Changeset 1654

Timestamp:

2000-11-03T15:22:00+13:00 (23 years ago)

Author:

paynter

Message:

Check .doc files to see if they are RTF files, Word 6/7/8 files that wv
handles, or "unknown" files (which we strip of binary characters and hope
the result is worthwhile).

File:

• trunk/gsdl/bin/script/gsConvert.pl (modified) (6 diffs)

trunk/gsdl/bin/script/gsConvert.pl

@@ -137 +137 @@
     ($input_filename, $output_filestem, $output_type) = @_;
 
+    # Many .doc files are not in fact word documents!
+    my $realtype = &find_docfile_type($input_filename);
+
+    if ($realtype eq "word678") {
+    return &convertWord678($input_filename, $output_filestem, $output_type);
+    } elsif ($realtype eq "rtf") {
+    return &convertRTF($input_filename, $output_filestem, $output_type);
+    } else {
+    return &convertAnything($input_filename, $output_filestem, $output_type);
+    }
+}
+
+# Convert a Microsoft word 6/7/8 document
+
+sub convertWord678 {
+    ($input_filename, $output_filestem, $output_type) = @_;
+
     my $success = 0;
 
@@ -147 +164 @@
     }
 
+    return &convertAnything($input_filename, $output_filestem, $output_type);
+}
+
+
+# Convert a Rich Text Format (RTF) file
+
+sub convertRTF {
+    ($input_filename, $output_filestem, $output_type) = @_;
+
+    my $success = 0;
+
+    # Attempt specialised conversion to HTML
+    if (!$output_type || ($output_type =~ /html/i)) {
+    $success = &rtf_to_html($input_filename, $output_filestem);
+    if ($success) {
+        return "html";
+    }
+    }
+
+    return &convertAnything($input_filename, $output_filestem, $output_type);
+}
+
+
+# Convert an unidentified file
+
+sub convertAnything {
+    ($input_filename, $output_filestem, $output_type) = @_;
+    
+    my $success = 0;
+    
     # Attempt simple conversion to HTML
     if (!$output_type || ($output_type =~ /html/i)) {
@@ -162 +209 @@
     }
     }
-
     return "fail";
-
-}
+}
+
 
 
@@ -213 +259 @@
     return "fail";
 
+}
+
+
+# Find the real type of a .doc file
+#
+# We seem to have alot of files with a .dco extension that are .rtf 
+# files or Word 5 files.  This function attempts to tell the difference.
+
+sub find_docfile_type {
+    ($input_filename) = @_;
+    
+    open(CHK, "<$input_filename");
+    my $line = "";
+    my $first = 1;
+
+    while (<CHK>) {
+    
+    $line = $_;
+
+    if ($first) {
+        # check to see if this is an rtf file
+        if ($line =~ /^\{\\rtf/) {
+        close(CHK);
+        return "rtf";
+        }
+    }
+    
+    # is theis a word 6/7/8 document?
+    if ($line =~ /Word\.Document\.[678]/) {
+        close(CHK);
+        return "word678";
+    }
+
+    $first = 0;
+
+    }
+
+    return "unknown";
 }
 
@@ -235 +319 @@
     my $wvWare = &util::filename_cat($wv_home, "bin", "wvWare");
     return 0 unless (-e "$wvWare");
-    $cmd = "$wvWare --charset utf-8 --config $wv_conf";
+    $cmd = "ulimit -t 20;";
+    $cmd .= "$wvWare --charset utf-8 --config $wv_conf";
     $cmd .= " \"$input_filename\" > \"$output_filestem.html\" 2>\"$output_filestem.err\"";
-    
+
     # execute the command
     if (system($cmd)>0)
@@ -262 +347 @@
 
 
+# Attempt to convert an RTF document to html with rtftohtml
+#
+# rtf2html isn't distributed with Greenstone because it is not 
+# distributed under teh GPL.  If you know of a better solution,
+# please let me know.
+
+sub rtf_to_html {
+    ($input_filename, $output_filestem) = @_;
+
+    # formulate the command
+    my $r_cmd = &util::filename_cat($ENV{'GSDLHOME'}, "packages", "unix", 
+                    "rtf2html", "rtf2html", "rtf2html");
+    $r_cmd = "rtf2html" unless (-e "$r_cmd");
+    return 0 unless (-e "$r_cmd");
+    $cmd = "ulimit -t 20;";
+    $cmd .= "$r_cmd";
+    $cmd .= " \"$input_filename\" > \"$output_filestem.html\" 2>\"$output_filestem.err\"";
+
+    # execute the command
+    if (system($cmd)>0)
+    {
+    print STDERR "Error executing rtf converter: $!. Continuing...\n";
+    }
+
+    # Was the conversion successful?
+    if (-e "$output_filestem.html") {
+    open(TMP, "$output_filestem.html");
+    $line = <TMP>;
+    close(TMP);
+    if ($line && $line =~ /DOCTYPE HTML/) {
+        &util::rm("$output_filestem.err");
+        return 1;
+    } else {
+        # An error of some sort occurred
+        &util::rm("$output_filestem.html");
+        &util::rm("$output_filestem.err");
+    }
+    }
+    return 0;
+}
+
+
 # Convert a pdf file to html with the pdftohtml command