root/main/trunk/greenstone2/perllib/plugins/ReadTextFile.pm @ 22857

Revision 22857, 19.0 KB (checked in by davidb, 10 years ago)

Further adjustments to our reading in of text files/data to be Unicode aware. These changes address the issue of when the data being read in is in a different encoding. The key 'problem' was that the (now) Unicode aware was being erroneously being re-encoded within HTMLPlugin

  • Property svn:executable set to *
Line 
1###########################################################################
2#
3# ReadTxtFile.pm -- base class for import plugins that have plain text files
4# A component of the Greenstone digital library software
5# from the New Zealand Digital Library Project at the
6# University of Waikato, New Zealand.
7#
8# Copyright (C) 1999-2005 New Zealand Digital Library Project
9#
10# This program is free software; you can redistribute it and/or modify
11# it under the terms of the GNU General Public License as published by
12# the Free Software Foundation; either version 2 of the License, or
13# (at your option) any later version.
14#
15# This program is distributed in the hope that it will be useful,
16# but WITHOUT ANY WARRANTY; without even the implied warranty of
17# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18# GNU General Public License for more details.
19#
20# You should have received a copy of the GNU General Public License
21# along with this program; if not, write to the Free Software
22# Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
23#
24###########################################################################
25
26package ReadTextFile;
27
28use strict;
29no strict 'subs';
30no strict 'refs'; # allow filehandles to be variables and viceversa
31
32use Encode;
33
34use multiread;
35use encodings;
36use unicode;
37use textcat;
38use doc;
39use ghtml;
40use gsprintf 'gsprintf';
41
42use AutoExtractMetadata;
43
44sub BEGIN {
45    @ReadTextFile::ISA = ( 'AutoExtractMetadata' );
46}
47
48my $encoding_plus_auto_list =
49    [ { 'name' => "auto",
50    'desc' => "{ReadTextFile.input_encoding.auto}" } ];
51push(@{$encoding_plus_auto_list},@{$BasePlugin::encoding_list});
52
53my $arguments =
54    [ { 'name' => "input_encoding",
55    'desc' => "{ReadTextFile.input_encoding}",
56    'type' => "enum",
57    'list' => $encoding_plus_auto_list,
58    'reqd' => "no" ,
59    'deft' => "auto" } ,
60      { 'name' => "default_encoding",
61    'desc' => "{ReadTextFile.default_encoding}",
62    'type' => "enum",
63    'list' => $BasePlugin::encoding_list,
64    'reqd' => "no",
65        'deft' => "utf8" },
66      { 'name' => "extract_language",
67    'desc' => "{ReadTextFile.extract_language}",
68    'type' => "flag",
69    'reqd' => "no" },
70      { 'name' => "default_language",
71    'desc' => "{ReadTextFile.default_language}",
72    'type' => "string",
73    'deft' => "en",
74    'reqd' => "no" }
75      ];
76
77
78my $options = { 'name'     => "ReadTextFile",
79        'desc'     => "{ReadTextFile.desc}",
80        'abstract' => "yes",
81        'inherits' => "no",
82        'args'     => $arguments };
83
84
85
86sub new {
87    my $class = shift (@_);
88    my ($pluginlist,$inputargs,$hashArgOptLists, $auxiliary) = @_;
89    push(@$pluginlist, $class);
90
91    push(@{$hashArgOptLists->{"ArgList"}},@{$arguments});
92    push(@{$hashArgOptLists->{"OptList"}},$options);
93
94    my $self = new AutoExtractMetadata($pluginlist, $inputargs, $hashArgOptLists, $auxiliary);
95
96    return bless $self, $class;
97   
98}
99
100
101
102# The ReadTextFile read_into_doc_obj() function. This function does all the
103# right things to make general options work for a given plugin.  It reads in
104# a file and sets up a slew of metadata all saved in doc_obj, which
105# it then returns as part of a tuple (process_status,doc_obj)
106#
107# Much of this functionality used to reside in read, but it was broken
108# down into a supporting routine to make the code more flexible. 
109#
110# recursive plugins (e.g. RecPlug) and specialized plugins like those
111# capable of processing many documents within a single file (e.g.
112# GMLPlug) will normally want to implement their own version of
113# read_into_doc_obj()
114#
115# Note that $base_dir might be "" and that $file might
116# include directories
117sub read_into_doc_obj {
118    my $self = shift (@_); 
119    my ($pluginfo, $base_dir, $file, $block_hash, $metadata, $processor, $maxdocs, $total_count, $gli) = @_;
120
121    my $outhandle = $self->{'outhandle'};
122    # should we move this to read? What about secondary plugins?
123    print STDERR "<Processing n='$file' p='$self->{'plugin_type'}'>\n" if ($gli);
124    print $outhandle "$self->{'plugin_type'} processing $file\n"
125        if $self->{'verbosity'} > 1;
126
127    my ($filename_full_path, $filename_no_path) =  &util::get_full_filenames($base_dir, $file);
128
129    # Do encoding stuff
130    my ($language, $encoding) = $self->textcat_get_language_encoding ($filename_full_path);
131    if ($self->{'verbosity'} > 2) {
132    print $outhandle "ReadTextFile: reading $file as ($encoding,$language)\n";
133    }
134
135    # create a new document
136    my $doc_obj = new doc ($filename_full_path, "indexed_doc", $self->{'file_rename_method'});
137    my $top_section = $doc_obj->get_top_section();
138
139    # this should look at the plugin option too...
140    $doc_obj->add_utf8_metadata($top_section, "Plugin", "$self->{'plugin_type'}");
141    $doc_obj->add_utf8_metadata($top_section, "FileSize", (-s $filename_full_path));
142    $self->set_Source_metadata($doc_obj, $filename_no_path, $encoding);
143
144    $doc_obj->add_utf8_metadata($top_section, "Language", $language);
145    $doc_obj->add_utf8_metadata($top_section, "Encoding", $encoding);
146   
147    # read in file ($text will be in utf8)
148    my $text = "";
149    $self->read_file ($filename_full_path, $encoding, $language, \$text);
150
151    if (!length ($text)) {
152    if ($gli) {
153        print STDERR "<ProcessingError n='$file' r='File contains no text'>\n";
154    }
155    gsprintf($outhandle, "$self->{'plugin_type'}: {ReadTextFile.file_has_no_text}\n", $filename_full_path) if $self->{'verbosity'};
156
157    my $failhandle = $self->{'failhandle'};
158    gsprintf($failhandle, "$file: " . ref($self) . ": {ReadTextFile.empty_file}\n");
159    # print $failhandle "$file: " . ref($self) . ": file contains no text\n";
160    $self->{'num_not_processed'} ++;
161
162    return (0,undef); # what should we return here?? error but don't want to pass it on
163    }
164   
165    # do plugin specific processing of doc_obj
166    unless (defined ($self->process (\$text, $pluginfo, $base_dir, $file, $metadata, $doc_obj, $gli))) {
167    $text = '';
168    undef $text;
169    print STDERR "<ProcessingError n='$file'>\n" if ($gli);
170    return (-1,undef);
171    }
172    $text='';
173    undef $text;
174   
175    # include any metadata passed in from previous plugins
176    # note that this metadata is associated with the top level section
177    $self->add_associated_files($doc_obj, $filename_full_path);
178    $self->extra_metadata ($doc_obj, $top_section, $metadata);
179
180     # do any automatic metadata extraction
181    $self->auto_extract_metadata ($doc_obj);
182
183
184    # if we haven't found any Title so far, assign one
185    $self->title_fallback($doc_obj,$top_section,$filename_no_path);
186
187    $self->add_OID($doc_obj);
188   
189    return (1,$doc_obj);
190}
191
192# uses the multiread package to read in the entire file pointed to
193# by filename and loads the resulting text into $$textref. Input text
194# may be in any of the encodings handled by multiread, output text
195# will be in utf8
196sub read_file {
197    my $self = shift (@_);
198    my ($filename, $encoding, $language, $textref) = @_;
199
200    if (!-r $filename)
201    {
202    my $outhandle = $self->{'outhandle'};
203    gsprintf($outhandle, "{ReadTextFile.read_denied}\n", $filename) if $self->{'verbosity'};
204    # print $outhandle "Read permission denied for $filename\n" if $self->{'verbosity'};
205    return;
206    }
207    $$textref = "";
208    if (!open (FILE, $filename)) {
209    gsprintf(STDERR, "ReadTextFile::read_file {ReadTextFile.could_not_open_for_reading} ($!)\n", $filename);
210    die "\n";
211    }
212     
213    if ($encoding eq "ascii") {
214    undef $/;
215    $$textref = <FILE>;
216    $/ = "\n";
217    } else {
218    my $reader = new multiread();
219    $reader->set_handle ('ReadTextFile::FILE');
220    $reader->set_encoding ($encoding);
221    $reader->read_file ($textref);
222    }
223
224    # At this point $$textref is a binary byte string
225    # => turn it into a Unicode aware string, so full
226    # Unicode aware pattern matching can be used.
227    # For instance: 's/\x{0101}//g' or '[[:upper:]]'
228    #
229
230    $$textref = decode("utf8",$$textref);
231
232    close FILE;
233}
234
235
236# Not currently used
237sub UNUSED_read_file_usingPerlsEncodeModule {
238##sub read_file {
239    my $self = shift (@_);
240    my ($filename, $encoding, $language, $textref) = @_;
241
242    if (!-r $filename)
243    {
244        my $outhandle = $self->{'outhandle'};
245        gsprintf($outhandle, "{ReadTextFile.read_denied}\n", $filename) if $self->{'verbosity'};
246        # print $outhandle "Read permission denied for $filename\n" if $self->{'verbosity'};
247        return;
248    }
249    $$textref = "";
250    if (!open (FILE, $filename)) {
251        gsprintf(STDERR, "ReadTextFile::read_file {ReadTextFile.could_not_open_f
252or_reading} ($!)\n", $filename);
253        die "\n";
254    }
255
256    my $store_slash = $/;
257    undef $/;
258    my $text = <FILE>;
259    $/ = $store_slash;
260
261    $$textref = decode($encoding,$text);
262
263    close FILE;
264}
265
266
267sub read_file_no_decoding {
268    my $self = shift (@_);
269    my ($filename, $textref) = @_;
270
271    if (!-r $filename)
272    {
273    my $outhandle = $self->{'outhandle'};
274    gsprintf($outhandle, "{ReadTextFile.read_denied}\n", $filename) if $self->{'verbosity'};
275    # print $outhandle "Read permission denied for $filename\n" if $self->{'verbosity'};
276    return;
277    }
278    $$textref = "";
279    if (!open (FILE, $filename)) {
280    gsprintf(STDERR, "ReadTextFile::read_file {ReadTextFile.could_not_open_for_reading} ($!)\n", $filename);
281    die "\n";
282    }
283     
284    my $reader = new multiread();
285    $reader->set_handle ('ReadTextFile::FILE');
286    $reader->read_file_no_decoding ($textref);
287   
288    $self->{'reader'} = $reader;
289
290    close FILE;
291}
292
293
294sub decode_text {
295    my $self = shift (@_);
296    my ($raw_text, $encoding, $language, $textref) = @_;
297
298    my $reader = $self->{'reader'};
299    if (!defined $reader) {
300    gsprintf(STDERR, "ReadTextFile::decode_text needs to call ReadTextFile::read_file_no_decoding first\n");
301    }
302    else {
303    $reader->set_encoding($encoding);
304    $reader->decode_text($raw_text,$textref);
305    }
306}
307
308
309sub textcat_get_language_encoding {
310    my $self = shift (@_);
311    my ($filename) = @_;
312
313    my ($language, $encoding, $extracted_encoding);
314    if ($self->{'input_encoding'} eq "auto") {
315        # use textcat to automatically work out the input encoding and language
316        ($language, $encoding) = $self->get_language_encoding ($filename);
317    } elsif ($self->{'extract_language'}) {
318    # use textcat to get language metadata
319        ($language, $extracted_encoding) = $self->get_language_encoding ($filename);
320        $encoding = $self->{'input_encoding'};
321    # don't print this message for english... english in utf8 is identical
322    # to english in iso-8859-1 (except for some punctuation). We don't have
323    # a language model for en_utf8, so textcat always says iso-8859-1!
324        if ($extracted_encoding ne $encoding && $language ne "en" && $self->{'verbosity'}) {
325        my $plugin_name = ref ($self);
326        my $outhandle = $self->{'outhandle'};
327        gsprintf($outhandle, "$plugin_name: {ReadTextFile.wrong_encoding}\n", $filename, $encoding, $extracted_encoding);
328        }
329    } else {
330        $language = $self->{'default_language'};
331        $encoding = $self->{'input_encoding'};
332    }
333   
334#    print STDERR "**** language encoding of contents of file $filename:\n\t****$language $encoding\n";
335
336    return ($language, $encoding);
337}
338
339
340# Uses textcat to work out the encoding and language of the text in
341# $filename. All html tags are removed before processing.
342# returns an array containing "language" and "encoding"
343sub get_language_encoding {
344    my $self = shift (@_);
345    my ($filename) = @_;
346    my $outhandle = $self->{'outhandle'};
347    my $unicode_format = "";
348    my $best_language = "";
349    my $best_encoding = "";
350   
351
352    # read in file
353    if (!open (FILE, $filename)) {
354    gsprintf(STDERR, "ReadTextFile::get_language_encoding {ReadTextFile.could_not_open_for_reading} ($!)\n", $filename);
355    # this is a pretty bad error, but try to continue anyway
356    return ($self->{'default_language'}, $self->{'input_encoding'});
357    }
358    undef $/;
359    my $text = <FILE>;
360    $/ = "\n";
361    close FILE;
362
363    # check if first few bytes have a Byte Order Marker
364    my $bom=substr($text,0,2); # check 16bit unicode
365    if ($bom eq "\xff\xfe") { # little endian 16bit unicode
366    $unicode_format="unicode";
367    } elsif ($bom eq "\xfe\xff") { # big endian 16bit unicode
368    $unicode_format="unicode";
369    } else {
370    $bom=substr($text,0,3); # check utf-8
371    if ($bom eq "\xef\xbb\xbf") { # utf-8 coded FEFF bom
372        $unicode_format="utf8";
373#   } elsif ($bom eq "\xef\xbf\xbe") { # utf-8 coded FFFE bom. Error!?
374#       $unicode_format="utf8";
375    }
376    }
377   
378    my $found_html_encoding = 0;
379    # handle html files specially
380    # XXX this doesn't match plugins derived from HTMLPlug (except ConvertTo)
381    if (ref($self) eq 'HTMLPlugin' ||
382    (exists $self->{'converted_to'} && $self->{'converted_to'} eq 'HTML')){
383
384    # remove comments in head, including multiline ones, so that we don't match on
385    # inactive tags (those that are nested inside comments)
386    my ($head) = ($text =~ m/<head>(.*)<\/head>/si);
387    $head = "" unless defined $head; # some files are not proper HTML eg php files
388    $head =~ s/<!--.*?-->//sg;
389
390    # remove <title>stuff</title> -- as titles tend often to be in English
391    # for foreign language documents
392    $text =~ s!<title>.*?</title>!!si;
393
394    # see if this html file specifies its encoding
395    if ($text =~ /^<\?xml.*encoding="(.+?)"/) {
396        $best_encoding = $1;
397    }
398    # check the meta http-equiv charset tag
399    elsif ($head =~ m/<meta http-equiv.*content-type.*charset=(.+?)\"/si) {           
400        $best_encoding = $1;
401    }
402    if ($best_encoding) { # we extracted an encoding
403        $best_encoding =~ s/-+/_/g;
404        $best_encoding = lc($best_encoding); # lowercase
405        if ($best_encoding eq "utf_8") { $best_encoding = "utf8" }
406        $found_html_encoding = 1;
407        # We shouldn't be modifying this here!!
408        #$self->{'input_encoding'} = $best_encoding;
409    }
410       
411    # remove all HTML tags
412    $text =~ s/<[^>]*>//sg;
413    }
414
415    # don't need to do textcat if we know the encoding now AND don't need to extract language
416    if($found_html_encoding && !$self->{'extract_language'}) { # encoding specified in html file
417    $best_language = $self->{'default_language'};
418    }
419
420    else { # need to use textcat to get either the language, or get both language and encoding
421    $self->{'textcat'} = new textcat() if (!defined($self->{'textcat'}));
422   
423    if($found_html_encoding) { # know encoding, find language by limiting search to known encoding
424        my $results = $self->{'textcat'}->classify_contents_for_encoding(\$text, $filename, $best_encoding);
425       
426        my $language;
427        ($language) = $results->[0] =~ m/^([^-]*)(?:-(?:.*))?$/ if (scalar @$results > 0);
428
429        if (!defined $language || scalar @$results > 3) {
430        # if there were too many results even when restricting results by encoding,
431        # or if there were no results, use default language with the known encoding
432        $best_language = $self->use_default_language($filename);
433        }
434        else { # fewer than 3 results means textcat is more certain, use the first result
435        $best_language = $language;     
436        }
437    }
438    else { # don't know encoding or language yet, therefore we use textcat
439        my $results = $self->{'textcat'}->classify_contents(\$text, $filename);
440       
441        # if textcat returns 3 or less possibilities we'll use the first one in the list
442        if (scalar @$results <= 3) { # results will be > 0 when we don't constrain textcat by an encoding
443        my ($language, $encoding) = $results->[0] =~ m/^([^-]*)(?:-(.*))?$/;
444
445        $language = $self->use_default_language($filename) unless defined $language;
446        $encoding = $self->use_default_encoding($filename) unless defined $encoding;
447
448        $best_language = $language;
449        $best_encoding = $encoding;
450        }
451        else { # if (scalar @$results > 3) {
452        if ($unicode_format) { # in case the first had a BOM
453            $best_encoding=$unicode_format;
454        }
455        else {
456            # Find the most frequent encoding in the textcat results returned
457            # Returns "" if there's no encoding more frequent than another
458            $best_encoding = $self->{'textcat'}->most_frequent_encoding($results);
459        }
460       
461        if ($best_encoding eq "") { # encoding still not set, use defaults
462            $best_language = $self->use_default_language($filename);
463            $best_encoding = $self->use_default_encoding($filename);
464        }
465        elsif (!$self->{'extract_language'}) { # know encoding but don't need to discover language
466            $best_language = $self->use_default_language($filename);
467        }
468        else { # textcat again using the most frequent encoding or the $unicode_format set above
469            $results = $self->{'textcat'}->classify_contents_for_encoding(\$text, $filename, $best_encoding);
470            my $language;
471            ($language) = $results->[0] =~ m/^([^-]*)(?:-(.*))?$/ if (scalar @$results > 0);
472            if (!defined $language || scalar @$results > 3) {
473            # if no result or too many results, use default language for the encoding previously found
474            $best_language = $self->use_default_language($filename);
475            }
476            else { # fewer than 3 results, use the language of the first result
477            $best_language = $language;
478            }
479        }
480        }
481    }
482    }
483
484    if($best_encoding eq "" || $best_language eq "") {
485    print STDERR "****Shouldn't happen: encoding and/or language still not set. Using defaults.\n";
486    $best_encoding = $self->use_default_encoding($filename) if $best_encoding eq "";
487    $best_language = $self->use_default_language($filename) if $best_language eq "";
488    }
489#    print STDERR "****Content language: $best_language; Encoding: $best_encoding.\n";
490
491
492    if ($best_encoding =~ /^iso_8859/ && &unicode::check_is_utf8($text)) {
493    # the text is valid utf8, so assume that's the real encoding
494    # (since textcat is based on probabilities)
495    $best_encoding = 'utf8';
496    }
497
498    # check for equivalents where textcat doesn't have some encodings...
499    # eg MS versions of standard encodings
500    if ($best_encoding =~ /^iso_8859_(\d+)/) {
501    my $iso = $1; # which variant of the iso standard?
502    # iso-8859 sets don't use chars 0x80-0x9f, windows codepages do
503    if ($text =~ /[\x80-\x9f]/) {
504        # Western Europe
505        if ($iso == 1 or $iso == 15) { $best_encoding = 'windows_1252' }
506        elsif ($iso == 2) {$best_encoding = 'windows_1250'} # Central Europe
507        elsif ($iso == 5) {$best_encoding = 'windows_1251'} # Cyrillic
508        elsif ($iso == 6) {$best_encoding = 'windows_1256'} # Arabic
509        elsif ($iso == 7) {$best_encoding = 'windows_1253'} # Greek
510        elsif ($iso == 8) {$best_encoding = 'windows_1255'} # Hebrew
511        elsif ($iso == 9) {$best_encoding = 'windows_1254'} # Turkish
512    }
513    }
514
515    if ($best_encoding !~ /^(ascii|utf8|unicode)$/ &&
516    !defined $encodings::encodings->{$best_encoding}) {
517    if ($self->{'verbosity'}) {
518        gsprintf($outhandle, "ReadTextFile: {ReadTextFile.unsupported_encoding}\n",
519             $filename, $best_encoding, $self->{'default_encoding'});
520    }
521    $best_encoding = $self->{'default_encoding'};
522    }
523
524    return ($best_language, $best_encoding);
525}
526
527
528sub use_default_language {
529    my $self = shift (@_);
530    my ($filename) = @_;
531
532    if ($self->{'verbosity'}>2) {
533    gsprintf($self->{'outhandle'},
534         "ReadTextFile: {ReadTextFile.could_not_extract_language}\n",
535         $filename, $self->{'default_language'});
536    }
537    return $self->{'default_language'};
538}
539
540sub use_default_encoding {
541    my $self = shift (@_);
542    my ($filename) = @_;
543
544    if ($self->{'verbosity'}>2) {
545    gsprintf($self->{'outhandle'},
546         "ReadTextFile: {ReadTextFile.could_not_extract_encoding}\n",
547         $filename, $self->{'default_encoding'});
548    }
549    return $self->{'default_encoding'};
550}
551
552# Overridden by exploding plugins (eg. ISISPlug)
553sub clean_up_after_exploding
554{
555    my $self = shift(@_);
556}
557
558
5591;
Note: See TracBrowser for help on using the browser.