Mercurial > repos > bgruening > openbabel_remove_protonation_state
view multi_obgrep.py @ 11:5dc2a6cd6da1 draft
"planemo upload for repository https://github.com/bgruening/galaxytools/tree/master/chemicaltoolbox/openbabel commit cda909c5e0b88fa3d12abe43fc72b8dd0729417a"
author | bgruening |
---|---|
date | Thu, 09 Apr 2020 10:11:55 -0400 |
parents | 0eabdfaef1d1 |
children | bf4e668b6690 |
line wrap: on
line source
#!/usr/bin/env python """ Input: Molecules in SDF, SMILES ... Output: Molecule file filtered with obgrep. Copyright 2013, Bjoern Gruening and Xavier Lucas """ import sys, os import argparse import openbabel openbabel.obErrorLog.StopLogging() import pybel import multiprocessing import tempfile import subprocess import shutil import shlex def parse_command_line(): parser = argparse.ArgumentParser() parser.add_argument('-i', '--infile', required=True, help='Molecule file.') parser.add_argument('-q', '--query', required=True, help='Query file, containing different SMARTS in each line.') parser.add_argument('-o', '--outfile', required=True, help='Path to the output file.') parser.add_argument("--iformat", help="Input format, like smi, sdf, inchi") parser.add_argument("--n-times", dest="n_times", type=int, default=0, help="Print a molecule only if the pattern occurs # times inside the molecule.") parser.add_argument('-p', '--processors', type=int, default=multiprocessing.cpu_count()) parser.add_argument("--invert-matches", dest="invert_matches", action="store_true", default=False, help="Invert the matching, print non-matching molecules.") parser.add_argument("--only-name", dest="only_name", action="store_true", default=False, help="Only print the name of the molecules.") parser.add_argument("--full-match", dest="full_match", action="store_true", default=False, help="Full match, print matching-molecules only when the number of heavy atoms is also equal to the number of atoms in the SMARTS pattern.") parser.add_argument("--number-of-matches", dest="number_of_matches", action="store_true", default=False, help="Print the number of matches.") return parser.parse_args() results = list() def mp_callback(res): results.append(res) def mp_helper( query, args ): """ Helper function for multiprocessing. That function is a wrapper around obgrep. """ cmd_list = [] if args.invert_matches: cmd_list.append('-v') if args.only_name: cmd_list.append('-n') if args.full_match: cmd_list.append('-f') if args.number_of_matches: cmd_list.append('-c') if args.n_times: cmd_list.append('-t %s' % str(args.n_times)) tmp = tempfile.NamedTemporaryFile(delete=False) cmd = 'obgrep %s "%s" %s' % (' '.join(cmd_list), query, args.infile) child = subprocess.Popen(shlex.split(cmd), stdout=open(tmp.name, 'w+'), stderr=subprocess.PIPE) stdout, stderr = child.communicate() return (tmp.name, query) def obgrep( args ): temp_file = tempfile.NamedTemporaryFile() temp_link = "%s.%s" % (temp_file.name, args.iformat) temp_file.close() os.symlink(args.infile, temp_link) args.infile = temp_link pool = multiprocessing.Pool( args.processors ) for query in open( args.query ): pool.apply_async(mp_helper, args=(query.strip(), args), callback=mp_callback) #mp_callback( mp_helper(query.strip(), args) ) pool.close() pool.join() out_handle = open( args.outfile, 'wb' ) for result_file, query in results: res_handle = open(result_file,'rb') shutil.copyfileobj( res_handle, out_handle ) res_handle.close() os.remove( result_file ) out_handle.close() os.remove( temp_link ) def __main__(): """ Multiprocessing obgrep search. """ args = parse_command_line() obgrep( args ) if __name__ == "__main__" : __main__()