crunch_scripts/decompress-all.py

   1 #!/usr/bin/env python
   2
   3 #
   4 # decompress-all.py
   5 #
   6 # Decompress all compressed files in the collection using the "dtrx" tool and
   7 # produce a new collection with the contents.  Uncompressed files
   8 # are passed through.
   9 #
  10 # input:
  11 # A collection at script_parameters["input"]
  12 #
  13 # output:
  14 # A manifest of the uncompressed contents of the input collection.
  15
  16 import arvados
  17 import re
  18 import subprocess
  19 import os
  20 import sys
  21
  22 arvados.job_setup.one_task_per_input_file(if_sequence=0, and_end_task=True,
  23                                           input_as_path=True)
  24
  25 task = arvados.current_task()
  26
  27 input_file = task['parameters']['input']
  28
  29 infile_parts = re.match(r"(^[a-f0-9]{32}\+\d+)(\+\S+)*(/.*)?(/[^/]+)$", input_file)
  30
  31 outdir = os.path.join(task.tmpdir, "output")
  32 os.makedirs(outdir)
  33 os.chdir(outdir)
  34
  35 if infile_parts == None:
  36     print >>sys.stderr, "Failed to parse input filename '%s' as a Keep file\n" % input_file
  37     sys.exit(1)
  38
  39 cr = arvados.CollectionReader(infile_parts.group(1))
  40 streamname = infile_parts.group(3)[1:]
  41 filename = infile_parts.group(4)[1:]
  42
  43 if streamname != None:
  44     subprocess.call(["mkdir", "-p", streamname])
  45     os.chdir(streamname)
  46 else:
  47     streamname = '.'
  48
  49 m = re.match(r'\.(gz|Z|bz2|tgz|tbz|zip|rar|7z|cab|deb|rpm|cpio|gem)$', arvados.get_task_param_mount('input'), re.IGNORECASE)
  50
  51 if m != None:
  52     rc = subprocess.call(["dtrx", "-r", "-n", "-q", arvados.get_task_param_mount('input')])
  53     if rc == 0:
  54         out = arvados.CollectionWriter()
  55         out.write_directory_tree(outdir, max_manifest_depth=0)
  56         task.set_output(out.finish())
  57     else:
  58         return rc
  59 else:
  60     streamreader = filter(lambda s: s.name() == streamname, cr.all_streams())[0]
  61     filereader = streamreader.files()[filename]
  62     task.set_output(streamname + filereader.as_manifest()[1:])