sdk/python/arvados/stream.py

   1 # Copyright (C) The Arvados Authors. All rights reserved.
   2 #
   3 # SPDX-License-Identifier: Apache-2.0
   4
   5 from __future__ import print_function
   6 from __future__ import absolute_import
   7 from future.utils import listvalues
   8 from builtins import object
   9 import collections
  10 import hashlib
  11 import os
  12 import re
  13 import threading
  14 import functools
  15 import copy
  16
  17 from ._ranges import locators_and_ranges, Range
  18 from .arvfile import StreamFileReader
  19 from arvados.retry import retry_method
  20 from arvados.keep import *
  21 from . import config
  22 from . import errors
  23 from ._normalize_stream import normalize_stream
  24
  25 class StreamReader(object):
  26     def __init__(self, tokens, keep=None, debug=False, _empty=False,
  27                  num_retries=10):
  28         self._stream_name = None
  29         self._data_locators = []
  30         self._files = collections.OrderedDict()
  31         self._keep = keep
  32         self.num_retries = num_retries
  33
  34         streamoffset = 0
  35
  36         # parse stream
  37         for tok in tokens:
  38             if debug: print('tok', tok)
  39             if self._stream_name is None:
  40                 self._stream_name = tok.replace('\\040', ' ')
  41                 continue
  42
  43             s = re.match(r'^[0-9a-f]{32}\+(\d+)(\+\S+)*$', tok)
  44             if s:
  45                 blocksize = int(s.group(1))
  46                 self._data_locators.append(Range(tok, streamoffset, blocksize, 0))
  47                 streamoffset += blocksize
  48                 continue
  49
  50             s = re.search(r'^(\d+):(\d+):(\S+)', tok)
  51             if s:
  52                 pos = int(s.group(1))
  53                 size = int(s.group(2))
  54                 name = s.group(3).replace('\\040', ' ')
  55                 if name not in self._files:
  56                     self._files[name] = StreamFileReader(self, [Range(pos, 0, size, 0)], name)
  57                 else:
  58                     filereader = self._files[name]
  59                     filereader.segments.append(Range(pos, filereader.size(), size))
  60                 continue
  61
  62             raise errors.SyntaxError("Invalid manifest format")
  63
  64     def name(self):
  65         return self._stream_name
  66
  67     def files(self):
  68         return self._files
  69
  70     def all_files(self):
  71         return listvalues(self._files)
  72
  73     def size(self):
  74         n = self._data_locators[-1]
  75         return n.range_start + n.range_size
  76
  77     def locators_and_ranges(self, range_start, range_size):
  78         return locators_and_ranges(self._data_locators, range_start, range_size)
  79
  80     @retry_method
  81     def _keepget(self, locator, num_retries=None):
  82         return self._keep.get(locator, num_retries=num_retries)
  83
  84     @retry_method
  85     def readfrom(self, start, size, num_retries=None):
  86         """Read up to 'size' bytes from the stream, starting at 'start'"""
  87         if size == 0:
  88             return b''
  89         if self._keep is None:
  90             self._keep = KeepClient(num_retries=self.num_retries)
  91         data = []
  92         for lr in locators_and_ranges(self._data_locators, start, size):
  93             data.append(self._keepget(lr.locator, num_retries=num_retries)[lr.segment_offset:lr.segment_offset+lr.segment_size])
  94         return b''.join(data)
  95
  96     def manifest_text(self, strip=False):
  97         manifest_text = [self.name().replace(' ', '\\040')]
  98         if strip:
  99             for d in self._data_locators:
 100                 m = re.match(r'^[0-9a-f]{32}\+\d+', d.locator)
 101                 manifest_text.append(m.group(0))
 102         else:
 103             manifest_text.extend([d.locator for d in self._data_locators])
 104         manifest_text.extend([' '.join(["{}:{}:{}".format(seg.locator, seg.range_size, f.name.replace(' ', '\\040'))
 105                                         for seg in f.segments])
 106                               for f in listvalues(self._files)])
 107         return ' '.join(manifest_text) + '\n'