject number in each participant number,
        for sesh in seshs[subj]:
            # add format-specific commands,
            job_cmd = deepcopy(cmd)
            job_cmd[job_cmd.index("<SUBJ>")] = subj
            if sesh is not None:
                job_cmd[job_cmd.index("<SESH>")] = sesh

            # then, grab the template,
            # add additional parameters,
            # make the json file for this iteration,
            # and add the path to its json file to `jobs`.
            job_json = deepcopy(template)
            ver = pynets.__version__.replace(".", "-")
            if dataset:
                name = f"pynets_{ver}_{dataset}_sub-{subj}"
            else:
                name = f"pynets_{ver}_sub-{subj}"
            if sesh is not None:
                name = f"{name}_ses-{sesh}"
            print(job_cmd)
            job_json["jobName"] = name
            job_json["containerOverrides"]["command"] = job_cmd
            job = os.path.join(jobdir, "jobs", name + ".json")
            with open(job, "w") as outfile:
                json.dump(job_json, outfile)
            jobs += [job]

    # return list of job jsons
    with open(jobsjson, "w") as f:
        json.dump(jobs, f)
    return jobs