#!/usr/bin/perl

=head1 NAME
                                                                                
dgfasplit.pl 

=head1 DESCRIPTION

Generates biosequence fasta data subsets from arguments.  Used
for data grid parallelization in Grid/cluster computing.

=head1 SYNOPSIS

 dgfasplit.pl -[npart|count] 10  
  [-infasta] infile
  [-indexfasta infile [-queryfasta '+docid:modSC'] ] 
  [-splitfasta infile]
  [-JAVALIB  path     $BLAST_LOCATION/lib/java]

 dgfasplit.pl -count 10  -in $bg/shared/prots/protfa2/modSC578.fa

=head1 AUTHOR
  
  Don Gilbert, gilbertd@indiana.edu, 2005/2006.

=cut


use strict;
use Getopt::Long;

my %opts;
my $ckopt = GetOptions( \%opts, 
  'count|npart=n', 
  'infasta=s', 
  'queryfa=s', 
  'JAVALIB=s',  
  'debug', 
  );

my $usage = 
"\nUsage: dgfasplit.pl -count 10 input.fasta
"; 

die "$usage" unless($ckopt);

my $error;
my $JAVALIB= 'lib/java';
my $JAVALIB_URL= 'ftp://iubio.bio.indiana.edu/biogrid/gridserver/lucegene/lib/java/';
my $URLGETDIR= "wget -nv -r -N -nH -nd --directory-prefix=";
my $URLCOPY= "globus-url-copy -dbg -vb"; ##  -dbg -vb 

my $TMPDIR='$PWD'; ### '$TG_NODE_SCRATCH'; # use TG_NODE_SCRATCH or in/out dir ??
my $JAVA= "java"; # ($ENV{JAVA_HOME}) ? "$ENV{JAVA_HOME}/bin/java" : "java";
my $GREP= "grep"; #  dont really need

my $debug = $opts{debug};
my $count  = $opts{count} or die "missing arg: -count=#parts $usage";
my $sessionid  = $opts{session};
my $queryfa     = $opts{queryfa};

$JAVALIB= $opts{'JAVALIB'} || $JAVALIB;
$TMPDIR = $opts{'TMPDIR'}  || $TMPDIR;
$URLCOPY= $opts{'URLCOPY'} || $URLCOPY;
chomp(my $host=`hostname`); # globus-hostname ?

# $error= indexFasta($opts{'infasta'}) if($defaultrun || $opts{'indexfasta'}); # force or do if needed?

## problem with all NNNN fasta parts -- blast segfaults ?? is this blast err or not?

my @dataparts= getQueryParts($count, $opts{'infasta'}, $sessionid);

print join("\n", @dataparts),"\n";
#exit;

#-------------------

sub fetchJavaLib {
  my($javalib)= @_;
  $javalib ||= evalEnv( $JAVALIB);
  unless(-d $javalib) {
    (my $topdir = $javalib) =~ s,/lib/java$,,;
    unless(-d $topdir) { die "Error: missing java lib topdir: $topdir"; }
    my $fullpath= mkdirPath("lib/java",$topdir);
    #mkdir("$topdir/lib"); mkdir("$topdir/lib/java");
    }
  my $url= evalEnv( $JAVALIB_URL);
  my $cmd="$URLGETDIR$javalib $url";
  ##warn "# fetchJavaLib: ",$cmd,"\n" if($debug);
  my $err= callSystem($cmd);
  return $err;
}

sub getJavaCp {
  my($fetchIfNeeded)= @_;
  my $javalib= evalEnv( $JAVALIB);
  my $cp="${javalib}/lusimple.jar";
  unless(-f $cp) {
    my $err= fetchJavaLib() if($fetchIfNeeded);
    die "Error: Missing javalib $cp" unless(-f $cp);
    }
  $cp .=":${javalib}/lucene.jar";
  return $cp; 
}

=item indexFasta($infile)

  create lucene index of fasta:
   java LuceneIndexerFasta -create  -index $infile.luindex $infile
  
=cut

sub indexFasta {
  my($infile)= @_;

  warn "# Error: missing fasta $infile" and return -1 unless (-f $infile);
  warn "# Warning: index exists: $infile.luindex" and return -1 
    if (-d "$infile.luindex");
  my $javalib= getJavaCp(1);
  my $deb=($debug) ? "-debug" : "";
  my $cmd="$JAVA -cp ${javalib} "
    ." LuceneIndexerFasta $deb -create -index $infile.luindex $infile";
  my $err= callSystem($cmd);  
  warn "# indexFasta() = $err\n" if($err || $debug);
  return $err;
}

=item getFastaPart($infile,$partfile,$nparts,$partnum)

  create part file with lucene indexed fasta:
  ( echo 'set parts $nparts' ; echo 'part $partnum' ) |
   java LuceneGetter -index $infile.luindex -out $partfile
  
=cut

sub getFastaPart {
  my($infile,$partfile,$nparts,$partnum)= @_;
  my $javalib= getJavaCp();
  my $deb=($debug) ? "-debug" : "";
  ## need option for query here: echo 'find $queryfa '
  my $fq=($queryfa) ? "echo 'find $queryfa' ;" : "";
  my $cmd="( echo 'set parts $nparts' ; $fq echo 'part $partnum' ) "
    ."| $JAVA -cp ${javalib} "
    ." LuceneGetter $deb -index $infile.luindex -out $partfile"; 
  my $err= callSystem($cmd);
  if($debug || $err) {
    my $ngot= `$GREP -c '^>' $partfile`; chomp($ngot);
    warn "# getFastaPart($partnum:$nparts,$queryfa,$infile)  got:$ngot err:$err\n";
    warn "#  -> $partfile\n";
    }
  return $err;
}

sub getQueryParts {
  my($nparts, $infile, $sessionid)= @_;
  my @parts=();
  warn "# Error: missing fasta $infile" and return @parts unless (-f $infile);
  my $dig= length($nparts);
  # my $tmpdir= getTMPDIR($sessionid);
  my $partfile= "$infile-"; ## "$tmpdir/$partfile-";
  indexFasta($infile) unless (-d "$infile.luindex");
  
  foreach my $partnum (1..$nparts) {
    my $ipart= $partfile . sprintf("%0".$dig."d",$partnum);  
    push(@parts, $ipart);
    my $err= getFastaPart($infile,$ipart,$nparts,$partnum)
      unless(-f $ipart);
    ## die if ($err);
    }
  return @parts;
}


sub evalEnv {
  my($s,$env)= @_;
  my %env= (defined $env && ref($env) =~ /HASH/) ? %$env : %ENV;
  $s =~ s/\$\{?(\w+)\}?/$env{$1}/g; 
  return $s;
}

sub callSystem {
  my(@args)= @_;
  warn("# DEBUG $host: ",join(" ",@args),"\n") if($debug);
  return system(@args);
}

sub mkdirPath {
  my($dir,$topdir)= @_;
  $topdir= "/" unless($topdir);
  $dir =~ s,^$topdir,,;
  $dir =~ s,^/,,; $dir =~ s,/$,,;
  $topdir =~ s,/$,,;
  my @dirpath= split("/",$dir);
  foreach my $d (@dirpath) {
    $topdir = "$topdir/$d";
    my $ok= (-d $topdir) ? 1 : mkdir($topdir); 
    warn "Error: mkdir $topdir" and return 0 unless($ok);
    }
  return $topdir;
}


1;

__END__