root/main/trunk/greenstone2/perllib/plugins/PDFv2Plugin.pm @ 32286

Revision 32286, 33.9 KB (checked in by ak19, 16 months ago)

PDFv2Plugin will only work out of the box for GS3 now: PDFBoxConverter is no longer loaded up via AutoLoadConverters?, but directly instantiated. Not yet tidied up

Line 
1###########################################################################
2#
3# PDFv2Plugin.pm -- pdf plugin that uses xpdftools and pdfbox to process PDFs.
4# It only works out of the box for GS3 since it assumes the pdfbox extension
5# is installed.
6# A component of the Greenstone digital library software
7# from the New Zealand Digital Library Project at the
8# University of Waikato, New Zealand.
9#
10# Copyright (C) 1999-2001 New Zealand Digital Library Project
11#
12# This program is free software; you can redistribute it and/or modify
13# it under the terms of the GNU General Public License as published by
14# the Free Software Foundation; either version 2 of the License, or
15# (at your option) any later version.
16#
17# This program is distributed in the hope that it will be useful,
18# but WITHOUT ANY WARRANTY; without even the implied warranty of
19# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
20# GNU General Public License for more details.
21#
22# You should have received a copy of the GNU General Public License
23# along with this program; if not, write to the Free Software
24# Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
25#
26###########################################################################
27package PDFv2Plugin;
28
29use strict;
30no strict 'refs'; # so we can use a var for filehandles (e.g. STDERR)
31no strict 'subs'; # allow filehandles to be variables and viceversa
32
33use ReadTextFile;
34use unicode;
35use Mojo::DOM; # for HTML parsing
36
37#use AutoLoadConverters;
38use PDFBoxConverter;
39use ConvertBinaryFile;
40
41#@PDFv2Plugin::ISA = ('ConvertBinaryFile', 'AutoLoadConverters', 'ReadTextFile');
42@PDFv2Plugin::ISA = ('ConvertBinaryFile', 'PDFBoxConverter', 'ReadTextFile');
43
44
45my $convert_to_list =
46    [ { 'name' => "auto", # pretty_html using xpdftools' pdftohtml
47    'desc' => "{ConvertBinaryFile.convert_to.auto}" },
48      { 'name' => "text", # xpdftools' pdftotext
49    'desc' => "{ConvertBinaryFile.convert_to.text}" },
50      { 'name' => "paged_text", # pdfbox
51    'desc' => "{ConvertBinaryFile.convert_to.paged_text}" },
52     
53      { 'name' => "html", # pdfbox ## TODO: rename this to html_without_imgs?
54    'desc' => "{PDFPlugin.convert_to.html}" },
55      { 'name' => "pretty_html", # xpdftools
56    'desc' => "{PDFPlugin.convert_to.pretty_html}" },
57      { 'name' => "paged_pretty_html", # xpdftools
58    'desc' => "{PDFPlugin.convert_to.paged_pretty_html}"},
59
60      # pdfbox for all pagedimg(txt) output formats:
61      { 'name' => "pagedimg_jpg",
62    'desc' => "{ConvertBinaryFile.convert_to.pagedimg_jpg}"},
63      { 'name' => "pagedimg_png",
64    'desc' => "{ConvertBinaryFile.convert_to.pagedimg_png}"},
65     
66      { 'name' => "pagedimgtxt_jpg",
67    'desc' => "{ConvertBinaryFile.convert_to.pagedimgtxt_jpg}"},
68      { 'name' => "pagedimgtxt_png",
69    'desc' => "{ConvertBinaryFile.convert_to.pagedimgtxt_png}"},
70      ];
71
72
73my $arguments =
74    [
75     { 'name' => "convert_to",
76       'desc' => "{ConvertBinaryFile.convert_to}",
77       'type' => "enum",
78       'reqd' => "yes",
79       'list' => $convert_to_list,
80       'deft' => "pretty_html" },   
81     { 'name' => "process_exp",
82       'desc' => "{BaseImporter.process_exp}",
83       'type' => "regexp",
84       'deft' => &get_default_process_exp(),
85       'reqd' => "no" },
86     { 'name' => "block_exp",
87       'desc' => "{CommonUtil.block_exp}",
88       'type' => "regexp",
89       'deft' => &get_default_block_exp() },
90     { 'name' => "metadata_fields",
91       'desc' => "{HTMLPlugin.metadata_fields}",
92       'type' => "string",
93       'deft' => "Title,Author,Subject,Keywords" },
94     { 'name' => "metadata_field_separator",
95    'desc' => "{HTMLPlugin.metadata_field_separator}",
96    'type' => "string",
97    'deft' => "" },
98     { 'name' => "dpi",
99       'desc' => "{PDFv2Plugin.dpi}",
100       'deft' => "96",
101       'type' => "int" }, # 72DPI is xpdf's pdftohtml's default. pdfbox' default is 96DPI in headless mode else detected from the screen resolution, see https://pdfbox.apache.org/2.0/commandline.html#pdftoimage
102#     { 'name' => "use_sections",
103#       'desc' => "{PDFPlugin.use_sections}",
104#       'type' => "flag" },
105#     { 'name' => "description_tags",
106#       'desc' => "{HTMLPlugin.description_tags}",
107#       'type' => "flag" },
108      { 'name' => "use_realistic_book",
109        'desc' => "{PDFPlugin.use_realistic_book}",
110    'type' => "flag"}
111     ];
112
113my $options = { 'name'     => "PDFv2Plugin",
114        'desc'     => "{PDFPlugin.desc}",
115        'abstract' => "no",
116        'inherits' => "yes",
117        'srcreplaceable' => "yes", # Source docs in PDF can be replaced with GS-generated html     
118        'args'     => $arguments };
119
120sub new {
121    my ($class) = shift (@_);
122    my ($pluginlist,$inputargs,$hashArgOptLists) = @_;
123    push(@$pluginlist, $class);
124
125#    push(@$inputargs,"-title_sub");
126#    push(@$inputargs,'^(Page\s+\d+)?(\s*1\s+)?');
127
128    push(@{$hashArgOptLists->{"ArgList"}},@{$arguments});
129    push(@{$hashArgOptLists->{"OptList"}},$options);
130
131    #    my $auto_converter_self = new AutoLoadConverters($pluginlist,$inputargs,$hashArgOptLists,["PDFBoxConverter"],1);
132    my $pdfbox_converter_self = new PDFBoxConverter($pluginlist, $inputargs, $hashArgOptLists);
133    my $cbf_self = new ConvertBinaryFile($pluginlist, $inputargs, $hashArgOptLists);
134    #    my $self = BaseImporter::merge_inheritance($auto_converter_self, $cbf_self);
135    my $self = BaseImporter::merge_inheritance($pdfbox_converter_self, $cbf_self);
136   
137    if ($self->{'info_only'}) {
138    # don't worry about any options etc
139    return bless $self, $class;
140    }
141   
142    $self = bless $self, $class;
143    $self->{'file_type'} = "PDF";
144
145    # convert_options are passed through to gsConvert.pl by ConvertBinaryFile.pm
146   
147    # the most important option is the tool that's used to do the conversion
148    $self->{'convert_options'} = "-pdf_tool xpdftools"; # default for PDFv2Plugin. If pdfbox_conversion is on, the pdfbpox GS extension sets pdf_tool to pdfbox
149
150    # Setting dpi has meaning for xpdftools pdftohtml (so paged_pretty_html and pretty_html)
151    # and for when pdfbox outputs an image for each page (pagedimg, pagedimgtxt).
152    # dpi has no effect on (paged_)text and html output modes.
153    my $dpi = $self->{"dpi"};
154    $self->{'convert_options'} .= " -pdf_dpi $dpi";
155
156    # The old pdftohtml tool used by PDFPlugin didn't do PDF to txt conversion on Windows
157    # But PDFv2Plugin now supports PDF to txt conversion on Windows too using XPDFTools' pdftotext
158
159    if ($self->{'convert_to'} eq "auto") {
160    # choose pretty_html is the best default option when using xpdftools
161    $self->{'convert_to'} = "pretty_html";
162    }
163    if ($self->{'use_realistic_book'}) {
164    if ($self->{'convert_to'} ne "html") {
165        print STDERR "PDFs will be converted to HTML for realistic book functionality\n";
166        $self->{'convert_to'} = "html";
167    }
168    }
169
170    # if pdfbox_conversion is not on, check convert_to to make sure that xpdftools can
171    # support the selected output format, or fallback on a sensible default
172    # Not all available conversion output options are possible with xpdftools, as some are
173    # only handled by pdfbox. If a format is unavailable with xpdftools, default to pretty_html
174    # if (!$self->{"pdfbox_conversion"}) {
175    #   my $convert_to = $self->{'convert_to'};
176    #   my $fallback_convert_to = $convert_to;
177    #   if($convert_to =~ /^html$/) {
178    #       $fallback_convert_to = "pretty_html";
179    #   }
180    #   elsif ($self->{'convert_to'} =~ /^pagedimg/) {
181    #       $fallback_convert_to = "paged_pretty_html";
182    #   }
183    #   elsif ($self->{'convert_to'} =~ /^paged_text$/) {
184    #       #   print STDERR "@@@ Conversion to " . $self->{'convert_to'} , " with Xpdf Tools is not yet implemented.\n";
185    #       $fallback_convert_to = "text";
186    #   }
187
188    #   if($convert_to =~ /^(html|pagedimg|paged_text)/) {
189    #       &gsprintf::gsprintf(STDERR, "{PDFv2Plugin.conversion_needs_pdfbox}\n", ($self->{'convert_to'}, $fallback_convert_to));
190    #       $self->{'convert_to'} = $fallback_convert_to;
191    #   }
192    # }
193   
194    # set convert_to_plugin and convert_to_ext
195    $self->set_standard_convert_settings();
196
197    my $secondary_plugin_name = $self->{'convert_to_plugin'};
198    my $secondary_plugin_options = $self->{'secondary_plugin_options'};
199
200    if (!defined $secondary_plugin_options->{$secondary_plugin_name}) {
201    $secondary_plugin_options->{$secondary_plugin_name} = [];
202    }
203    my $specific_options = $secondary_plugin_options->{$secondary_plugin_name};
204
205    # following title_sub removes "Page 1" added by pdftohtml, and a leading
206    # "1", which is often the page number at the top of the page. Bad Luck
207    # if your document title actually starts with "1 " - is there a better way?
208#    push(@$specific_options , "-title_sub", '^(Page\s+\d+)?(\s*1\s+)?');
209    my $associate_tail_re = $self->{'associate_tail_re'};
210    if ((defined $associate_tail_re) && ($associate_tail_re ne "")) {
211    push(@$specific_options, "-associate_tail_re", $associate_tail_re);
212    }
213    push(@$specific_options, "-file_rename_method", "none");
214   
215    if ($secondary_plugin_name eq "HTMLPlugin") {
216    # pdftohtml always produces utf8 - What about pdfbox???
217    # push(@$specific_options, "-input_encoding", "utf8");
218    push(@$specific_options, "-extract_language") if $self->{'extract_language'};
219    push(@$specific_options, "-processing_tmp_files");
220    # Instruct HTMLPlug (when eventually accessed through read_into_doc_obj)
221    # to extract these metadata fields from the HEAD META fields
222    if (defined $self->{'metadata_fields'} && $self->{'metadata_fields'} =~ /\S/) {
223        push(@$specific_options,"-metadata_fields",$self->{'metadata_fields'});
224    } else {
225        push(@$specific_options,"-metadata_fields","Title,GENERATOR,date,author<Creator>");
226    }
227    if (defined $self->{'metadata_field_separator'} && $self->{'metadata_field_separator'} =~ /\S/) {
228        push(@$specific_options,"-metadata_field_separator",$self->{'metadata_field_separator'});
229    }
230    if ($self->{'use_sections'} || $self->{'description_tags'}) {
231        $self->{'description_tags'} = 1;
232        push(@$specific_options, "-description_tags");
233    }
234    if ($self->{'use_realistic_book'}) {
235        push(@$specific_options, "-use_realistic_book");
236    }
237    if($self->{'convert_to'} eq "paged_pretty_html") { # for paged pretty html, the default should be to sectionalise
238        # the single superpage, the one containing divs representing individual pages as sections, on headings
239        push(@$specific_options, "sectionalise_using_h_tags");
240    }
241    }
242    elsif ($secondary_plugin_name eq "PagedImagePlugin") {
243    push(@$specific_options, "-screenviewsize", "1000");
244    push(@$specific_options, "-enable_cache");
245    push(@$specific_options, "-processing_tmp_files");
246    }
247
248    $self = bless $self, $class;
249    $self->load_secondary_plugins($class,$secondary_plugin_options,$hashArgOptLists);
250    return $self;
251}
252
253sub get_default_process_exp {
254    my $self = shift (@_);
255
256    return q^(?i)\.pdf$^;
257}
258
259# so we don't inherit HTMLPlug's block exp...
260sub get_default_block_exp {
261    return "";
262}
263
264sub init {
265    my $self = shift (@_);
266
267    # ConvertBinaryFile init
268    $self->SUPER::init(@_);
269#   $self->AutoLoadConverters::init(@_);
270    $self->PDFBoxConverter::init(@_);
271
272}
273
274sub begin {
275    my $self = shift (@_);
276
277#   $self->AutoLoadConverters::begin(@_);
278    $self->PDFBoxConverter::begin(@_);
279    $self->SUPER::begin(@_);
280
281}
282
283sub deinit {
284    my $self = shift (@_);
285
286    $self->PDFBoxConverter::deinit(@_);
287#   $self->AutoLoadConverters::deinit(@_);
288    $self->SUPER::deinit(@_);
289
290}
291 
292# By setting hashing to be on ga xml this ensures that two
293# PDF files that are identical except for the metadata
294# to hash to different values. Without this, when each PDF
295# file is converted to HTML there is a chance that they
296# will both be *identical* if the conversion utility does
297# not embed the metadata in the generated HTML. This is
298# certainly the case when PDFBOX is being used.
299
300# This change makes this convert to based plugin more
301# consistent with the original vision that the same document
302# with different metadata should
303# be seen as different.
304
305sub get_oid_hash_type {
306    my $self = shift (@_);
307    return "hash_on_ga_xml";
308}
309 
310
311sub tmp_area_convert_file {
312
313    my $self = shift (@_);
314    my ($output_ext, $input_filename, $textref) = @_;
315   
316    if($self->{'convert_to'} eq "text" || $self->{'convert_to'} =~ m/pretty_html$/) { # use xpdftools
317    return $self->ConvertBinaryFile::tmp_area_convert_file(@_);
318    }
319   
320    # for all other output formats, use pdfbox:   
321    #return $self->AutoLoadConverters::tmp_area_convert_file(@_);
322    # Here, we now do what AutoLoadConverters::tmp_area_convert_file(@_) does:
323    my ($result, $result_str, $new_filename) = $self->PDFBoxConverter::convert($input_filename, $output_ext);
324    if (defined $result && $result != 0) {
325    return $new_filename;
326    }
327    my $outhandle=$self->{'outhandle'};
328    print $outhandle "PDFBoxConverter had a conversion error\n";
329    print $outhandle "$@\n";
330    if (defined $result_str) {
331    print $outhandle "$result_str\n";
332    }
333    return "";
334}
335
336# Overriding to do some extra handling for pretty_html/paged_pretty_html output mode
337sub run_conversion_command {
338    my $self = shift (@_);
339    my ($tmp_dirname, $tmp_inputPDFname, $utf8_tailname, $lc_suffix, $tailname, $suffix) = @_;   
340
341    if($self->{'convert_to'} !~ m/pretty_html$/) {
342    return $self->ConvertBinaryFile::run_conversion_command(@_);
343    }
344
345    # else, paged_pretty_html or pretty_html
346
347    # if output mode is (paged_)pretty_html, we use Xpdf tools' pdftohtml and tell it
348    # to create a subdir called "pages" in the tmp area to puts its products
349    # in there. (Xpdf's pdftohtml needs to be passed a *non-existent* directory
350    # parameter, the "pages" subdir). If Xpdf's pdftohtml has successfully run,
351    # the intermediary output file tmp/<random-num>/pages/index.html should
352    # exist (besides other output products there)
353
354    # We let ConvertBinaryFile proceed normally, but the return value should reflect
355    # that on success it should expect the intermediary product tmpdir/pages/index.html
356    # (which is the product of xpdftohtml conversion).
357    my $output_filename = $self->ConvertBinaryFile::run_conversion_command(@_);
358    $output_filename = &FileUtils::filenameConcatenate($tmp_dirname, "pages", "index.html");
359
360    # However, when convert_post_process() is done, it should have output the final
361    # product of the (paged_)pretty_html conversion: an html file of the same name and in the
362    # same tmp location as the input PDF file.
363
364    my ($name_prefix, $output_dir, $ext)
365    = &File::Basename::fileparse($tmp_inputPDFname, "\\.[^\\.]+\$");
366    $self->{'conv_filename_after_post_process'} = &FileUtils::filenameConcatenate($output_dir, $name_prefix.".html");
367#    print STDERR "@@@@@ final paged html file will be: " . $self->{'conv_filename_after_post_process'} . "\n";
368
369    return $output_filename;
370}
371
372sub convert_post_process
373{
374    my $self = shift (@_);
375    my ($conv_filename) = @_;
376
377    my $outhandle=$self->{'outhandle'};
378
379    if($self->{'convert_to'} =~ /pretty_html/) { # (paged_)pretty_html
380    # special post-processing for (paged_)pretty_html mode, as HTML pages generated
381    # by xpdf's pdftohtml need to be massaged into the form we want
382    $self->xpdftohtml_convert_post_process($conv_filename);
383    }
384    else { # use PDFPlugin's usual post processing
385    $self->default_convert_post_process($conv_filename);
386    }
387}
388
389# Called after gsConvert.pl has been run to convert a PDF to (paged_)pretty_html
390# using Xpdftools' pdftohtml
391# This method will do some cleanup of the HTML files produced after XPDF has produced
392# an HTML doc for each PDF page: it first gets rid of the default index.html.
393# Instead, it constructs a single html page containing each original HTML page
394# <body> nested as divs instead, with simple section information inserted at the top
395# of each 'page' <div> and some further styling customisation. This HTML manipulation
396# is to be done with the Mojo::DOM perl package.
397# Note that since xpdf's pdftohtml would have failed if the output dir already
398# existed and for simpler naming, the output files are created in a new "pages"
399# subdirectory of the tmp location parent of $conv_filename instead
400sub xpdftohtml_convert_post_process
401{
402    my $self = shift (@_);
403    my ($pages_index_html) = @_; # = tmp/<rand>/pages/index.html for (paged_)pretty_html output mode
404    my $output_filename = $self->{'conv_filename_after_post_process'};
405
406    # Read in all the html files in tmp's "pages" subdir, except for index.html.
407    # and use it to create a new html file called $self->{'conv_filename_after_post_process'}
408    # which will consist of a slightly modified version of
409    # each of the other html files concatenated together.
410
411    my $outhandle=$self->{'outhandle'};
412
413    my ($tailname, $pages_subdir, $suffix)
414    = &File::Basename::fileparse($pages_index_html, "\\.[^\\.]+\$");
415
416    # Code from util::create_itemfile()
417    # Read in all the files
418    opendir(DIR, $pages_subdir) || die "can't opendir $pages_subdir: $!";
419    my @page_files = grep {-f "$pages_subdir/$_"} readdir(DIR);
420    closedir DIR;
421    # Sort files in the directory by page_num
422    # files are named index.html, page1.html, page2.html, ..., pagen.html
423    sub page_number {
424    my ($dir) = @_;
425    my ($pagenum) =($dir =~ m/^page(\d+)\.html?$/i);
426    $pagenum = 0 unless defined $pagenum; # index.html will be given pagenum=0
427    return $pagenum;
428    }
429    # sort the files in the directory in the order of page_num rather than lexically.
430    @page_files = sort { page_number($a) <=> page_number($b) } @page_files;
431
432    #my $num_html_pages = (scalar(@page_files) - 1)/2; # skip index file.
433              # For every html file there's an img file, so halve the total num.
434              # What about other file types that may potentially be there too???
435    my $num_html_pages = 0;
436    foreach my $pagefile (@page_files) {
437    $num_html_pages++ if $pagefile =~ m/\.html?$/ && $pagefile !~ /^index\.html?/i;
438    }
439
440    # Prepare to create our new html page that will contain all the individual
441    # htmls generated by xpdf's pdftohtml in sequence.
442    # First write the opening html tags out to the output file. These are the
443    # same tags and their contents, including <meta>, as is generated by
444    # Xpdf's pdftohtml for each of its individual html pages.
445    my $start_text = "<html>\n<head>\n";
446    my ($output_tailname, $tmp_subdir, $html_suffix)
447    = &File::Basename::fileparse($output_filename, "\\.[^\\.]+\$");
448    $start_text .= "<title>$output_tailname</title>\n";
449    $start_text .= "<meta http-equiv=\"Content-Type\" content=\"text/html; charset=UTF-8\">\n";
450    $start_text .= "</head>\n<body>\n\n";
451   
452    if($self->{'convert_to'} =~ /paged_pretty_html/) { # then add the <h>tags for sectionalising
453    $start_text .= "<h1>$output_tailname</h1>\n\n";
454    }
455
456    #handle content encodings the same way that default_convert_post_process does
457    # $self->utf8_write_file ($start_text, $conv_filename); # will close file after write   
458    # Don't want to build a giant string in memory of all the pages concatenated
459    # and then write it out in one go. Instead, build up the final single page
460    # by writing each modified (paged_)pretty_html file out to it as this is processed.
461    # Copying file open/close code from CommonUtil::utf8_write_file()
462    if (!open (OUTFILE, ">:utf8", $output_filename)) {
463    gsprintf(STDERR, "PDFv2Plugin::xpdftohtml_convert_post_process {CommonUtil.could_not_open_for_writing} ($!)\n", $output_filename);
464    die "\n";
465    }
466    print OUTFILE $start_text;
467
468    # Get the contents of each individual HTML page generated by Xpdf, after first
469    # modifying each, and write each out into our single all-encompassing html
470    foreach my $pagefile (@page_files) {
471    if ($pagefile =~ m/\.html?$/ && $pagefile !~ /^index\.html?/i) {
472        my $page_num = page_number($pagefile);   
473        # get full path to pagefile
474        $pagefile = &FileUtils::filenameConcatenate($pages_subdir, $pagefile);
475#       print STDERR "@@@ About to process html file $pagefile (num $page_num)\n";
476        my $modified_page_contents = $self->_process_pretty_html_page($pagefile, $page_num, $num_html_pages);
477        print OUTFILE "$modified_page_contents\n\n";
478    }
479    }
480
481    # we've now created a single HTML file by concatenating (a modified version)
482    # of each paged html file
483    print OUTFILE "</body>\n</html>\n"; # write out closing tags
484    close OUTFILE; # done
485
486    # Get rid of all the htm(l) files incl index.html in the associated "pages"
487    # subdir, since we've now processed them all into a single html file
488    # one folder level up and we don't want HTMLPlugin to process all of them next.
489    &FileUtils::removeFilesFiltered($pages_subdir, "\.html?\$"); #  no specific whitelist, but blacklist htm(l)
490
491    # now the tmp area should contain a single html file contain all the html pages'
492    # contents in sequence, and a "pages" subdir containing the screenshot images
493    # of each page.   
494    # HTMLPlugin will process these further in the plugin pipeline
495}
496
497# For whatever reason, most html <tags> don't get printed out in GLI
498# So when debugging, use this function to print them out as [tags] instead.
499sub _debug_print_html
500{
501    my $self = shift (@_);
502    my ($string_or_dom) = @_;
503
504    # can't seem to determine type of string with ref/reftype
505    # https://stackoverflow.com/questions/1731333/how-do-i-tell-what-type-of-value-is-in-a-perl-variable
506    # Not needed, as $dom objects seem to get correctly stringified in string contexts
507    # $dom.to_string/$dom.stringify seem to get called, no need to call them
508    # https://stackoverflow.com/questions/5214543/what-is-stringification-in-perl
509    my $escapedTxt = $string_or_dom;
510    $escapedTxt =~ s@\<@[@sg;
511    $escapedTxt =~ s@\>@]@sg;
512
513    print STDERR "#### $escapedTxt\n";
514}
515
516# Helper function for (paged_)pretty_html
517# to read in each page of pretty_html generated by Xpdf's pdftohtml
518# then modify the html suitably using the HTML parsing functions offered by
519# Mojo::DOM, then return the modified HTML content as a string.
520# For paged_pretty_html, some additional modification is done to sectionalise the final html
521# See https://mojolicious.org/perldoc/Mojo/DOM
522sub _process_pretty_html_page
523{
524    my $self = shift (@_);
525    my ($pagefile, $page_num, $num_html_pages) = @_;
526
527    my $text = "";
528
529    # handling content encoding the same way default_convert_post_process does
530    $self->read_file ($pagefile, "utf8", "", \$text);
531
532    my $dom = Mojo::DOM->new($text);
533
534#    $self->_debug_print_html($dom);
535
536    # there's a <style> element on the <html>, we need to shift it into the <div>
537    # tag that we'll be creating. We'll first slightly modify the <style> element
538    # store the first style element, which is the only one and in the <body>
539    # we'll later insert it as child of an all-encompassing div that we'll create
540    my $page_style_tag_str = $dom->at('html')->at('style')->to_string;
541    # In the style tag, convert id style references to class style references
542    my $css_class = ".p".$page_num."f";
543    $page_style_tag_str =~ s@\#f@$css_class@sg;
544    my $style_element = Mojo::DOM->new($page_style_tag_str)->at('style'); # modified   
545#$self->_debug_print_html($style_element);
546
547    # need to know the image's height to set the height of the surrounding
548    # div that's to replace this page's <body>:
549    my $img_height = $dom->find('img')->[0]{height};
550
551    # 2. Adjust the img#background src attribute to point to the pages subdir for imgs
552    # 3. Set that img tag's class=background, and change its id to background+$page_num
553    my $bg_img_tag=$dom->find('img#background')->[0];
554    my $img_src_str = $bg_img_tag->{src};
555    $img_src_str = "pages/$img_src_str";
556    $bg_img_tag->attr(src => $img_src_str); # reset
557#$self->_debug_print_html($bg_img_tag);
558    # set both class and modified id attributes in one step:
559    $bg_img_tag->attr({class => "background", id => "background".$page_num});
560#$self->_debug_print_html($bg_img_tag);
561
562    # get all the <span> nested inside <div class="txt"> elements and
563    # 1. set their class attr to be "p + page_num + id-of-the-span",
564    # 2. then delete the id, because the span ids have been reused when element
565    # ids ought to be unique. Which is why we set the modified ids to be the
566    # value of the class attribute instead
567    $dom->find('div.txt span')->each(sub {
568    $_->attr(class => "p". $page_num. $_->{id});
569    delete $_->{id};
570                     }); # both changes done in one find() operation
571#$self->_debug_print_html($dom->find('div.txt span')->last);
572
573    # Finally can create our new dom, starting with a div tag for the current page
574    # Must be: <div id="$page_num" style="position:relative; height:$img_height;"/>
575#    my $new_dom = Mojo::DOM->new_tag('div', id => "page".$page_num, style => "position: relative; height: ".$img_height."px;" )
576    my $new_dom = Mojo::DOM->new_tag('div', style => "position: relative; height: ".$img_height."px;" );
577#$self->_debug_print_html($new_dom);
578    $new_dom->at('div')->append_content($style_element)->root;
579
580
581#$self->_debug_print_html($new_dom);
582    # Copy across all the old html's body tag's child nodes into the new dom's new div tag
583    $dom->at('body')->child_nodes->each(sub { $new_dom->at('div')->append_content($_)}); #$_->to_string
584#$self->_debug_print_html($new_dom);
585
586    # build up the outer div
587    my $inner_div_str = $new_dom->to_string;
588    my $page_div = "<div id=\"page".$page_num."\">\n";
589
590    # If paged_pretty_html, then add <h>tags for sectionalising
591    if($self->{'convert_to'} =~ /paged_pretty_html/) {
592   
593    # Append a page range bucket heading if applicable: if we have more than 10 pages
594    # to display in the current bucket AND we're on the first page of each bucket of 10 pages.
595    # Dr Bainbridge thinks for now we need only consider PDFs where the
596    # total number of pages < 1000 and create buckets of size 10 (e.g. 1-10, ... 51-60, ...)
597    # If number of remaining pages >= 10, then create new bucket heading
598    # e.g. "Pages 30-40"
599    if(($page_num % 10) == 1 && ($num_html_pages - $page_num) > 10) {
600        # Double-digit page numbers that start with 2
601        # i.e. 21 to 29 (and 30) should be in 21 to 30 range
602        my $start_range = $page_num - ($page_num % 10) + 1;
603        my $end_range = $page_num + 10 - ($page_num % 10);
604        $page_div .= "<h2 style=\"font-size:1em;font-weight:normal;\">Pages ".$start_range . "-" . $end_range."</h2>\n";
605    }
606
607    # Every page is now a section too, not just buckets.
608    # Whether we're starting a new bucket or not, add a simpler heading: just the pagenumber, "Page #"
609    # However, this should be <H3> when there are buckets and <H2> when there aren't any.
610    if($num_html_pages > 10) {
611        $page_div .= "<h3 style=\"font-size:1em;font-weight:normal;\">Page ".$page_num."</h3>\n";       
612    } else { # PDF has less than 10 pages in total
613        $page_div .= "<h2 style=\"font-size:1em;font-weight:normal;\">Page ".$page_num."</h2>\n";       
614    }
615    }
616
617    $page_div .= $inner_div_str;
618    $page_div .= "\n</div>";
619
620    # Finished processing a single html page of the (paged_)pretty_html output generated by
621    # Xpdf's pdftohtml: finished massaging that single html page into the right form
622    return $page_div;
623}
624
625# This subroutine is called to do the PDFv2Plugin post-processing for all cases
626# except the "pretty_html" or "paged_pretty_html" conversion modes.
627# This is what PDFPlugin always used to do:
628sub default_convert_post_process
629{
630    my $self = shift (@_);
631    my ($conv_filename) = @_;
632    my $outhandle=$self->{'outhandle'};
633
634    #$self->{'input_encoding'} = "utf8"; # TODO: The output is always in utf8 (is it?? it is for html, but what about other types?)
635    #my ($language, $encoding) = $self->textcat_get_language_encoding ($conv_filename);
636
637    # read in file ($text will be in utf8)
638    my $text = "";
639    # encoding will be utf8 for html files - what about other types? will we do this step for them anyway?
640    $self->read_file ($conv_filename, "utf8", "", \$text);
641
642    # To support the use_sections option with PDFBox: Greenstone splits PDFs into pages for
643    # sections. The PDFPlugin code wants each new page to be prefixed with <a name=pagenum></a>,
644    # which it then splits on to generate page-based sections. However, that's not what PDFBox
645    # generates in its HTML output. Fortunately, PDFBox does have its own page-separator: it
646    # embeds each page in an extra div. The div opener is:
647    # <div style=\"page-break-before:always; page-break-after:always\">
648    # The PDFPlugin now looks for this and prefixes <a name=0></a> to each such div. (The
649    # pagenumber is fixed at 0 since I'm unable to work out how to increment the pagenum during
650    # a regex substitution even with regex extensions on.) Later, when we process each section
651    # to get the pagenum, PDFBox's output for this is pre-processed by having a loopcounter
652    # that increments the pagenum for each subsequent section.
653
654    #$pdfbox_pageheader="\<div style=\"page-break-before:always; page-break-after:always\">";
655    my $loopcounter = 0; # used later on!
656    $text =~ s@\<div style=\"page-break-before:always; page-break-after:always\">@<a name=$loopcounter></a><div style=\"page-break-before:always; page-break-after:always\">@g;
657
658
659    # Calculate number of pages based on <a ...> tags (we have a <a name=1> etc
660    # for each page).  Metadata based on this calculation not set until process()
661    #
662    # Note: this is done even if we are not breaking the document into pages as it might
663    # be useful to give an indication of document length in browser through setting
664    # num_pages as metadata.
665    # Clean html from low and hight surrogates D800–DFFF
666    $text =~ s@[\N{U+D800}-\N{U+DFFF}]@\ @g;
667    my @pages = ($text =~ m/\<[Aa] name=\"?\w+\"?>/ig); #<div style=\"?page-break-before:always; page-break-after:always\"?>
668    my $num_pages = scalar(@pages);
669    $self->{'num_pages'} = $num_pages;
670
671    if ($self->{'use_sections'}
672    && $self->{'converted_to'} eq "HTML") {
673
674    print $outhandle "PDFv2Plugin: Calculating sections...\n";
675
676    # we have "<a name=1></a>" etc for each page
677    # it may be <A name=
678    my @sections = split('<[Aa] name=', $text);
679
680    my $top_section = "";
681
682    if (scalar (@sections) == 1) { #only one section - no split!
683        print $outhandle "PDFv2Plugin: warning - no sections found\n";
684    } else {
685        $top_section .= shift @sections; # keep HTML header etc as top_section
686    }
687
688    # handle first section specially for title? Or all use first 100...
689   
690    my $title = $sections[0];
691    $title =~ s/^\"?\w+\"?>//; # specific for pdftohtml...
692    $title =~ s/<\/([^>]+)><\1>//g; # (eg) </b><b> - no space
693    $title =~ s/<[^>]*>/ /g;
694    $title =~ s/(?:&nbsp;|\xc2\xa0)/ /g; # utf-8 for nbsp...
695    $title =~ s/^\s+//s;
696    $title =~ s/\s+$//;
697    $title =~ s/\s+/ /gs;
698    $title =~ s/^$self->{'title_sub'}// if ($self->{'title_sub'});
699    $title =~ s/^\s+//s; # in case title_sub introduced any...
700    $title = substr ($title, 0, 100);
701    $title =~ s/\s\S*$/.../;
702
703
704    if (scalar (@sections) == 1) { # no sections found
705        $top_section .= $sections[0];
706        @sections=();
707    } else {
708        $top_section .= "<!--<Section>\n<Metadata name=\"Title\">$title</Metadata>\n-->\n <!--</Section>-->\n";
709    }
710
711    # add metadata per section...
712    foreach my $section (@sections) {
713        # section names are not always just digits, may be like "outline"
714        $section =~ s@^\"?(\w+)\"?></a>@@; # leftover from split expression...
715
716        $title = $1; # Greenstone does magic if sections are titled digits
717
718        # A title of pagenum=0 means use_sections is being applied on output from PDFBox,
719        # which didn't originally have a <a name=incremented pagenumber></a> to split each page.
720        # Our Perl code then prefixed <a name=0></a> to it. Now need to increment the pagenum here:
721        if($loopcounter > 0 || ($title eq 0 && $loopcounter == 0)) { # implies use_sections with PDFBox
722        $title = ++$loopcounter;
723        }
724
725        if (! defined($title) ) {
726        print STDERR "no title: $section\n";
727        $title = " "; # get rid of the undefined warning in next line
728        }
729        my $newsection = "<!-- from PDFv2Plugin -->\n<!-- <Section>\n";
730        $newsection .= "<Metadata name=\"Title\">" . $title
731        . "</Metadata>\n--><br />\n";
732        $newsection .= $section;
733        $newsection .= "<!--</Section>-->\n";
734        $section = $newsection;
735    }
736
737    $text=join('', ($top_section, @sections));
738    }
739
740    if ($self->{'use_sections'}
741    && $self->{'converted_to'} eq "text") {
742    print STDERR "**** When converting PDF to text, cannot apply use_sections\n";
743    }
744
745
746    # The following should no longer be needed, now that strings
747    # read in are Unicode aware (in the Perl sense) rather than
748    # raw binary strings that just happen to be UTF-8 compliant
749
750    # turn any high bytes that aren't valid utf-8 into utf-8.
751##    unicode::ensure_utf8(\$text);
752
753    # Write it out again!
754    $self->utf8_write_file (\$text, $conv_filename);
755}
756
757
758# do plugin specific processing of doc_obj for HTML type
759sub process {
760    my $self = shift (@_);
761    my ($pluginfo, $base_dir, $file, $metadata, $doc_obj, $gli) = @_;
762
763    my $result = $self->process_type($base_dir,$file,$doc_obj);
764
765    # fix up the extracted date metadata to be in Greenstone date format,
766    # and fix the capitalisation of 'date'
767    my $cursection = $doc_obj->get_top_section();
768    foreach my $datemeta (@{$doc_obj->get_metadata($cursection, "date")}) {
769    $doc_obj->delete_metadata($cursection, "date", $datemeta);
770
771    # We're just interested in the date bit, not the time
772    # some pdf creators (eg "Acrobat 5.0 Scan Plug-in for Windows")
773    # set a /CreationDate, and set /ModDate to 000000000. pdftohtml
774    # extracts the ModDate, so it is 0...
775    $datemeta =~ /(\d+)-(\d+)-(\d+)/;
776    my ($year, $month, $day) = ($1,$2,$3);
777    if (defined($year) && defined($month) && defined($day)) {
778        if ($year == 0) {next}
779        if ($year < 100) {$year += 1900} # just to be safe
780        if ($month =~ /^\d$/) {$month="0$month"} # single digit
781        if ($day =~ /^\d$/) {$day="0$day"} # single digit
782        my $date="$year$month$day";
783        $doc_obj->add_utf8_metadata($cursection, "Date", $date);
784    }
785    }
786
787    $doc_obj->add_utf8_metadata($cursection, "NumPages", $self->{'num_pages'}) if defined $self->{'num_pages'};
788   
789    if ($self->{'use_sections'} && $self->{'converted_to'} eq "HTML") {
790    # For gs2 we explicitly make it a paged document, cos greenstone won't get it
791    # right if any section has an empty title, or one with letters in it
792    if (&util::is_gs3()) {
793        # but for gs3, paged docs currently use image slider which is ugly if there are no images
794        $doc_obj->set_utf8_metadata_element ($cursection, "gsdlthistype", "Hierarchy");
795    } else {
796        $doc_obj->set_utf8_metadata_element ($cursection, "gsdlthistype", "Paged");
797    }
798    }
799
800    return $result;
801}
802
8031;
Note: See TracBrowser for help on using the browser.