Changeset 2012

Timestamp:

2001-02-19T18:06:16+13:00 (23 years ago)

Author:

jrm21

Message:

re-added the crappy PS text-stripper, and made the error handling
a bit more robust.

File:

• trunk/gsdl/bin/script/gsConvert.pl (modified) (2 diffs)

trunk/gsdl/bin/script/gsConvert.pl

@@ -276 +276 @@
 # Find the real type of a .doc file
 #
-# We seem to have alot of files with a .dco extension that are .rtf 
+# We seem to have a lot of files with a .doc extension that are .rtf 
 # files or Word 5 files.  This function attempts to tell the difference.
 
@@ -487 +487 @@
 }
 
-# Convert a PostScript document to text with ps2ascii
+# Convert a PostScript document to text
+# note - just using "ps2ascii" isn't good enough, as it
+# returns 0 for a postscript interpreter error. ps2ascii is just
+# a wrapper to "gs" anyway, so we use that cmd here.
 
 sub ps_to_text {
     ($input_filename, $output_filestem) = @_;
 
-    my $cmd = "ps2ascii \"$input_filename\" > \"$output_filestem.text\"";
+    my $cmd = "gs -q -dNODISPLAY -dNOBIND -dWRITESYSTEMDICT -dSIMPLE -c save ";
+    $cmd .= "-f ps2ascii \"$input_filename\" -c quit > \"$output_filestem.text\"";
     $cmd .= " 2> $output_filestem.err";
-    
-    if (system($cmd)>0)
-    {
-    print STDERR "Error executing $cmd: $!\n";
+    $!=0;
+    my $retcode=system($cmd);
+    $retcode = $? >> 8;  # see man perlfunc - system for this...
+    # if system returns -1 | 127 (couldn't start program), look at $! for message
+    my $error="";
+    if ($retcode!=0) {if ($!) {$error=$!;} else {$error="couldn't run.\n";}}
+    elsif (! -e "$output_filestem.text") {
+    $error="did not create output file.\n";
+    }
+    else 
+    { # make sure the interpreter didn't get an error. It is technically
+    # possible for the actual text to start with this, but....
+    open PSOUT, "$output_filestem.text";
+    if (<PSOUT> =~ /^Error: (.*)/) {
+        $error="interpreter error - \"$1\"";
+    }
+    close PSOUT;
+    }
+#    print STDERR "retcode=$retcode, error=\"$!\"\n";
+    if ($error ne "")
+    {
+    print STDERR "PSPLUG: WARNING: Error executing gs: $error\n";
     &util::rm("$output_filestem.text") if (-e "$output_filestem.text");
     &util::rm("$output_filestem.err") if (-e "$output_filestem.err");
-    return 0;
+
+    # Fine then. We'll just do a lousy job by ourselves...
+    # Based on regexps nicked from:
+    # http://snark.ptc.spbu.ru/mail-archives/lout/brown/msg00003.html
+    # 
+    print STDERR "PSPlug: Stripping text from postscript\n";
+    my $errorcode=0;
+    open (IN, "$input_filename") 
+        ||  ($errorcode=1, warn "Couldn't read file: $!");
+    open (OUT, ">$output_filestem.text") 
+        ||  ($errorcode=1, warn "Couldn't write file: $!");
+    if ($errorcode) {print STDERR "errors\n";return 0;}
+    
+    my $in_a_sentence=0;
+    while (<IN>) {
+        if (/^[^\(\)]+$/ && !$in_a_sentence) {next ;} # no brackets in line
+        # attempt to add whitespace between different lines...
+        s/F.?\(/\( /g; # this might break up some other words though...
+        ### remove all postscript control data
+        if (!$in_a_sentence) {
+        s/^[^\(\)]*?\(//;  # rm start of line up to first open bracket
+        }
+        s/\\\(/\{/g;s/\\\)/\}/g ; # change quoted braces
+        s/\)([^\(\)])*?\(//g ;   # close bracket up to next open unquoted bracket
+        if (s/\)[^\(\)]*?$//g)  # last close bracket to end of line
+        {$in_a_sentence=0;chomp;}     
+        if (s/\\$//) # if line is a continuation
+        {$in_a_sentence=1;chomp;} 
+        s/^$//g ;             # remove empty lines
+        ### ligatures have special characters...
+        s/\\214/fi/g;
+        s/\\215/fl/g;
+        print OUT "$_";
+    }
+    close IN; close OUT;
     }