Warning

This document is for an in-development version of Galaxy. You can alternatively view this page in the latest release if it exists or view the top of the latest release's documentation.

Source code for galaxy.datatypes.blast

# This file is now part of the Galaxy Project, but due to historical reasons
# reflecting time developed outside of the Galaxy Project, this file is under
# the MIT license.
#
# The MIT License (MIT)
# Copyright (c) 2012,2013,2014,2015,2016 Peter Cock
# Copyright (c) 2012 Edward Kirton
# Copyright (c) 2013 Nicola Soranzo
# Copyright (c) 2014 Bjoern Gruening
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to
# deal in the Software without restriction, including without limitation the
# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
# sell copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
# DEALINGS IN THE SOFTWARE.
#
"""NCBI BLAST datatypes.

Covers the ``blastxml`` format and the BLAST databases.
"""
import io
import logging
import os
from time import sleep

from galaxy.util import smart_str
from .data import (
    Data,
    get_file_peek,
    Text
)
from .sniff import build_sniff_from_prefix
from .xml import GenericXml

log = logging.getLogger(__name__)


[docs]@build_sniff_from_prefix class BlastXml(GenericXml): """NCBI Blast XML Output data""" file_ext = "blastxml" edam_format = "format_3331" edam_data = "data_0857"
[docs] def set_peek(self, dataset, is_multi_byte=False): """Set the peek and blurb text""" if not dataset.dataset.purged: dataset.peek = get_file_peek(dataset.file_name) dataset.blurb = 'NCBI Blast XML data' else: dataset.peek = 'file does not exist' dataset.blurb = 'file purged from disk'
[docs] def sniff_prefix(self, file_prefix): """Determines whether the file is blastxml >>> from galaxy.datatypes.sniff import get_test_fname >>> fname = get_test_fname('megablast_xml_parser_test1.blastxml') >>> BlastXml().sniff(fname) True >>> fname = get_test_fname('tblastn_four_human_vs_rhodopsin.blastxml') >>> BlastXml().sniff(fname) True >>> fname = get_test_fname('interval.interval') >>> BlastXml().sniff(fname) False """ handle = file_prefix.string_io() line = handle.readline() if line.strip() != '<?xml version="1.0"?>': return False line = handle.readline() if line.strip() not in ['<!DOCTYPE BlastOutput PUBLIC "-//NCBI//NCBI BlastOutput/EN" "http://www.ncbi.nlm.nih.gov/dtd/NCBI_BlastOutput.dtd">', '<!DOCTYPE BlastOutput PUBLIC "-//NCBI//NCBI BlastOutput/EN" "NCBI_BlastOutput.dtd">']: return False line = handle.readline() if line.strip() != '<BlastOutput>': return False return True
[docs] def merge(split_files, output_file): """Merging multiple XML files is non-trivial and must be done in subclasses.""" if len(split_files) == 1: # For one file only, use base class method (move/copy) return Text.merge(split_files, output_file) if not split_files: raise ValueError("Given no BLAST XML files, %r, to merge into %s" % (split_files, output_file)) with open(output_file, "w") as out: h = None old_header = None for f in split_files: if not os.path.isfile(f): log.warning("BLAST XML file %s missing, retry in 1s..." % f) sleep(1) if not os.path.isfile(f): log.error("BLAST XML file %s missing" % f) raise ValueError("BLAST XML file %s missing" % f) h = open(f) header = h.readline() if not header: h.close() # Retry, could be transient error with networked file system... log.warning("BLAST XML file %s empty, retry in 1s..." % f) sleep(1) h = open(f) header = h.readline() if not header: log.error("BLAST XML file %s was empty" % f) raise ValueError("BLAST XML file %s was empty" % f) if header.strip() != '<?xml version="1.0"?>': out.write(header) # for diagnosis h.close() raise ValueError("%s is not an XML file!" % f) line = h.readline() header += line if line.strip() not in ['<!DOCTYPE BlastOutput PUBLIC "-//NCBI//NCBI BlastOutput/EN" "http://www.ncbi.nlm.nih.gov/dtd/NCBI_BlastOutput.dtd">', '<!DOCTYPE BlastOutput PUBLIC "-//NCBI//NCBI BlastOutput/EN" "NCBI_BlastOutput.dtd">']: out.write(header) # for diagnosis h.close() raise ValueError("%s is not a BLAST XML file!" % f) while True: line = h.readline() if not line: out.write(header) # for diagnosis h.close() raise ValueError("BLAST XML file %s ended prematurely" % f) header += line if "<Iteration>" in line: break if len(header) > 10000: # Something has gone wrong, don't load too much into memory! # Write what we have to the merged file for diagnostics out.write(header) h.close() raise ValueError("The header in BLAST XML file %s is too long" % f) if "<BlastOutput>" not in header: h.close() raise ValueError("%s is not a BLAST XML file:\n%s\n..." % (f, header)) if f == split_files[0]: out.write(header) old_header = header elif old_header is not None and old_header[:300] != header[:300]: # Enough to check <BlastOutput_program> and <BlastOutput_version> match h.close() raise ValueError("BLAST XML headers don't match for %s and %s - have:\n%s\n...\n\nAnd:\n%s\n...\n" % (split_files[0], f, old_header[:300], header[:300])) else: out.write(" <Iteration>\n") for line in h: if "</BlastOutput_iterations>" in line: break # TODO - Increment <Iteration_iter-num> and if required automatic query names # like <Iteration_query-ID>Query_3</Iteration_query-ID> to be increasing? out.write(line) h.close() out.write(" </BlastOutput_iterations>\n") out.write("</BlastOutput>\n")
merge = staticmethod(merge)
class _BlastDb(Data): """Base class for BLAST database datatype.""" def set_peek(self, dataset, is_multi_byte=False): """Set the peek and blurb text.""" if not dataset.dataset.purged: dataset.peek = "BLAST database (multiple files)" dataset.blurb = "BLAST database (multiple files)" else: dataset.peek = 'file does not exist' dataset.blurb = 'file purged from disk' def display_peek(self, dataset): """Create HTML content, used for displaying peek.""" try: return dataset.peek except Exception: return "BLAST database (multiple files)" def display_data(self, trans, data, preview=False, filename=None, to_ext=None, size=None, offset=None, **kwd): """ If preview is `True` allows us to format the data shown in the central pane via the "eye" icon. If preview is `False` triggers download. """ if not preview: return super(_BlastDb, self).display_data(trans, data=data, preview=preview, filename=filename, to_ext=to_ext, size=size, offset=offset, **kwd) if self.file_ext == "blastdbn": title = "This is a nucleotide BLAST database" elif self.file_ext == "blastdbp": title = "This is a protein BLAST database" elif self.file_ext == "blastdbd": title = "This is a domain BLAST database" else: # Error? title = "This is a BLAST database." msg = "" try: # Try to use any text recorded in the dummy index file: with io.open(data.file_name, "rU", encoding='utf-8') as handle: msg = handle.read().strip() except Exception: pass if not msg: msg = title # Galaxy assumes HTML for the display of composite datatypes, return smart_str("<html><head><title>%s</title></head><body><pre>%s</pre></body></html>" % (title, msg)) def merge(split_files, output_file): """Merge BLAST databases (not implemented for now).""" raise NotImplementedError("Merging BLAST databases is non-trivial (do this via makeblastdb?)") def split(cls, input_datasets, subdir_generator_function, split_params): """Split a BLAST database (not implemented for now).""" if split_params is None: return None raise NotImplementedError("Can't split BLAST databases")
[docs]class BlastNucDb(_BlastDb, Data): """Class for nucleotide BLAST database files.""" file_ext = 'blastdbn' allow_datatype_change = False composite_type = 'basic'
[docs] def __init__(self, **kwd): Data.__init__(self, **kwd) self.add_composite_file('blastdb.nhr', is_binary=True) # sequence headers self.add_composite_file('blastdb.nin', is_binary=True) # index file self.add_composite_file('blastdb.nsq', is_binary=True) # nucleotide sequences self.add_composite_file('blastdb.nal', is_binary=False, optional=True) # alias ( -gi_mask option of makeblastdb) self.add_composite_file('blastdb.nhd', is_binary=True, optional=True) # sorted sequence hash values ( -hash_index option of makeblastdb) self.add_composite_file('blastdb.nhi', is_binary=True, optional=True) # index of sequence hash values ( -hash_index option of makeblastdb) self.add_composite_file('blastdb.nnd', is_binary=True, optional=True) # sorted GI values ( -parse_seqids option of makeblastdb and gi present in the description lines) self.add_composite_file('blastdb.nni', is_binary=True, optional=True) # index of GI values ( -parse_seqids option of makeblastdb and gi present in the description lines) self.add_composite_file('blastdb.nog', is_binary=True, optional=True) # OID->GI lookup file ( -hash_index or -parse_seqids option of makeblastdb) self.add_composite_file('blastdb.nsd', is_binary=True, optional=True) # sorted sequence accession values ( -hash_index or -parse_seqids option of makeblastdb) self.add_composite_file('blastdb.nsi', is_binary=True, optional=True) # index of sequence accession values ( -hash_index or -parse_seqids option of makeblastdb) # self.add_composite_file('blastdb.00.idx', is_binary=True, optional=True) # first volume of the MegaBLAST index generated by makembindex # The previous line should be repeated for each index volume, with filename extensions like '.01.idx', '.02.idx', etc. self.add_composite_file('blastdb.shd', is_binary=True, optional=True) # MegaBLAST index superheader (-old_style_index false option of makembindex)
# self.add_composite_file('blastdb.naa', is_binary=True, optional=True) # index of a WriteDB column for e.g. mask data # self.add_composite_file('blastdb.nab', is_binary=True, optional=True) # data of a WriteDB column # self.add_composite_file('blastdb.nac', is_binary=True, optional=True) # multiple byte order for a WriteDB column # The previous 3 lines should be repeated for each WriteDB column, with filename extensions like ('.nba', '.nbb', '.nbc'), ('.nca', '.ncb', '.ncc'), etc.
[docs]class BlastProtDb(_BlastDb, Data): """Class for protein BLAST database files.""" file_ext = 'blastdbp' allow_datatype_change = False composite_type = 'basic'
[docs] def __init__(self, **kwd): Data.__init__(self, **kwd) # Component file comments are as in BlastNucDb except where noted self.add_composite_file('blastdb.phr', is_binary=True) self.add_composite_file('blastdb.pin', is_binary=True) self.add_composite_file('blastdb.psq', is_binary=True) # protein sequences self.add_composite_file('blastdb.phd', is_binary=True, optional=True) self.add_composite_file('blastdb.phi', is_binary=True, optional=True) self.add_composite_file('blastdb.pnd', is_binary=True, optional=True) self.add_composite_file('blastdb.pni', is_binary=True, optional=True) self.add_composite_file('blastdb.pog', is_binary=True, optional=True) self.add_composite_file('blastdb.psd', is_binary=True, optional=True) self.add_composite_file('blastdb.psi', is_binary=True, optional=True)
# self.add_composite_file('blastdb.paa', is_binary=True, optional=True) # self.add_composite_file('blastdb.pab', is_binary=True, optional=True) # self.add_composite_file('blastdb.pac', is_binary=True, optional=True) # The last 3 lines should be repeated for each WriteDB column, with filename extensions like ('.pba', '.pbb', '.pbc'), ('.pca', '.pcb', '.pcc'), etc.
[docs]class BlastDomainDb(_BlastDb, Data): """Class for domain BLAST database files.""" file_ext = 'blastdbd' allow_datatype_change = False composite_type = 'basic'
[docs] def __init__(self, **kwd): Data.__init__(self, **kwd) self.add_composite_file('blastdb.phr', is_binary=True) self.add_composite_file('blastdb.pin', is_binary=True) self.add_composite_file('blastdb.psq', is_binary=True) self.add_composite_file('blastdb.freq', is_binary=True, optional=True) self.add_composite_file('blastdb.loo', is_binary=True, optional=True) self.add_composite_file('blastdb.psd', is_binary=True, optional=True) self.add_composite_file('blastdb.psi', is_binary=True, optional=True) self.add_composite_file('blastdb.rps', is_binary=True, optional=True) self.add_composite_file('blastdb.aux', is_binary=True, optional=True)