annotate filtering.py @ 22:2615b9ede2c4

Uploaded 1filtering.tar
author greg
date Thu, 04 Aug 2011 13:56:33 -0400
parents 4f07f3a33605
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
4f07f3a33605 Uploaded
greg
parents:
diff changeset
1 #!/usr/bin/env python
4f07f3a33605 Uploaded
greg
parents:
diff changeset
2 # This tool takes a tab-delimited text file as input and creates filters on columns based on certain properties.
4f07f3a33605 Uploaded
greg
parents:
diff changeset
3 # The tool will skip over invalid lines within the file, informing the user about the number of lines skipped.
4f07f3a33605 Uploaded
greg
parents:
diff changeset
4
4f07f3a33605 Uploaded
greg
parents:
diff changeset
5 from __future__ import division
4f07f3a33605 Uploaded
greg
parents:
diff changeset
6 import sys, re, os.path
4f07f3a33605 Uploaded
greg
parents:
diff changeset
7 from galaxy import eggs
4f07f3a33605 Uploaded
greg
parents:
diff changeset
8
4f07f3a33605 Uploaded
greg
parents:
diff changeset
9 # Older py compatibility
4f07f3a33605 Uploaded
greg
parents:
diff changeset
10 try:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
11 set()
4f07f3a33605 Uploaded
greg
parents:
diff changeset
12 except:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
13 from sets import Set as set
4f07f3a33605 Uploaded
greg
parents:
diff changeset
14
4f07f3a33605 Uploaded
greg
parents:
diff changeset
15 assert sys.version_info[:2] >= ( 2, 4 )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
16
4f07f3a33605 Uploaded
greg
parents:
diff changeset
17 def get_operands( filter_condition ):
4f07f3a33605 Uploaded
greg
parents:
diff changeset
18 # Note that the order of all_operators is important
4f07f3a33605 Uploaded
greg
parents:
diff changeset
19 items_to_strip = ['+', '-', '**', '*', '//', '/', '%', '<<', '>>', '&', '|', '^', '~', '<=', '<', '>=', '>', '==', '!=', '<>', ' and ', ' or ', ' not ', ' is ', ' is not ', ' in ', ' not in ']
4f07f3a33605 Uploaded
greg
parents:
diff changeset
20 for item in items_to_strip:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
21 if filter_condition.find( item ) >= 0:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
22 filter_condition = filter_condition.replace( item, ' ' )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
23 operands = set( filter_condition.split( ' ' ) )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
24 return operands
4f07f3a33605 Uploaded
greg
parents:
diff changeset
25
4f07f3a33605 Uploaded
greg
parents:
diff changeset
26 def stop_err( msg ):
4f07f3a33605 Uploaded
greg
parents:
diff changeset
27 sys.stderr.write( msg )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
28 sys.exit()
4f07f3a33605 Uploaded
greg
parents:
diff changeset
29
4f07f3a33605 Uploaded
greg
parents:
diff changeset
30 in_fname = sys.argv[1]
4f07f3a33605 Uploaded
greg
parents:
diff changeset
31 out_fname = sys.argv[2]
4f07f3a33605 Uploaded
greg
parents:
diff changeset
32 cond_text = sys.argv[3]
4f07f3a33605 Uploaded
greg
parents:
diff changeset
33 try:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
34 in_columns = int( sys.argv[4] )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
35 assert sys.argv[5] #check to see that the column types varaible isn't null
4f07f3a33605 Uploaded
greg
parents:
diff changeset
36 in_column_types = sys.argv[5].split( ',' )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
37 except:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
38 stop_err( "Data does not appear to be tabular. This tool can only be used with tab-delimited data." )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
39
4f07f3a33605 Uploaded
greg
parents:
diff changeset
40 # Unescape if input has been escaped
4f07f3a33605 Uploaded
greg
parents:
diff changeset
41 mapped_str = {
4f07f3a33605 Uploaded
greg
parents:
diff changeset
42 '__lt__': '<',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
43 '__le__': '<=',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
44 '__eq__': '==',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
45 '__ne__': '!=',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
46 '__gt__': '>',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
47 '__ge__': '>=',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
48 '__sq__': '\'',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
49 '__dq__': '"',
4f07f3a33605 Uploaded
greg
parents:
diff changeset
50 }
4f07f3a33605 Uploaded
greg
parents:
diff changeset
51 for key, value in mapped_str.items():
4f07f3a33605 Uploaded
greg
parents:
diff changeset
52 cond_text = cond_text.replace( key, value )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
53
4f07f3a33605 Uploaded
greg
parents:
diff changeset
54 # Attempt to determine if the condition includes executable stuff and, if so, exit
4f07f3a33605 Uploaded
greg
parents:
diff changeset
55 secured = dir()
4f07f3a33605 Uploaded
greg
parents:
diff changeset
56 operands = get_operands(cond_text)
4f07f3a33605 Uploaded
greg
parents:
diff changeset
57 for operand in operands:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
58 try:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
59 check = int( operand )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
60 except:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
61 if operand in secured:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
62 stop_err( "Illegal value '%s' in condition '%s'" % ( operand, cond_text ) )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
63
4f07f3a33605 Uploaded
greg
parents:
diff changeset
64 # Prepare the column variable names and wrappers for column data types
4f07f3a33605 Uploaded
greg
parents:
diff changeset
65 cols, type_casts = [], []
4f07f3a33605 Uploaded
greg
parents:
diff changeset
66 for col in range( 1, in_columns + 1 ):
4f07f3a33605 Uploaded
greg
parents:
diff changeset
67 col_name = "c%d" % col
4f07f3a33605 Uploaded
greg
parents:
diff changeset
68 cols.append( col_name )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
69 col_type = in_column_types[ col - 1 ]
4f07f3a33605 Uploaded
greg
parents:
diff changeset
70 type_cast = "%s(%s)" % ( col_type, col_name )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
71 type_casts.append( type_cast )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
72
4f07f3a33605 Uploaded
greg
parents:
diff changeset
73 col_str = ', '.join( cols ) # 'c1, c2, c3, c4'
4f07f3a33605 Uploaded
greg
parents:
diff changeset
74 type_cast_str = ', '.join( type_casts ) # 'str(c1), int(c2), int(c3), str(c4)'
4f07f3a33605 Uploaded
greg
parents:
diff changeset
75 assign = "%s = line.split( '\\t' )" % col_str
4f07f3a33605 Uploaded
greg
parents:
diff changeset
76 wrap = "%s = %s" % ( col_str, type_cast_str )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
77 skipped_lines = 0
4f07f3a33605 Uploaded
greg
parents:
diff changeset
78 first_invalid_line = 0
4f07f3a33605 Uploaded
greg
parents:
diff changeset
79 invalid_line = None
4f07f3a33605 Uploaded
greg
parents:
diff changeset
80 lines_kept = 0
4f07f3a33605 Uploaded
greg
parents:
diff changeset
81 total_lines = 0
4f07f3a33605 Uploaded
greg
parents:
diff changeset
82 out = open( out_fname, 'wt' )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
83
4f07f3a33605 Uploaded
greg
parents:
diff changeset
84 # Read and filter input file, skipping invalid lines
4f07f3a33605 Uploaded
greg
parents:
diff changeset
85 code = '''
4f07f3a33605 Uploaded
greg
parents:
diff changeset
86 for i, line in enumerate( file( in_fname ) ):
4f07f3a33605 Uploaded
greg
parents:
diff changeset
87 total_lines += 1
4f07f3a33605 Uploaded
greg
parents:
diff changeset
88 line = line.rstrip( '\\r\\n' )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
89 if not line or line.startswith( '#' ):
4f07f3a33605 Uploaded
greg
parents:
diff changeset
90 skipped_lines += 1
4f07f3a33605 Uploaded
greg
parents:
diff changeset
91 if not invalid_line:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
92 first_invalid_line = i + 1
4f07f3a33605 Uploaded
greg
parents:
diff changeset
93 invalid_line = line
4f07f3a33605 Uploaded
greg
parents:
diff changeset
94 continue
4f07f3a33605 Uploaded
greg
parents:
diff changeset
95 try:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
96 %s
4f07f3a33605 Uploaded
greg
parents:
diff changeset
97 %s
4f07f3a33605 Uploaded
greg
parents:
diff changeset
98 if %s:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
99 lines_kept += 1
4f07f3a33605 Uploaded
greg
parents:
diff changeset
100 print >> out, line
4f07f3a33605 Uploaded
greg
parents:
diff changeset
101 except:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
102 skipped_lines += 1
4f07f3a33605 Uploaded
greg
parents:
diff changeset
103 if not invalid_line:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
104 first_invalid_line = i + 1
4f07f3a33605 Uploaded
greg
parents:
diff changeset
105 invalid_line = line
4f07f3a33605 Uploaded
greg
parents:
diff changeset
106 ''' % ( assign, wrap, cond_text )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
107
4f07f3a33605 Uploaded
greg
parents:
diff changeset
108 valid_filter = True
4f07f3a33605 Uploaded
greg
parents:
diff changeset
109 try:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
110 exec code
4f07f3a33605 Uploaded
greg
parents:
diff changeset
111 except Exception, e:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
112 out.close()
4f07f3a33605 Uploaded
greg
parents:
diff changeset
113 if str( e ).startswith( 'invalid syntax' ):
4f07f3a33605 Uploaded
greg
parents:
diff changeset
114 valid_filter = False
4f07f3a33605 Uploaded
greg
parents:
diff changeset
115 stop_err( 'Filter condition "%s" likely invalid. See tool tips, syntax and examples.' % cond_text )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
116 else:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
117 stop_err( str( e ) )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
118
4f07f3a33605 Uploaded
greg
parents:
diff changeset
119 if valid_filter:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
120 out.close()
4f07f3a33605 Uploaded
greg
parents:
diff changeset
121 valid_lines = total_lines - skipped_lines
4f07f3a33605 Uploaded
greg
parents:
diff changeset
122 print 'Filtering with %s, ' % cond_text
4f07f3a33605 Uploaded
greg
parents:
diff changeset
123 if valid_lines > 0:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
124 print 'kept %4.2f%% of %d lines.' % ( 100.0*lines_kept/valid_lines, total_lines )
4f07f3a33605 Uploaded
greg
parents:
diff changeset
125 else:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
126 print 'Possible invalid filter condition "%s" or non-existent column referenced. See tool tips, syntax and examples.' % cond_text
4f07f3a33605 Uploaded
greg
parents:
diff changeset
127 if skipped_lines > 0:
4f07f3a33605 Uploaded
greg
parents:
diff changeset
128 print 'Skipped %d invalid lines starting at line #%d: "%s"' % ( skipped_lines, first_invalid_line, invalid_line )