| 1 |
#!/usr/bin/perl |
| 2 |
use strict; |
| 3 |
use Getopt::Std; |
| 4 |
use FindBin;use lib $FindBin::Bin; |
| 5 |
|
| 6 |
my $usage = q/ |
| 7 |
Split a multi-fasta file in multiple smaller parts. |
| 8 |
Usage: |
| 9 |
fasplit {-n <seqs_per_part> | -s <partsize>} [-o <outname>] <multifasta.fa> |
| 10 |
|
| 11 |
Options: |
| 12 |
There are two modes of operation (mutually exclusive): |
| 13 |
-n split by number of sequences in a part |
| 14 |
-s split by part size; <partsize> is a numeric argument representing |
| 15 |
the desired part size in Megabytes |
| 16 |
|
| 17 |
-o specify the name prefix for the output files (parts);a 0-padded number |
| 18 |
and a .fa extension will be appended to <outname> for each part |
| 19 |
/; |
| 20 |
umask 0002; |
| 21 |
getopts('Nn:s:o:') || die($usage."\n"); |
| 22 |
my $infile=$ARGV[0] || die("No multifasta input file provided!\n"); |
| 23 |
my $outprefix=$Getopt::Std::opt_o; |
| 24 |
my $perSeq=$Getopt::Std::opt_N; |
| 25 |
unless ($outprefix) { |
| 26 |
$outprefix=$infile; |
| 27 |
$outprefix=~s/\.\w+$//; |
| 28 |
} |
| 29 |
my $partsize=$Getopt::Std::opt_s; |
| 30 |
my $numseqs=$Getopt::Std::opt_n; |
| 31 |
die("$usage\nOnly one of -n or -s options can be given!\n") |
| 32 |
if ($numseqs && $partsize); |
| 33 |
if ($perSeq) { $numseqs=1; $partsize=''; } |
| 34 |
my $bufProcess = $numseqs ? \&nBufProc : \&sBufProc; |
| 35 |
open(INFILE, $infile) || die("Input file $infile cannot be opened!"); |
| 36 |
$partsize=int($partsize*1024*1024); |
| 37 |
#my $rsize=0; #accumulated read size so far since the last file open |
| 38 |
my $rsize=0; |
| 39 |
my $wsize=0; #last written size |
| 40 |
#my $partbuf; |
| 41 |
my $scount=1; |
| 42 |
my $partnum=1; #current part number |
| 43 |
my $pfname=sprintf('%s%02d',$outprefix,$partnum).'.fa'; |
| 44 |
open(PFNAME, '>'.$pfname) || die ("Error creating file $pfname!\n"); |
| 45 |
while (<INFILE>) { |
| 46 |
tr/\n\r//d; |
| 47 |
next unless $_; |
| 48 |
if (m/^>/) { |
| 49 |
#starting record: |
| 50 |
#print STDERR "* starting record: $_\n"; |
| 51 |
&$bufProcess(); |
| 52 |
} |
| 53 |
$rsize+=length($_)+1; |
| 54 |
print PFNAME $_."\n"; |
| 55 |
|
| 56 |
} |
| 57 |
close(INFILE); |
| 58 |
#&$bufProcess(1); |
| 59 |
close(PFNAME); |
| 60 |
|
| 61 |
sub nBufProc { |
| 62 |
return unless $rsize>0; |
| 63 |
if ($scount>=$numseqs) { |
| 64 |
close(PFNAME); |
| 65 |
$partnum++; |
| 66 |
$pfname=sprintf('%s%02d',$outprefix,$partnum).'.fa'; |
| 67 |
#print STDERR "$scount > $numseqs : write to $pfname now.\n"; |
| 68 |
open(PFNAME, '>'.$pfname) || die ("Error creating file $pfname!\n"); |
| 69 |
$wsize=$rsize; |
| 70 |
$rsize=0; |
| 71 |
$scount=0; |
| 72 |
} |
| 73 |
$scount++; |
| 74 |
} |
| 75 |
|
| 76 |
sub sBufProc { |
| 77 |
return unless $rsize>0; |
| 78 |
if ($rsize>=$partsize) { |
| 79 |
#close previous part, open new one |
| 80 |
close(PFNAME); |
| 81 |
$partnum++; |
| 82 |
$pfname=sprintf('%s%02d',$outprefix,$partnum).'.fa'; |
| 83 |
open(PFNAME, '>'.$pfname) || die ("Error creating file $pfname!\n"); |
| 84 |
$wsize=$rsize; |
| 85 |
$rsize=0; |
| 86 |
$scount=0; |
| 87 |
} |
| 88 |
$scount++; |
| 89 |
} |
