X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/7a041a2b0068e02bbb9743407d915522a1e2d1dc..0ff7b94edaaaa07932ae757c0a2b7ba3fde026cb:/sdk/cwl/arvados_cwl/runner.py diff --git a/sdk/cwl/arvados_cwl/runner.py b/sdk/cwl/arvados_cwl/runner.py index d7d5d2b32d..2e52b63ed6 100644 --- a/sdk/cwl/arvados_cwl/runner.py +++ b/sdk/cwl/arvados_cwl/runner.py @@ -4,6 +4,7 @@ from functools import partial import logging import json import re +from cStringIO import StringIO import cwltool.draft2tool from cwltool.draft2tool import CommandLineTool @@ -13,6 +14,7 @@ from cwltool.load_tool import fetch_document from cwltool.pathmapper import adjustFileObjs, adjustDirObjs import arvados.collection +import ruamel.yaml as yaml from .arvdocker import arv_docker_get_image from .pathmapper import ArvPathMapper @@ -21,6 +23,88 @@ logger = logging.getLogger('arvados.cwl-runner') cwltool.draft2tool.ACCEPTLIST_RE = re.compile(r"^[a-zA-Z0-9._+-]+$") +def upload_dependencies(arvrunner, name, document_loader, + workflowobj, uri, loadref_run): + """Upload the dependencies of the workflowobj document to Keep. + + Returns a pathmapper object mapping local paths to keep references. Also + does an in-place update of references in "workflowobj". + + Use scandeps to find $import, $include, $schemas, run, File and Directory + fields that represent external references. + + If workflowobj has an "id" field, this will reload the document to ensure + it is scanning the raw document prior to preprocessing. + """ + + loaded = set() + def loadref(b, u): + joined = urlparse.urljoin(b, u) + defrg, _ = urlparse.urldefrag(joined) + if defrg not in loaded: + loaded.add(defrg) + # Use fetch_text to get raw file (before preprocessing). + text = document_loader.fetch_text(defrg) + if isinstance(text, bytes): + textIO = StringIO(text.decode('utf-8')) + else: + textIO = StringIO(text) + return yaml.safe_load(textIO) + else: + return {} + + if loadref_run: + loadref_fields = set(("$import", "run")) + else: + loadref_fields = set(("$import",)) + + scanobj = workflowobj + if "id" in workflowobj: + # Need raw file content (before preprocessing) to ensure + # that external references in $include and $mixin are captured. + scanobj = loadref("", workflowobj["id"]) + + sc = scandeps(uri, scanobj, + loadref_fields, + set(("$include", "$schemas", "location")), + loadref) + + files = [] + def visitFiles(path): + files.append(path) + + adjustFileObjs(sc, visitFiles) + adjustDirObjs(sc, visitFiles) + + normalizeFilesDirs(files) + + if "id" in workflowobj: + files.append({"class": "File", "location": workflowobj["id"]}) + + mapper = ArvPathMapper(arvrunner, files, "", + "keep:%s", + "keep:%s/%s", + name=name) + + def setloc(p): + if not p["location"].startswith("keep:"): + p["location"] = mapper.mapper(p["location"]).resolved + adjustFileObjs(workflowobj, setloc) + adjustDirObjs(workflowobj, setloc) + + return mapper + + +def upload_docker(arvrunner, tool): + if isinstance(tool, CommandLineTool): + (docker_req, docker_is_req) = get_feature(tool, "DockerRequirement") + if docker_req: + arv_docker_get_image(arvrunner.api, docker_req, True, arvrunner.project_uuid) + elif isinstance(tool, cwltool.workflow.Workflow): + for s in tool.steps: + upload_docker(arvrunner, s.embedded_tool) + + class Runner(object): def __init__(self, runner, tool, job_order, enable_reuse): self.arvrunner = runner @@ -33,67 +117,24 @@ class Runner(object): def update_pipeline_component(self, record): pass - def upload_docker(self, tool): - if isinstance(tool, CommandLineTool): - (docker_req, docker_is_req) = get_feature(tool, "DockerRequirement") - if docker_req: - arv_docker_get_image(self.arvrunner.api, docker_req, True, self.arvrunner.project_uuid) - elif isinstance(tool, cwltool.workflow.Workflow): - for s in tool.steps: - self.upload_docker(s.embedded_tool) - - def arvados_job_spec(self, *args, **kwargs): - self.upload_docker(self.tool) - - workflowfiles = [] - jobfiles = [] - workflowfiles.append({"class":"File", "location": self.tool.tool["id"]}) + upload_docker(self.arvrunner, self.tool) self.name = os.path.basename(self.tool.tool["id"]) - def visitFiles(files, path): - files.append(path) - - document_loader, workflowobj, uri = fetch_document(self.tool.tool["id"]) - loaded = set() - def loadref(b, u): - joined = urlparse.urljoin(b, u) - if joined not in loaded: - loaded.add(joined) - return document_loader.fetch(urlparse.urljoin(b, u)) - else: - return {} - - sc = scandeps(uri, workflowobj, - set(("$import", "run")), - set(("$include", "$schemas", "path", "location")), - loadref) - adjustFileObjs(sc, partial(visitFiles, workflowfiles)) - adjustFileObjs(self.job_order, partial(visitFiles, jobfiles)) - adjustDirObjs(sc, partial(visitFiles, workflowfiles)) - adjustDirObjs(self.job_order, partial(visitFiles, jobfiles)) - - normalizeFilesDirs(jobfiles) - normalizeFilesDirs(workflowfiles) - - keepprefix = kwargs.get("keepprefix", "") - workflowmapper = ArvPathMapper(self.arvrunner, workflowfiles, "", - keepprefix+"%s", - keepprefix+"%s/%s", - name=self.name, - **kwargs) - - jobmapper = ArvPathMapper(self.arvrunner, jobfiles, "", - keepprefix+"%s", - keepprefix+"%s/%s", - name=os.path.basename(self.job_order.get("id", "#")), - **kwargs) - - def setloc(p): - p["location"] = jobmapper.mapper(p["location"])[1] - adjustFileObjs(self.job_order, setloc) - adjustDirObjs(self.job_order, setloc) + workflowmapper = upload_dependencies(self.arvrunner, + self.name, + self.tool.doc_loader, + self.tool.tool, + self.tool.tool["id"], + True) + + jobmapper = upload_dependencies(self.arvrunner, + os.path.basename(self.job_order.get("id", "#")), + self.tool.doc_loader, + self.job_order, + self.job_order.get("id", "#"), + False) if "id" in self.job_order: del self.job_order["id"]