Mercurial > repos > nick > duplex
annotate planemo-template/random_lines_two_pass.py @ 4:af383638de66 draft
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
author | nick |
---|---|
date | Mon, 23 Nov 2015 18:44:23 -0500 |
parents | |
children |
rev | line source |
---|---|
4
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
1 #!/usr/bin/env python |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
2 #Dan Blankenberg |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
3 #Selects N random lines from a file and outputs to another file, maintaining original line order |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
4 #allows specifying a seed |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
5 #does two passes to determine line offsets/count, and then to output contents |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
6 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
7 import optparse, random |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
8 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
9 def get_random_by_subtraction( line_offsets, num_lines ): |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
10 while len( line_offsets ) > num_lines: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
11 del line_offsets[ random.randint( 0, len( line_offsets ) - 1 ) ] |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
12 return line_offsets |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
13 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
14 def get_random_by_sample( line_offsets, num_lines ): |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
15 line_offsets = random.sample( line_offsets, num_lines ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
16 line_offsets.sort() |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
17 return line_offsets |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
18 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
19 def get_random( line_offsets, num_lines ): |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
20 if num_lines > ( len( line_offsets ) / 2 ): |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
21 return get_random_by_subtraction( line_offsets, num_lines ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
22 else: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
23 return get_random_by_sample( line_offsets, num_lines ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
24 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
25 def __main__(): |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
26 #Parse Command Line |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
27 parser = optparse.OptionParser() |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
28 parser.add_option( '-s', '--seed', dest='seed', action='store', type="string", default=None, help='Set the random seed.' ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
29 (options, args) = parser.parse_args() |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
30 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
31 assert len( args ) == 3, "Invalid command line specified." |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
32 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
33 input = open( args[0], 'rb' ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
34 output = open( args[1], 'wb' ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
35 num_lines = int( args[2] ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
36 assert num_lines > 0, "You must select at least one line." |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
37 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
38 if options.seed is not None: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
39 random.seed( options.seed ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
40 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
41 #get line offsets |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
42 line_offsets = [] |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
43 teller = input.tell |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
44 readliner = input.readline |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
45 appender = line_offsets.append |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
46 while True: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
47 offset = teller() |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
48 if readliner(): |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
49 appender( offset ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
50 else: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
51 break |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
52 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
53 total_lines = len( line_offsets ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
54 assert num_lines <= total_lines, "Error: asked to select more lines (%i) than there were in the file (%i)." % ( num_lines, total_lines ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
55 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
56 #get random line offsets |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
57 line_offsets = get_random( line_offsets, num_lines ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
58 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
59 #write out random lines |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
60 seeker = input.seek |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
61 writer = output.write |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
62 for line_offset in line_offsets: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
63 seeker( line_offset ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
64 writer( readliner() ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
65 input.close() |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
66 output.close() |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
67 print "Kept %i of %i total lines." % ( num_lines, total_lines ) |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
68 if options.seed is not None: |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
69 print 'Used random seed of "%s".' % options.seed |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
70 |
af383638de66
planemo upload commit 022984f323d3da44f70b3bf79c684cfd8dda3f61-dirty
nick
parents:
diff
changeset
|
71 if __name__=="__main__": __main__() |