sdk/cwl/arvados_cwl/arvworkflow.py

   1 import os
   2 import json
   3 import copy
   4 import logging
   5
   6 from schema_salad.sourceline import SourceLine, cmap
   7
   8 from cwltool.pack import pack
   9 from cwltool.load_tool import fetch_document
  10 from cwltool.process import shortname
  11 from cwltool.workflow import Workflow, WorkflowException
  12 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
  13
  14 import ruamel.yaml as yaml
  15
  16 from .runner import upload_dependencies, trim_listing, packed_workflow, upload_workflow_collection
  17 from .arvtool import ArvadosCommandTool
  18 from .perf import Perf
  19
  20 logger = logging.getLogger('arvados.cwl-runner')
  21 metrics = logging.getLogger('arvados.cwl-runner.metrics')
  22
  23 def upload_workflow(arvRunner, tool, job_order, project_uuid, uuid=None,
  24                     submit_runner_ram=0, name=None):
  25
  26     packed = packed_workflow(arvRunner, tool)
  27
  28     adjustDirObjs(job_order, trim_listing)
  29
  30     main = [p for p in packed["$graph"] if p["id"] == "#main"][0]
  31     for inp in main["inputs"]:
  32         sn = shortname(inp["id"])
  33         if sn in job_order:
  34             inp["default"] = job_order[sn]
  35
  36     if not name:
  37         name = tool.tool.get("label", os.path.basename(tool.tool["id"]))
  38
  39     upload_dependencies(arvRunner, name, tool.doc_loader,
  40                         packed, tool.tool["id"], False)
  41
  42     # TODO nowhere for submit_runner_ram to go.
  43
  44     body = {
  45         "workflow": {
  46             "name": name,
  47             "description": tool.tool.get("doc", ""),
  48             "definition":yaml.round_trip_dump(packed)
  49         }}
  50     if project_uuid:
  51         body["workflow"]["owner_uuid"] = project_uuid
  52
  53     if uuid:
  54         call = arvRunner.api.workflows().update(uuid=uuid, body=body)
  55     else:
  56         call = arvRunner.api.workflows().create(body=body)
  57     return call.execute(num_retries=arvRunner.num_retries)["uuid"]
  58
  59 def dedup_reqs(reqs):
  60     dedup = {}
  61     for r in reversed(reqs):
  62         if r["class"] not in dedup and not r["class"].startswith("http://arvados.org/cwl#"):
  63             dedup[r["class"]] = r
  64     return [dedup[r] for r in sorted(dedup.keys())]
  65
  66 class ArvadosWorkflow(Workflow):
  67     """Wrap cwltool Workflow to override selected methods."""
  68
  69     def __init__(self, arvrunner, toolpath_object, **kwargs):
  70         super(ArvadosWorkflow, self).__init__(toolpath_object, **kwargs)
  71         self.arvrunner = arvrunner
  72         self.work_api = kwargs["work_api"]
  73         self.wf_pdh = None
  74
  75     def job(self, joborder, output_callback, **kwargs):
  76         kwargs["work_api"] = self.work_api
  77         req, _ = self.get_requirement("http://arvados.org/cwl#RunInSingleContainer")
  78         if req:
  79             with SourceLine(self.tool, None, WorkflowException):
  80                 if "id" not in self.tool:
  81                     raise WorkflowException("%s object must have 'id'" % (self.tool["class"]))
  82             document_loader, workflowobj, uri = (self.doc_loader, self.doc_loader.fetch(self.tool["id"]), self.tool["id"])
  83
  84             with Perf(metrics, "subworkflow upload_deps"):
  85                 upload_dependencies(self.arvrunner,
  86                                     os.path.basename(joborder.get("id", "#")),
  87                                     document_loader,
  88                                     joborder,
  89                                     joborder.get("id", "#"),
  90                                     False)
  91
  92                 if self.wf_pdh is None:
  93                     workflowobj["requirements"] = dedup_reqs(self.requirements)
  94                     workflowobj["hints"] = dedup_reqs(self.hints)
  95
  96                     packed = pack(document_loader, workflowobj, uri, self.metadata)
  97
  98                     upload_dependencies(self.arvrunner,
  99                                         kwargs.get("name", ""),
 100                                         document_loader,
 101                                         packed,
 102                                         uri,
 103                                         False)
 104
 105             with Perf(metrics, "subworkflow adjust"):
 106                 joborder_keepmount = copy.deepcopy(joborder)
 107
 108                 def keepmount(obj):
 109                     with SourceLine(obj, None, WorkflowException):
 110                         if "location" not in obj:
 111                             raise WorkflowException("%s object is missing required 'location' field: %s" % (obj["class"], obj))
 112                     with SourceLine(obj, "location", WorkflowException):
 113                         if obj["location"].startswith("keep:"):
 114                             obj["location"] = "/keep/" + obj["location"][5:]
 115                             if "listing" in obj:
 116                                 del obj["listing"]
 117                         elif obj["location"].startswith("_:"):
 118                             del obj["location"]
 119                         else:
 120                             raise WorkflowException("Location is not a keep reference or a literal: '%s'" % obj["location"])
 121
 122                 adjustFileObjs(joborder_keepmount, keepmount)
 123                 adjustDirObjs(joborder_keepmount, keepmount)
 124
 125                 if self.wf_pdh is None:
 126                     adjustFileObjs(packed, keepmount)
 127                     adjustDirObjs(packed, keepmount)
 128                     self.wf_pdh = upload_workflow_collection(self.arvrunner, shortname(self.tool["id"]), packed)
 129
 130             wf_runner = cmap({
 131                 "class": "CommandLineTool",
 132                 "baseCommand": "cwltool",
 133                 "inputs": self.tool["inputs"],
 134                 "outputs": self.tool["outputs"],
 135                 "stdout": "cwl.output.json",
 136                 "requirements": workflowobj["requirements"]+[
 137                     {
 138                     "class": "InitialWorkDirRequirement",
 139                     "listing": [{
 140                             "entryname": "workflow.cwl",
 141                             "entry": {
 142                                 "class": "File",
 143                                 "location": "keep:%s/workflow.cwl" % self.wf_pdh
 144                             }
 145                         }, {
 146                             "entryname": "cwl.input.yml",
 147                             "entry": json.dumps(joborder_keepmount, indent=2, sort_keys=True, separators=(',',': ')).replace("\\", "\\\\").replace('$(', '\$(').replace('${', '\${')
 148                         }]
 149                 }],
 150                 "hints": workflowobj["hints"],
 151                 "arguments": ["--no-container", "--move-outputs", "--preserve-entire-environment", "workflow.cwl#main", "cwl.input.yml"]
 152             })
 153             kwargs["loader"] = self.doc_loader
 154             kwargs["avsc_names"] = self.doc_schema
 155             return ArvadosCommandTool(self.arvrunner, wf_runner, **kwargs).job(joborder, output_callback, **kwargs)
 156         else:
 157             return super(ArvadosWorkflow, self).job(joborder, output_callback, **kwargs)