[2] | 1 | #!/usr/bin/env python |
---|
| 2 | # Greg Von Kuster |
---|
| 3 | |
---|
| 4 | """ |
---|
| 5 | Subtract an entire query from another query |
---|
| 6 | usage: %prog in_file_1 in_file_2 begin_col end_col output |
---|
| 7 | """ |
---|
| 8 | import sys, re |
---|
| 9 | from galaxy import eggs |
---|
| 10 | import pkg_resources; pkg_resources.require( "bx-python" ) |
---|
| 11 | from bx.cookbook import doc_optparse |
---|
| 12 | |
---|
| 13 | # Older py compatibility |
---|
| 14 | try: |
---|
| 15 | set() |
---|
| 16 | except: |
---|
| 17 | from sets import Set as set |
---|
| 18 | |
---|
| 19 | assert sys.version_info[:2] >= ( 2, 4 ) |
---|
| 20 | |
---|
| 21 | def get_lines(fname, begin_col='', end_col=''): |
---|
| 22 | lines = set([]) |
---|
| 23 | i = 0 |
---|
| 24 | for i, line in enumerate(file(fname)): |
---|
| 25 | line = line.rstrip('\r\n') |
---|
| 26 | if line and not line.startswith('#'): |
---|
| 27 | if begin_col and end_col: |
---|
| 28 | """Both begin_col and end_col must be integers at this point.""" |
---|
| 29 | try: |
---|
| 30 | line = line.split('\t') |
---|
| 31 | line = '\t'.join([line[j] for j in range(begin_col-1, end_col)]) |
---|
| 32 | lines.add( line ) |
---|
| 33 | except: pass |
---|
| 34 | else: |
---|
| 35 | lines.add( line ) |
---|
| 36 | if i: return (i+1, lines) |
---|
| 37 | else: return (i, lines) |
---|
| 38 | |
---|
| 39 | def main(): |
---|
| 40 | |
---|
| 41 | # Parsing Command Line here |
---|
| 42 | options, args = doc_optparse.parse( __doc__ ) |
---|
| 43 | |
---|
| 44 | try: |
---|
| 45 | inp1_file, inp2_file, begin_col, end_col, out_file = args |
---|
| 46 | except: |
---|
| 47 | doc_optparse.exception() |
---|
| 48 | |
---|
| 49 | begin_col = begin_col.strip() |
---|
| 50 | end_col = end_col.strip() |
---|
| 51 | |
---|
| 52 | if begin_col != 'None' or end_col != 'None': |
---|
| 53 | """ |
---|
| 54 | The user selected columns for restriction. We'll allow default |
---|
| 55 | values for both begin_col and end_col as long as the user selected |
---|
| 56 | at least one of them for restriction. |
---|
| 57 | """ |
---|
| 58 | if begin_col == 'None': |
---|
| 59 | begin_col = end_col |
---|
| 60 | elif end_col == 'None': |
---|
| 61 | end_col = begin_col |
---|
| 62 | begin_col = int(begin_col) |
---|
| 63 | end_col = int(end_col) |
---|
| 64 | """Make sure that begin_col <= end_col (switch if not)""" |
---|
| 65 | if begin_col > end_col: |
---|
| 66 | tmp_col = end_col |
---|
| 67 | end_col = begin_col |
---|
| 68 | begin_col = tmp_col |
---|
| 69 | else: |
---|
| 70 | begin_col = end_col = '' |
---|
| 71 | |
---|
| 72 | try: |
---|
| 73 | fo = open(out_file,'w') |
---|
| 74 | except: |
---|
| 75 | print >> sys.stderr, "Unable to open output file" |
---|
| 76 | sys.exit() |
---|
| 77 | |
---|
| 78 | """ |
---|
| 79 | len1 is the number of lines in inp1_file |
---|
| 80 | lines1 is the set of unique lines in inp1_file |
---|
| 81 | diff1 is the number of duplicate lines removed from inp1_file |
---|
| 82 | """ |
---|
| 83 | len1, lines1 = get_lines(inp1_file, begin_col, end_col) |
---|
| 84 | diff1 = len1 - len(lines1) |
---|
| 85 | len2, lines2 = get_lines(inp2_file, begin_col, end_col) |
---|
| 86 | |
---|
| 87 | lines1.difference_update(lines2) |
---|
| 88 | """lines1 is now the set of unique lines in inp1_file - the set of unique lines in inp2_file""" |
---|
| 89 | |
---|
| 90 | for line in lines1: |
---|
| 91 | print >> fo, line |
---|
| 92 | |
---|
| 93 | fo.close() |
---|
| 94 | |
---|
| 95 | info_msg = 'Subtracted %d lines. ' %((len1 - diff1) - len(lines1)) |
---|
| 96 | |
---|
| 97 | if begin_col and end_col: |
---|
| 98 | info_msg += 'Restricted to columns c' + str(begin_col) + ' thru c' + str(end_col) + '. ' |
---|
| 99 | |
---|
| 100 | if diff1 > 0: |
---|
| 101 | info_msg += 'Eliminated %d duplicate/blank/comment/invalid lines from first query.' %diff1 |
---|
| 102 | |
---|
| 103 | print info_msg |
---|
| 104 | |
---|
| 105 | if __name__ == "__main__": |
---|
| 106 | main() |
---|