Moving non tei2html tools to pptools

jhellingman · Jan 6, 2025 · cece34d · cece34d
1 parent 12bca16
commit cece34d
Show file tree

Hide file tree

Showing 34 changed files with 4,440 additions and 0 deletions.
diff --git a/pptools/broken-images.pl b/pptools/broken-images.pl
@@ -0,0 +1,67 @@
+#!/usr/bin/perl -w
+
+#
+# Test JPEG files in directories.
+#
+
+use strict;
+use warnings;
+use File::Basename;
+use Image::Magick;
+
+sub list_recursively($);
+
+sub list_recursively($) {
+    my ($directory) = @_;
+    my @files = (  );
+
+    unless (opendir(DIRECTORY, $directory)) {
+        print "Cannot open directory $directory!\n";
+        exit;
+    }
+
+    # Read the directory, ignoring special entries "." and ".."
+    @files = grep (!/^\.\.?$/, readdir(DIRECTORY));
+
+    closedir(DIRECTORY);
+
+    foreach my $file (@files) {
+        if (-f "$directory\\$file") {
+            handle_file("$directory\\$file");
+        } elsif (-d "$directory\\$file") {
+            list_recursively("$directory\\$file");
+        }
+    }
+}
+
+
+sub handle_file($) {
+    my ($file) = @_;
+    if ($file =~ m/^(.*)\.(jpg|png|gif)$/) {
+        my $image = new Image::Magick;
+        my $error = $image->Read($file);
+        if ($error) {
+            print "BROKEN $file\n";
+        } else {
+            print "OK     $file\n";
+        }
+    }
+}
+
+
+sub main() {
+    ## initial call ... $ARGV[0] is the first command line argument
+    my $file = $ARGV[0];
+
+    if (!defined $file) {
+        $file = ".";
+    }
+    if (-d $file) {
+        list_recursively($file);
+    } else {
+        handle_file($file);
+    }
+}
+
+
+main();
diff --git a/pptools/checkArchives.pl b/pptools/checkArchives.pl
@@ -0,0 +1,61 @@
+#!/usr/bin/perl -w
+
+#
+# Check the consistency of archive files.
+#
+
+use strict;
+use warnings;
+use File::Basename;
+use File::Temp;
+
+my $zip = "zip";
+my $sevenZip = "\"C:\\Program Files\\7-Zip\\7z\"";
+# $sevenZip = "7z";
+
+my $logFile = "checkArchives.log";
+
+main();
+
+sub main {
+    ## initial call ... $ARGV[0] is the first command line argument
+    list_recursively($ARGV[0]);
+}
+
+sub list_recursively($);
+
+sub list_recursively($) {
+    my ($directory) = @_;
+    my @files = (  );
+
+    unless (opendir(DIRECTORY, $directory)) {
+        logError("Cannot open directory $directory!");
+        exit;
+    }
+
+    # Read the directory, ignoring special entries "." and ".."
+    @files = grep (!/^\.\.?$/, readdir(DIRECTORY));
+
+    closedir(DIRECTORY);
+
+    foreach my $file (@files) {
+        if (-f "$directory/$file") {
+            handle_file("$directory/$file");
+        } elsif (-d "$directory/$file") {
+            list_recursively("$directory/$file");
+        }
+    }
+}
+
+
+sub handle_file($) {
+    my ($file) = @_;
+
+    if ($file =~ m/^(.*)\.(7z)$/) {
+        system ("$sevenZip t \"$file\" 1>>$logFile");
+    }
+
+    if ($file =~ m/^(.*)\.(zip)$/) {
+        system ("$zip -T \"$file\" 1>>$logFile");
+    }
+}
diff --git a/pptools/cleanHtml.pl b/pptools/cleanHtml.pl
@@ -0,0 +1,19 @@
+
+use strict;
+use warnings;
+
+while (<>) {
+
+    my $line = $_;
+
+    $line =~ s/<\/link>//g;
+    $line =~ s/<\/meta>//g;
+    $line =~ s/<\/img>//g;
+    $line =~ s/<\/hr>//g;
+
+    $line =~ s/<br\/>/<br>/g;
+
+    $line =~ s/<style><\/style>//g;
+
+    print $line;
+}
diff --git a/pptools/divn2div.pl b/pptools/divn2div.pl
@@ -0,0 +1,35 @@
+# divn2div.pl -- change from numbered to unnumbered TEI divs.
+
+use strict;
+use warnings;
+
+my $inputFile = $ARGV[0];
+
+open(INPUTFILE, $inputFile) || die("Could not open $inputFile");
+
+print STDERR "Handling $inputFile\n";
+
+my $previousLevel = 0;
+
+while (<INPUTFILE>) {
+    my $remainder = $_;
+
+    if ($remainder =~ m/<(body|front|back)(.*?)>/i) {
+        $previousLevel = 0;
+    }
+
+    while ($remainder =~ m/<div([0-9])(.*?)>/i) {
+        my $before = $`;
+        my $level = $1;
+        my $attrs = $2;
+        $remainder = $';
+
+        my $close = $previousLevel - $level;
+        $previousLevel = $level;
+        for ( ; $close >= 0; $close--) {
+            print $before . "</div>";
+        }
+        print "<div$attrs>";
+    }
+    print $remainder;
+}
diff --git a/pptools/extract-page.pl b/pptools/extract-page.pl
@@ -0,0 +1,17 @@
+# extract-page.pl
+
+use strict;
+use warnings;
+use FindBin qw($Bin);
+
+my $saxon           = "java -jar " . $Bin . "/lib/saxon9he.jar ";
+my $xsldir          = $Bin . "/..";                    # location of xsl stylesheets
+
+my $filename = $ARGV[0];
+my $page = $ARGV[1];
+
+if ($page eq '') {
+    system ("$saxon \"$filename\" $xsldir/extract-page.xsl");
+} else {
+    system ("$saxon \"$filename\" $xsldir/extract-page.xsl n=\"$page\"");
+}
diff --git a/pptools/extractPdf.pl b/pptools/extractPdf.pl
@@ -0,0 +1,97 @@
+#!/usr/bin/perl -w
+
+#
+# Extract images from JPEG files in directories.
+#
+
+use strict;
+use warnings;
+use File::Basename;
+use Getopt::Long;
+
+my $pdfimages = "pdfimages.exe"; # See http://www.foolabs.com/xpdf/download.html -> http://www.xpdfreader.com/download.html
+my $pdftopng = "pdftopng.exe";
+my $pdfcount = 1000;
+my $resolutionDpi = 300;
+my $extractActualImages = 0;
+my $showHelp = 0;
+
+GetOptions(
+    'j' => \$extractActualImages,
+    'r=i' => \$resolutionDpi,
+    'q' => \$showHelp,
+    'help' => \$showHelp
+    );
+
+
+if ($showHelp == 1) {
+    print "extractPdf.pl -- Wrapper around pdfimages.exe to extract images from a PDF file.\n\n";
+    print "Usage: extractPdf.pl [-jq] [-r=i] <file>\n\n";
+    print "Options:\n";
+    print "    j         Extract actual embedded images.\n";
+    print "    r=i       Output resolution in DPI.\n";
+    print "    q         Print this help and exit.\n";
+
+    exit(0);
+}
+
+
+sub list_recursively($);
+
+sub list_recursively($) {
+    my ($directory) = @_;
+    my @files = (  );
+
+    unless (opendir(DIRECTORY, $directory)) {
+        print "Cannot open directory $directory!\n";
+        exit;
+    }
+
+    # Read the directory, ignoring special entries "." and ".."
+    @files = grep (!/^\.\.?$/, readdir(DIRECTORY));
+
+    closedir(DIRECTORY);
+
+    foreach my $file (@files) {
+        if (-f "$directory\\$file") {
+            handle_file("$directory\\$file");
+        } elsif (-d "$directory\\$file") {
+            list_recursively("$directory\\$file");
+        }
+    }
+}
+
+
+sub handle_file($) {
+    my ($file) = @_;
+    if ($file =~ m/^(.*)\.(pdf)$/) {
+        my $extension = $2;
+        print "Extracting images from PDF: $file\n";
+        if ($extension eq 'pdf') {
+            $pdfcount++;
+            if ($extractActualImages == 0) {
+                system ("$pdftopng -r $resolutionDpi $file $pdfcount");
+            } else {
+                system ("$pdfimages -j -list \"$file\" $pdfcount");
+            }
+        }
+    }
+}
+
+
+sub main() {
+    ## initial call ... $ARGV[0] is the first command line argument
+    my $file = $ARGV[0];
+
+    if (!defined $file) {
+        $file = ".";
+    }
+    if (-d $file) {
+        list_recursively($file);
+    } else {
+        handle_file($file);
+    }
+}
+
+
+main();
diff --git a/pptools/fb2tei.pl b/pptools/fb2tei.pl
@@ -0,0 +1,40 @@
+# fb2tei.pl -- covert fictionbook to TEI.
+
+use strict;
+use warnings;
+use MIME::Base64;
+use FindBin qw($Bin);
+
+my $toolsdir        = $Bin;                                 # location of tools
+my $xsldir          = $toolsdir . "/..";                    # location of xsl stylesheets
+my $saxon           = "java -jar " . $toolsdir . "/lib/saxon9he.jar ";
+
+my $filename = $ARGV[0];
+$filename =~ /^(.*)\.fb2$/;
+my $basename    = $1;
+
+system ("$saxon $filename $xsldir/fb2tei.xsl > $basename.xml");
+
+# convert extracted .hex files to binary
+my @files = <*.hex>;
+foreach my $file (@files) {
+    convertFile($file);
+}
+
+system ("perl -S tei2html.pl -h $basename.xml");
+
+sub convertFile($) {
+    my $filename = shift;
+
+    open INFILE, $filename or die "Unable to open file: $filename"; 
+    my $string = join("", <INFILE>); 
+    close INFILE;
+    my $binary = decode_base64($string);
+
+    $filename =~ /^(.*)\.hex$/;
+    my $outputFilename = $1;
+
+    open(OUTFILE, '>:raw', $outputFilename) or die "Unable to open: $outputFilename";
+    print OUTFILE $binary ;
+    close(OUTFILE);
+}
diff --git a/pptools/fixAnchors.pl b/pptools/fixAnchors.pl
@@ -0,0 +1,55 @@
+#
+# fixAnchors.pl -- provide anchors with sequential numbers per page in TEI tagged files
+#
+
+use strict;
+use warnings;
+
+use Roman;                          # Roman.pm version 1.1 by OZAWA Sakuro <ozawa@aisoft.co.jp>
+use SgmlSupport qw/getAttrVal/;
+
+my $inputFile   = $ARGV[0];
+my $currentPage = 0;
+my $currentAnchor = 1;
+my $prefix = "a";
+
+
+open(INPUTFILE, $inputFile) || die("Could not open $inputFile");
+
+print STDERR "Fixing anchors in $inputFile\n";
+
+while (<INPUTFILE>) {
+    my $line = $_;
+    my $remainder = $line;
+    while ($remainder =~ m/<pb(.*?)>/) {
+        my $before = $`;
+        my $attrs = $1;
+        $remainder = $';
+        handleAnchors($before);
+        print "<pb$1>";
+        $currentAnchor = 1;
+        $currentPage = getAttrVal('n', $attrs);
+        $currentPage = isroman($currentPage) ? arabic($currentPage) : $currentPage;
+    }
+    handleAnchors($remainder);
+}
+
+
+sub handleAnchors($) {
+    my $remainder = shift;
+    while ($remainder =~ m/<anchor(.*?)>/) {
+        my $before = $`;
+        my $attrs = $1;
+        $remainder = $';
+        print $before;
+        print "<anchor id=$prefix$currentPage.$currentAnchor>";
+        if ($remainder =~ m/<ab type=lineNum>([0-9]+)<\/ab>/) {
+            my $lineNum = $1;
+            if ($lineNum != $currentAnchor) {
+                print STDERR "WARNING: anchor at $currentPage.$currentAnchor doesn't match lineNum: $lineNum\n";
+            }
+        }
+        $currentAnchor++;
+    }
+    print $remainder;
+}