Changeset 2031

Timestamp:

2001-02-20T18:54:46+13:00 (23 years ago)

Author:

jrm21

Message:

Improved postscript to text handling a little bit better.
Also, system($cmd) return value can be "-1", not just ">0"....

File:

• trunk/gsdl/bin/script/gsConvert.pl (modified) (5 diffs)

trunk/gsdl/bin/script/gsConvert.pl

@@ -418 +418 @@
     $cmd .= " \"$input_filename\" \"$output_filestem\"";
 
-    if (system($cmd)>0)
+    if (system($cmd)!=0)
     {
     print STDERR "Error executing $cmd: $!\n";
@@ -489 +489 @@
 
     my $cmd = "gs -q -dNODISPLAY -dNOBIND -dWRITESYSTEMDICT -dSIMPLE -c save ";
-    $cmd .= "-f ps2ascii \"$input_filename\" -c quit > \"$output_filestem.text\"";
+    $cmd .= "-f ps2ascii.ps \"$input_filename\" -c quit > \"$output_filestem.text\"";
     $cmd .= " 2> $output_filestem.err";
     $!=0;
@@ -509 +509 @@
     close PSOUT;
     }
-#    print STDERR "retcode=$retcode, error=\"$!\"\n";
     if ($error ne "")
     {
@@ -517 +516 @@
 
     # Fine then. We'll just do a lousy job by ourselves...
-    # Based on regexps nicked from:
+    # Based on 5-line regexp sed script found at:
     # http://snark.ptc.spbu.ru/mail-archives/lout/brown/msg00003.html
     # 
@@ -528 +527 @@
     if ($errorcode) {print STDERR "errors\n";return 0;}
     
-    my $in_a_sentence=0;
+    my $text="";  # this is for whole .ps file...
     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;
-    }
-
+        $text.=$_;
+    }
+    close IN;
+
+    # if ps has Page data, then use it to delete all stuff before it.
+    $text =~ s/^.*?%%Page:.*?\n//s; # treat string as single line
+    
+    # remove all leading non-data stuff
+    $text =~ s/^.*?\(//s;
+
+    # remove all newline chars for easier processing
+    $text =~ s/\n//g;
+    
+    # Big assumption here - assume that if any co-ordinates are
+    # given, then we are at the end of a sentence.
+    $text =~ s/\)-?\d+\ -?\d+/\) \(\n\)/g;
+
+    # special characters--
+    $text =~ s/\(\|\)/\(\ - \)/g; # j -> em-dash?
+
+    # ? ps text formatting (eg italics?) ?
+    $text =~ s/Fn\(f\)/\(\{\)/g; # f -> {
+    $text =~ s/Fn\(g\)/\(\}\)/g; # g -> }
+    $text =~ s/Fn\(j\)/\(\|\)/g; # j -> |
+    # default - remove the rest
+    $text =~ s/\ ?F.\((.+?)\)/\($1\)/g;
+
+    # attempt to add whitespace between words... 
+    # this is based purely on observation, and may be completely wrong...
+    $text =~ s/([^F])[defghijkuy]\(/$1 \( /g;
+    # eg I notice "b(" is sometimes NOT a space if preceded by a 
+    # negative number.
+    $text =~ s/\)\d+ ?b\(/\) \( /g;
+
+    # change quoted braces to brackets
+    $text =~ s/([^\\])\\\(/$1\{/g;
+    $text =~ s/([^\\])\\\)/$1\}/g ;
+
+    # remove everything that is not between braces
+    $text =~ s/\)([^\(\)])+?\(//sg ;
+    
+    # remove any Trailer eof stuff.
+    $text =~ s/\)[^\)]*$//sg;
+
+    ### ligatures have special characters...
+    $text =~ s/\\013/ff/g;
+    $text =~ s/\\014/fi/g;
+    $text =~ s/\\015/fl/g;
+    $text =~ s/\\016/ffi/g;
+    $text =~ s/\\214/fi/g;
+    $text =~ s/\\215/fl/g;
+    $text =~ s/\\017/\n\* /g; # asterisk?
+    $text =~ s/\\023/\023/g;  # e acute ('e)
+    $text =~ s/\\177/\252/g;  # u"
+#   $text =~ s/ ?? /\344/g;  # a"
+
+    print OUT "$text";
+    close OUT;
+    }
     &util::rm("$output_filestem.err") if (-e "$output_filestem.err");
     return 1;