source: other-projects/nightly-tasks/diffcol/trunk/diffcol/gdbdiff.pm@ 28661

Last change on this file since 28661 was 28661, checked in by ak19, 7 years ago

Committing the next installment of code to handle diffcol for GS3. Now it successfully compiles up GS3, while diffcol still works for GS2.

File size: 14.2 KB
Line 
1package gdbdiff;
2
3BEGIN {
4 die "GSDLHOME not set\n" unless defined $ENV{'GSDLHOME'};
5 die "GSDLOS not set\n" unless defined $ENV{'GSDLOS'};
6 unshift (@INC, "$ENV{'GSDLHOME'}/perllib");
7 unshift (@INC, "$ENV{'GSDLHOME'}/perllib/cpan");
8}
9
10use util;
11use diffutil;
12use Text::Diff;
13use Cwd;
14
15if ($^O =~ m/mswin/i) {
16 require Win32; # for working out Windows Long Filenames from Win 8.3 short filenames
17}
18
19sub readin_gdb
20{
21 my ($cmd) = @_;
22
23 open(PIN,"$cmd|")
24 || die "Unable to open pipe to $cmd: $!\n";
25
26 my $text_content = "";
27
28 while (defined (my $line = <PIN>)) {
29 $text_content .= $line;
30 }
31
32 close(PIN);
33 return $text_content;
34}
35
36sub read_db
37{
38 # need to sort text output of both test and model col database files, to normalise them for the comparison
39 # the -sort option to db2txt was added specifically to support diffcol
40 my($db_file) = @_;
41 my $db_cmd = "db2txt -sort $db_file 2>&1";
42 if($db_file =~ m/\.jdb$/) {
43 print STDERR "NOT YET IMPLEMENTED\n";
44 #$db_cmd = "jdb2txt -sort $db_file 2>&1";
45 }
46
47 my $db_text = readin_gdb($db_cmd);
48 return $db_text;
49}
50
51sub text_to_db_to_text
52{
53 my($db_text, $db_filename) = @_;
54
55 # http://stackoverflow.com/questions/1909262/how-can-i-pipe-input-into-a-java-command-from-perl
56 open PIPE, "| txt2db $db_filename";
57 print PIPE "$db_text";
58 close(PIPE);
59
60 return &read_db("$db_filename");
61}
62
63# for debugging. Prints txt contents of db to file
64sub print_string_to_file
65{
66 my ($text, $outfile) = @_;
67
68 open(FOUT, ">$outfile") or die "ERROR failed to write to $outfile: $!\n";
69 print FOUT $text;
70 close(FOUT);
71}
72
73sub test_gdb
74{
75 my ($full_modeldb, $full_testdb, $strColName, $test_os, $model_os, $strTestCol, $strModelCol) = @_;
76
77 # print "Now is testing database\n";
78 my ($dbname, $dirname, $suffix)= &File::Basename::fileparse($full_testdb, "\\.[^\\.]+\$");
79
80 my $model_text = read_db($full_modeldb);
81 my $test_text = read_db($full_testdb);
82
83# my $savepath = &getcwd."/../"; # TASK_HOME env var does not exist at this stage, but it's one level up from current directory
84# print_string_to_file($test_text, $savepath.$dbname."_test.out1");
85# print_string_to_file($model_text, $savepath.$dbname."_model.out1");
86
87 # filter out the fields that can be ignored in the two database files
88 # The total_numbytes field can vary depending on how many backslashes exist in the urls in the main body text, as each
89 # of these windows slashes get escaped with another backslash, and the resulting string is used as key into rel link db
90 my $ignore_line_re = "\n<(FileSize|lastmodified|lastmodifieddate|oailastmodified|oailastmodifieddate|ex.File.FileModifyDate|ex.File.FilePermissions|total_numbytes|ex.Composite.LightValue)>([^\n])*";
91 $model_text =~ s/$ignore_line_re//g;
92 $test_text =~ s/$ignore_line_re//g;
93
94 # tmp dirs have subdirs with random numbers in name, remove subdir
95 # these tmpdirs are located inside the collection directory
96 $model_text =~ s@(tmp[\\\/])(\d*[\\\/])@$1@g;
97 $test_text =~ s@(tmp[\\\/])(\d*[\\\/])@$1@g;
98
99 # if the OS doesn't match and one of them is windows, extra work needs to be done to bring the db files
100 # in test and model collection to an even base for comparison
101
102 my $testIsWin = ($test_os ne "compute") ? ($test_os eq "windows") : &isDBWindowsSensitive($dbname, $test_text);
103 my $modelIsWin = ($model_os ne "compute") ? ($model_os eq "windows") : &isDBWindowsSensitive($dbname, $model_text);
104
105 if($testIsWin == $modelIsWin) {
106 # both linux or both windows, do the basic test we did on linux machines:
107 # ignore absolute path prefixes in modelcol and testcol (necessary for archiveinf-doc and -src.gdb files)
108
109 # Remember the original model col on SVN could have been built anywhere,
110 # and in the gdb files, absolute paths are stored to the collection location.
111 # Crop these paths to the collect/<colname> point.
112
113 # Entries are of the form [Entry] or <Entry>. In order to do a sensible diff,
114 # need to remove the prefix to the collect/colname folder in any (absolute) path that occurs in Entry
115 # E.g. [/full/path/collect/colname/import/file.ext] should become [collect/colname/import/file.ext]
116 # Better regex is of the form /BEGIN((?:(?!BEGIN).)*)END/, see http://docstore.mik.ua/orelly/perl/cookbook/ch06_16.htm
117
118 $model_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
119 $test_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
120 #$model_text =~ s@^([^\\//]*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$3$5@mg;
121 #$test_text =~ s@^([^\\//]*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$3$5@mg;
122 }
123
124 else { # one of the collections was built on windows
125 # handling slashes and other differences between a model coll built on one OS (e.g. linux)
126 # and a test collection built and diffed on another OS (windows)
127
128 my ($win_text, $lin_text); # references
129 my $collection_path = $strTestCol; # full path to a windows collection
130
131 if($testIsWin) {
132 $collection_path = $strTestCol; # test collection path is windows
133 $win_text = \$test_text;
134 $lin_text = \$model_text;
135 } else {
136 $collection_path = $strModelCol; # model collection path is windows
137 $win_text = \$model_text;
138 $lin_text = \$test_text;
139 }
140
141 if($dbname =~ m/archiveinf-doc/) {
142
143 (my $collection_path_re = $collection_path) =~ s@\\@\\\\@g;
144
145 my $tmp = ""; # rebuild windows file's set of lines after processing them one by one
146
147 # convert short filenames to long perl:
148 # http://www.mombu.com/programming/perl/t-convert-dos-83-filenames-to-win32-long-filenames-using-perl-525448.html
149 for my $line (split /^/, $$win_text) { # split the string into newlines
150
151 # assoc-file and meta-file contain filepaths, ensure these are long windows file paths now (will later convert to linux slashes)
152 if($line =~ m@^<(assoc-file|meta-file|src-file)>(.*)(\s+)@s) {
153 my ($field, $value, $suffix) = ($1, $2, $3);
154 $line = $value; # it may be a short file name
155
156 if($line !~ m/^\@/) { # if the path doesn't use a "relative" @GSPATH@ placeholder string, but is an absolute path instead
157 # perhaps test here if it is a shortfilename? should match /CAPS....~number(.ext)/
158 $line = "<$field>".&Win32::GetLongPathName($line)."$suffix"; # make it a long file name and prefix assoc-file/meta-file tagname to it again
159 }
160 else { # if $line contains @THISCOLLECTIONPATH@, still need to deal with DOS filenames suffixes:
161 # replace placeholder with absolute path and expand to long filename, then insert placeholder in its original place again
162 $line =~ s/\@THISCOLLECTPATH\@/$collection_path/;
163 $line = &Win32::GetLongPathName($line);
164 $line =~ s/^$collection_path_re/\@THISCOLLECTPATH\@/;
165 $line = "<$field>".$line."$suffix";
166 }
167 }
168 $tmp .= $line;
169 }
170 $$win_text = $tmp;
171 }
172
173
174 # index gdb file
175 if($dbname =~ m/$strColName/) {
176 my $tmp = ""; # rebuild windows file's set of lines after processing them one by one
177 for my $line (split /^/, $$win_text) { # split the string into newlines
178
179 # In the following regex, add any .gdb fieldnames that represent a path and so would contain double backslashes
180 # on Windows (to escape the single backlash of win filepaths). They will be turned into single-backslashes here,
181 # and converted into single forward slashes futher below when the txt version of the win gdb file is normalised
182 # to compare it with the linux version.
183 # E.g. On windows, the Word-PDF collection(s) contains double backslashes in the ex.File.Directory field
184 # the MARC-Exploded collection contains double backslashes in the null_file entry field of the .gdb file
185 if($line =~ m@^<(ex.File.Directory|null_file)>(.*)@s) {
186 my ($fieldname, $escaped_path) = ($1, $2);
187 $escaped_path =~ s@\\\\@/@g; #(my $escaped_path = $2) =~ s@\\\\@\\@g;
188 $line = "<$fieldname>$escaped_path";
189 }
190 elsif($line =~ m@^<Title>(.*)@s) {
191# print STDERR "***** TITLE: |$1|\n";
192
193 # word-pdf collection: Title of ps files contain new lines at end when
194 # GreenstoneXMLPlugin::xml_end_tag() writes the Title back out after utf8 decode
195 # if($metadata_name eq "Title") { $metadata_value =~ s/[\n\r]*$//; }
196
197 (my $title = $1) =~ s@(\r|\n|\\n)*$@@; # get rid of trailing newlines/carriage returns
198 $line = "<Title>$title\n"; # add single newline
199 }
200 $tmp .= $line;
201 }
202 $$win_text = $tmp;
203
204 # slashes in windows metadata text need to be turned into linux style slashes.
205 # index\col.gdb uses double backslashes, and single for \n,\t
206 #$$win_text =~ s@\\\\@/@g;
207 }
208 else { # archiveinf gdb file
209
210 # slashes in windows metadata text need to be turned into linux style slashes.
211 # In the two archivesinf gdb files, filepaths may use single backslashes
212 $$win_text =~ s@\\@/@g; #$$win_text =~ s@\\([^n|r|\|"])@/$1@g; # filepath something\rtf remains something\rtf
213 }
214
215 # cut down absolute paths to files to just collect/colname/.../file, same as before
216 $$lin_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg; # $$lin_text =~ s@^([^\\\/]*(//)?).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
217 $$win_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
218
219 # for the windows text, need to further get rid of the driveletter after [ or <meta>
220 $$win_text =~ s@^(\[|<[^>]*>)[a-zA-Z]:collect@$1collect@mg;
221
222 } # end of equalising differences between a windows collection's db file and linux coll's db file
223
224 # The following block of code is necessary to deal with tmp (html) source files generated when using PDFBox
225 # These tmpdirs are located inside the toplevel *greenstone* directory
226 (my $gsdlhome_re = $ENV{'GSDLHOME'}) =~ s@\\@\/@g;
227 $gsdlhome_re = ".*" unless $$ENV{'GSDLHOME'};
228 my $tmpfile_regex = "<URL>http://$gsdlhome_re/tmp/([^\.]*?)(\..{3,4})"; # $gsdlhome/tmp/randomfilename.html, file ext can be 3 or 4 chars long
229 if($test_text =~ m@$tmpfile_regex@g) {
230 # found a match, replace the tmp file name with "random", keeping the original file extension
231 # in <OrigSource|URL|UTF8URL|gsdlconvertedfilename>
232
233 # This code is slightly different from doc.xml because each document has its own doc.xml, so this needs to be done
234 # only once for doc.xml, but multiple times in index/col.gdb since it contains the random filenames of all docs in the col
235 #my ($old_tmp_filename, $ext) = ($1, $2);
236
237 my $new_tmp_filename = "random";
238
239
240 $tmpfile_regex = "(<(URL|UTF8URL|gsdlconvertedfilename|OrigSource)>(http://)?)($gsdlhome_re)?(/tmp/)?.*?(\..{3,4})";
241 if($5) {
242 $test_text =~ s@$tmpfile_regex@$1$5$new_tmp_filename$6@mg;
243 } else { # OrigSource contains only the filename
244 $test_text =~ s@$tmpfile_regex@$1$new_tmp_filename$6@mg;
245 }
246
247 # modelcol used a different gsdlhome, but also a tmp dir, so make the same changes to its random filename
248 $tmpfile_regex = "(<(URL|UTF8URL|gsdlconvertedfilename|OrigSource)>(http://)?)(.*)?(/tmp/)?.*?(\..{3,4})";
249 if($5) {
250 $model_text =~ s@$tmpfile_regex@$1$5$new_tmp_filename$6@mg;
251 } else { # OrigSource contains only the filename
252 $model_text =~ s@$tmpfile_regex@$1$new_tmp_filename$6@mg;
253 }
254
255 # index/col.gdb also has entries for the random tmp file names in the form: [http://research/ak19/GS2bin_5July2013/tmp/F639.html]
256 # need to equalise these also. Sadly, when there are multiple intermediate files, their random tmp filenames are not
257 # guaranteed to be generated in the same (alphabetical/numerical) order between model and test collection, so the
258 # HASH OIDs, although all of them accounted for, appear in a different order. So we have to remove the Hash OIDs.
259 #$test_text =~ s@\[http://.*/tmp/.*(\..{3,4})\]@tmp/random$1@mg; # HASH OIDs can appear in different order
260 #$model_text =~ s@\[http://.*/tmp/.*(\..{3,4})\]@tmp/random$1@mg;
261
262 $test_text =~ s@\[http://[^\n]*?/tmp/.*?(\..{3,4})\]\n<section>([^\n]*?)\n@[tmp/random$1\n<section>$2]\n@sg;
263 $model_text =~ s@\[http://[^\n]*?/tmp/.*?(\..{3,4})\]\n<section>([^\n]*?)\n@[tmp/random$1\n<section>$2]\n@sg;
264
265 # need to re- sort the keys, now that the absolute paths to tmp locations has been removed
266 # so that we get the tmp files in the same order in both model and test collections
267
268 $model_text = text_to_db_to_text($model_text, "model.gdb");
269 $test_text = text_to_db_to_text($test_text, "test.gdb");
270 }
271
272 # now can go back to using $model_text and $test_text
273# print_string_to_file($test_text, $savepath.$dbname."_test.out");
274# print_string_to_file($model_text, $savepath.$dbname."_model.out");
275
276 my $report_type = "OldStyle"; # Can not change this type.
277 my $diff_gdb = diff \$model_text, \$test_text, { STYLE => $report_type };
278
279 # leaving the ignore regex as it used to be in the following, in case it helps with single line comparisons
280 $diff_gdb = &diffutil::GenerateOutput($diff_gdb,"^<(lastmodified|lastmodifieddate|oailastmodified|oailastmodifieddate|ImageSize)>.*");
281
282 if($diff_gdb eq "")
283 {
284 return "";
285 }
286 else
287 {
288 return "Difference Report: Differences found in the Database file: \n$diff_gdb";
289 }
290 # Call diff?
291}
292
293# returns true if the contents are windows AND it matters for the diffing on the db that it's windows
294# For col.gdb it does not seem to matter so far, if it is generated on a windows machine and to be compared to a linux-generated col.gdb
295sub isDBWindowsSensitive
296{
297 my ($dbtailname, $db_contents) = @_; # db filename without suffix
298
299# return ($db_contents =~ m/\\/) ? 1 : 0; # windows slashes detected. Better test would be: [Something\something] OR <tag>something\something
300
301 if($dbtailname =~ m/^archiveinf-doc/) {
302 return ($db_contents =~ m@<src-file>[a-zA-Z]:\\@) ? 1 : 0; # <src-file>C:\path
303 }
304 elsif($dbtailname =~ m/^archiveinf-src/) { # <src-file>C:\path
305 return ($db_contents =~ m@\[[a-zA-Z]:\\@) ? 1 : 0; # [C:\path]
306 }
307 else { # index/col.gdb file
308 if ($db_contents =~ m@<URL>http://[a-zA-Z]:/@) { # <URL>http://C:/path
309 return 1;
310 }
311 elsif ($db_contents =~ m@^(<URL>http://[a-zA-Z]:/)|(<null_file>[^\\]*\\)@m) { # <URL>http://C:/path OR <null_file>CMSwp-all.00000001\\00000035.nul
312 return 1;
313 }
314 elsif ($db_contents =~ m@^(<ex.File.Directory>[a-zA-Z]:\\\\)@m) { # <ex.File.Directory>C:\\path\\path for OAI collection
315 return 1;
316 }
317 return 0;
318 }
319}
320
3211;
Note: See TracBrowser for help on using the repository browser.