htsworkflow/pipelines/sequences.py

   1 """
   2 Utilities to work with the various eras of sequence archive files
   3 """
   4 import collections
   5 import logging
   6 import os
   7 import types
   8 import re
   9
  10 LOGGER = logging.getLogger(__name__)
  11
  12 eland_re = re.compile('s_(?P<lane>\d)(_(?P<read>\d))?_eland_')
  13 raw_seq_re = re.compile('woldlab_[0-9]{6}_[^_]+_[\d]+_[\dA-Za-z]+')
  14 qseq_re = re.compile('woldlab_[0-9]{6}_[^_]+_[\d]+_[\dA-Za-z]+_l[\d]_r[\d].tar.bz2')
  15
  16 SEQUENCE_TABLE_NAME = "sequences"
  17 def create_sequence_table(cursor):
  18     """
  19     Create a SQL table to hold  SequenceFile entries
  20     """
  21     sql = """
  22 CREATE TABLE %(table)s (
  23   filetype   CHAR(8),
  24   path       TEXT,
  25   flowcell   CHAR(8),
  26   lane       INTEGER,
  27   read       INTEGER,
  28   pf         BOOLEAN,
  29   cycle      CHAR(8)
  30 );
  31 """ %( {'table': SEQUENCE_TABLE_NAME} )
  32     return cursor.execute(sql)
  33
  34 FlowcellPath = collections.namedtuple('FlowcellPath',
  35                                       'flowcell start stop project')
  36
  37 class SequenceFile(object):
  38     """
  39     Simple container class that holds the path to a sequence archive
  40     and basic descriptive information.
  41     """
  42     def __init__(self, filetype, path, flowcell, lane,
  43                  read=None,
  44                  pf=None,
  45                  cycle=None,
  46                  project=None,
  47                  index=None,
  48                  split=None):
  49         """Store various fields used in our sequence files
  50
  51         filetype is one of 'qseq', 'srf', 'fastq'
  52         path = location of file
  53         flowcell = files flowcell id
  54         lane = which lane
  55         read = which sequencer read, usually 1 or 2
  56         pf = did it pass filter
  57         cycle = cycle dir name e.g. C1-202
  58         project = projed name from HiSeq, probably library ID
  59         index = HiSeq barcode index sequence
  60         split = file fragment from HiSeq (Since one file is split into many)
  61         """
  62         self.filetype = filetype
  63         self.path = path
  64         self.flowcell = flowcell
  65         self.lane = lane
  66         self.read = read
  67         self.pf = pf
  68         self.cycle = cycle
  69         self.project = project
  70         self.index = index
  71         self.split = split
  72
  73     def __hash__(self):
  74         return hash(self.key())
  75
  76     def key(self):
  77         return (self.flowcell, self.lane, self.read, self.project, self.split)
  78
  79     def unicode(self):
  80         return unicode(self.path)
  81
  82     def __eq__(self, other):
  83         """
  84         Equality is defined if everything but the path matches
  85         """
  86         attributes = ['filetype','flowcell', 'lane', 'read', 'pf', 'cycle', 'project', 'index']
  87         for a in attributes:
  88             if getattr(self, a) != getattr(other, a):
  89                 return False
  90
  91         return True
  92
  93     def __repr__(self):
  94         return u"<%s %s %s %s>" % (self.filetype, self.flowcell, self.lane, self.path)
  95
  96     def make_target_name(self, root):
  97         """
  98         Create target name for where we need to link this sequence too
  99         """
 100         path, basename = os.path.split(self.path)
 101         # Because the names aren't unque we include the flowcel name
 102         # because there were different eland files for different length
 103         # analyses, we include the cycle length in the name.
 104         if self.filetype == 'eland':
 105             template = "%(flowcell)s_%(cycle)s_%(eland)s"
 106             basename = template % { 'flowcell': self.flowcell,
 107                                     'cycle': self.cycle,
 108                                     'eland': basename }
 109         # else:
 110         # all the other file types have names that include flowcell/lane
 111         # information and thus are unique so we don't have to do anything
 112         return os.path.join(root, basename)
 113
 114     def save(self, cursor):
 115         """
 116         Add this entry to a DB2.0 database.
 117         """
 118         #FIXME: NEEDS SQL ESCAPING
 119         header_macro = {'table': SEQUENCE_TABLE_NAME }
 120         sql_header = "insert into %(table)s (" % header_macro
 121         sql_columns = ['filetype','path','flowcell','lane']
 122         sql_middle = ") values ("
 123         sql_values = [self.filetype, self.path, self.flowcell, self.lane]
 124         sql_footer = ");"
 125         for name in ['read', 'pf', 'cycle']:
 126             value = getattr(self, name)
 127             if value is not None:
 128                 sql_columns.append(name)
 129                 sql_values.append(value)
 130
 131         sql = " ".join([sql_header,
 132                         ", ".join(sql_columns),
 133                         sql_middle,
 134                         # note the following makes a string like ?,?,?
 135                         ",".join(["?"] * len(sql_values)),
 136                         sql_footer])
 137
 138         return cursor.execute(sql, sql_values)
 139
 140 def get_flowcell_cycle(path):
 141     """
 142     Extract flowcell, cycle from pathname
 143     """
 144     path = os.path.normpath(path)
 145     project = None
 146     rest, tail = os.path.split(path)
 147     if tail.startswith('Project_'):
 148         # we're in a multiplexed sample
 149         project = tail
 150         rest, cycle = os.path.split(rest)
 151     else:
 152         cycle = tail
 153
 154     rest, flowcell = os.path.split(rest)
 155     cycle_match = re.match("C(?P<start>[0-9]+)-(?P<stop>[0-9]+)", cycle)
 156     if cycle_match is None:
 157         raise ValueError(
 158             "Expected .../flowcell/cycle/ directory structure in %s" % \
 159             (path,))
 160     start = cycle_match.group('start')
 161     if start is not None:
 162         start = int(start)
 163     stop = cycle_match.group('stop')
 164     if stop is not None:
 165         stop = int(stop)
 166
 167     return FlowcellPath(flowcell, start, stop, project)
 168
 169 def parse_srf(path, filename):
 170     flowcell_dir, start, stop, project = get_flowcell_cycle(path)
 171     basename, ext = os.path.splitext(filename)
 172     records = basename.split('_')
 173     flowcell = records[4]
 174     lane = int(records[5][0])
 175     fullpath = os.path.join(path, filename)
 176
 177     if flowcell_dir != flowcell:
 178         LOGGER.warn("flowcell %s found in wrong directory %s" % \
 179                          (flowcell, path))
 180
 181     return SequenceFile('srf', fullpath, flowcell, lane, cycle=stop)
 182
 183 def parse_qseq(path, filename):
 184     flowcell_dir, start, stop, project = get_flowcell_cycle(path)
 185     basename, ext = os.path.splitext(filename)
 186     records = basename.split('_')
 187     fullpath = os.path.join(path, filename)
 188     flowcell = records[4]
 189     lane = int(records[5][1])
 190     read = int(records[6][1])
 191
 192     if flowcell_dir != flowcell:
 193         LOGGER.warn("flowcell %s found in wrong directory %s" % \
 194                          (flowcell, path))
 195
 196     return SequenceFile('qseq', fullpath, flowcell, lane, read, cycle=stop)
 197
 198 def parse_fastq(path, filename):
 199     """Parse fastq names
 200     """
 201     flowcell_dir, start, stop, project = get_flowcell_cycle(path)
 202     basename = re.sub('\.fastq(\.gz|\.bz2)?$', '', filename)
 203     records = basename.split('_')
 204     fullpath = os.path.join(path, filename)
 205     if project is not None:
 206         # demultiplexed sample!
 207         flowcell = flowcell_dir
 208         lane = int(records[2][-1])
 209         read = int(records[3][-1])
 210         pf = True # as I understand it hiseq runs toss the ones that fail filter
 211         index = records[1]
 212         project_id = records[0]
 213         split = records[4]
 214         sequence_type = 'split_fastq'
 215     else:
 216         flowcell = records[4]
 217         lane = int(records[5][1])
 218         read = int(records[6][1])
 219         pf = parse_fastq_pf_flag(records)
 220         index = None
 221         project_id = None
 222         split = None
 223         sequence_type = 'fastq'
 224
 225     if flowcell_dir != flowcell:
 226         LOGGER.warn("flowcell %s found in wrong directory %s" % \
 227                          (flowcell, path))
 228
 229     return SequenceFile(sequence_type, fullpath, flowcell, lane, read,
 230                         pf=pf,
 231                         cycle=stop,
 232                         project=project_id,
 233                         index=index,
 234                         split=split)
 235
 236 def parse_fastq_pf_flag(records):
 237     """Take a fastq filename split on _ and look for the pass-filter flag
 238     """
 239     if len(records) < 8:
 240         pf = None
 241     else:
 242         fastq_type = records[-1].lower()
 243         if fastq_type.startswith('pass'):
 244             pf = True
 245         elif fastq_type.startswith('nopass'):
 246             pf = False
 247         elif fastq_type.startswith('all'):
 248             pf = None
 249         else:
 250             raise ValueError("Unrecognized fastq name %s at %s" % \
 251                              (records[-1], os.path.join(path,filename)))
 252
 253     return pf
 254
 255 def parse_eland(path, filename, eland_match=None):
 256     if eland_match is None:
 257         eland_match = eland_re.match(filename)
 258     fullpath = os.path.join(path, filename)
 259     flowcell, start, stop, project = get_flowcell_cycle(path)
 260     if eland_match.group('lane'):
 261         lane = int(eland_match.group('lane'))
 262     else:
 263         lane = None
 264     if eland_match.group('read'):
 265         read = int(eland_match.group('read'))
 266     else:
 267         read = None
 268     return SequenceFile('eland', fullpath, flowcell, lane, read, cycle=stop)
 269
 270 def scan_for_sequences(dirs):
 271     """
 272     Scan through a list of directories for sequence like files
 273     """
 274     sequences = []
 275     if type(dirs) in types.StringTypes:
 276         raise ValueError("You probably want a list or set, not a string")
 277
 278     for d in dirs:
 279         LOGGER.info("Scanning %s for sequences" % (d,))
 280         if not os.path.exists(d):
 281             LOGGER.warn("Flowcell directory %s does not exist" % (d,))
 282             continue
 283
 284         for path, dirname, filenames in os.walk(d):
 285             for f in filenames:
 286                 seq = None
 287                 # find sequence files
 288                 if f.endswith('.md5'):
 289                     continue
 290                 elif f.endswith('.srf') or f.endswith('.srf.bz2'):
 291                     seq = parse_srf(path, f)
 292                 elif qseq_re.match(f):
 293                     seq = parse_qseq(path, f)
 294                 elif f.endswith('.fastq') or \
 295                      f.endswith('.fastq.bz2') or \
 296                      f.endswith('.fastq.gz'):
 297                     seq = parse_fastq(path, f)
 298                 eland_match = eland_re.match(f)
 299                 if eland_match:
 300                     if f.endswith('.md5'):
 301                         continue
 302                     seq = parse_eland(path, f, eland_match)
 303                 if seq:
 304                     sequences.append(seq)
 305                     LOGGER.debug("Found sequence at %s" % (f,))
 306
 307     return sequences