source: other-projects/nightly-tasks/diffcol/trunk/diffcol/gdbdiff.pm@ 29495

Last change on this file since 29495 was 29495, checked in by ak19, 6 years ago

Related to previous commit 29494 accidentally committed under sjs49 and which was prematurely committed. That commit and this one adds a debug flag to the run_test action of the diffcol task, that will store the intermediate debug files in the top level diffcol folder for inspection.

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, $debugging) = @_;
76
77 # print "Now is testing database\n";
78 my ($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 $savepath = &getcwd."/../"; # TASK_HOME env var does not exist at this stage, but it's one level up from current directory
84 if($debugging) {
85 print_string_to_file($test_text, $savepath.$dbname."_test.out1");
86 print_string_to_file($model_text, $savepath.$dbname."_model.out1");
87 }
88
89 # filter out the fields that can be ignored in the two database files
90 # The total_numbytes field can vary depending on how many backslashes exist in the urls in the main body text, as each
91 # of these windows slashes get escaped with another backslash, and the resulting string is used as key into rel link db
92 my $ignore_line_re = "\n<(FileSize|lastmodified|lastmodifieddate|oailastmodified|oailastmodifieddate|ex.File.FileModifyDate|ex.File.FilePermissions|total_numbytes|ex.Composite.LightValue)>([^\n])*";
93 $model_text =~ s/$ignore_line_re//g;
94 $test_text =~ s/$ignore_line_re//g;
95
96 # tmp dirs have subdirs with random numbers in name, remove subdir
97 # these tmpdirs are located inside the collection directory
98 $model_text =~ s@(tmp[\\\/])(\d*[\\\/])@$1@g;
99 $test_text =~ s@(tmp[\\\/])(\d*[\\\/])@$1@g;
100
101 # if the OS doesn't match and one of them is windows, extra work needs to be done to bring the db files
102 # in test and model collection to an even base for comparison
103
104 my $testIsWin = ($test_os ne "compute") ? ($test_os eq "windows") : &isDBWindowsSensitive($dbname, $test_text);
105 my $modelIsWin = ($model_os ne "compute") ? ($model_os eq "windows") : &isDBWindowsSensitive($dbname, $model_text);
106
107 if($testIsWin == $modelIsWin) {
108 # both linux or both windows, do the basic test we did on linux machines:
109 # ignore absolute path prefixes in modelcol and testcol (necessary for archiveinf-doc and -src.gdb files)
110
111 # Remember the original model col on SVN could have been built anywhere,
112 # and in the gdb files, absolute paths are stored to the collection location.
113 # Crop these paths to the collect/<colname> point.
114
115 # Entries are of the form [Entry] or <Entry>. In order to do a sensible diff,
116 # need to remove the prefix to the collect/colname folder in any (absolute) path that occurs in Entry
117 # E.g. [/full/path/collect/colname/import/file.ext] should become [collect/colname/import/file.ext]
118 # Better regex is of the form /BEGIN((?:(?!BEGIN).)*)END/, see http://docstore.mik.ua/orelly/perl/cookbook/ch06_16.htm
119
120 $model_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
121 $test_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
122 #$model_text =~ s@^([^\\//]*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$3$5@mg;
123 #$test_text =~ s@^([^\\//]*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$3$5@mg;
124 }
125
126 else { # one of the collections was built on windows
127 # handling slashes and other differences between a model coll built on one OS (e.g. linux)
128 # and a test collection built and diffed on another OS (windows)
129
130 my ($win_text, $lin_text); # references
131 my $collection_path = $strTestCol; # full path to a windows collection
132
133 if($testIsWin) {
134 $collection_path = $strTestCol; # test collection path is windows
135 $win_text = \$test_text;
136 $lin_text = \$model_text;
137 } else {
138 $collection_path = $strModelCol; # model collection path is windows
139 $win_text = \$model_text;
140 $lin_text = \$test_text;
141 }
142
143 if($dbname =~ m/archiveinf-doc/) {
144
145 (my $collection_path_re = $collection_path) =~ s@\\@\\\\@g;
146
147 my $tmp = ""; # rebuild windows file's set of lines after processing them one by one
148
149 # convert short filenames to long perl:
150 # http://www.mombu.com/programming/perl/t-convert-dos-83-filenames-to-win32-long-filenames-using-perl-525448.html
151 for my $line (split /^/, $$win_text) { # split the string into newlines
152
153 # assoc-file and meta-file contain filepaths, ensure these are long windows file paths now (will later convert to linux slashes)
154 if($line =~ m@^<(assoc-file|meta-file|src-file)>(.*)(\s+)@s) {
155 my ($field, $value, $suffix) = ($1, $2, $3);
156 $line = $value; # it may be a short file name
157
158 if($line !~ m/^\@/) { # if the path doesn't use a "relative" @GSPATH@ placeholder string, but is an absolute path instead
159 # perhaps test here if it is a shortfilename? should match /CAPS....~number(.ext)/
160 $line = "<$field>".&Win32::GetLongPathName($line)."$suffix"; # make it a long file name and prefix assoc-file/meta-file tagname to it again
161 }
162 else { # if $line contains @THISCOLLECTIONPATH@, still need to deal with DOS filenames suffixes:
163 # replace placeholder with absolute path and expand to long filename, then insert placeholder in its original place again
164 $line =~ s/\@THISCOLLECTPATH\@/$collection_path/;
165 $line = &Win32::GetLongPathName($line);
166 $line =~ s/^$collection_path_re/\@THISCOLLECTPATH\@/;
167 $line = "<$field>".$line."$suffix";
168 }
169 }
170 $tmp .= $line;
171 }
172 $$win_text = $tmp;
173 }
174
175
176 # index gdb file
177 if($dbname =~ m/$strColName/) {
178 my $tmp = ""; # rebuild windows file's set of lines after processing them one by one
179 for my $line (split /^/, $$win_text) { # split the string into newlines
180
181 # In the following regex, add any .gdb fieldnames that represent a path and so would contain double backslashes
182 # on Windows (to escape the single backlash of win filepaths). They will be turned into single-backslashes here,
183 # and converted into single forward slashes futher below when the txt version of the win gdb file is normalised
184 # to compare it with the linux version.
185 # E.g. On windows, the Word-PDF collection(s) contains double backslashes in the ex.File.Directory field
186 # the MARC-Exploded collection contains double backslashes in the null_file entry field of the .gdb file
187 if($line =~ m@^<(ex.File.Directory|null_file)>(.*)@s) {
188 my ($fieldname, $escaped_path) = ($1, $2);
189 $escaped_path =~ s@\\\\@/@g; #(my $escaped_path = $2) =~ s@\\\\@\\@g;
190 $line = "<$fieldname>$escaped_path";
191 }
192 elsif($line =~ m@^<Title>(.*)@s) {
193# print STDERR "***** TITLE: |$1|\n";
194
195 # word-pdf collection: Title of ps files contain new lines at end when
196 # GreenstoneXMLPlugin::xml_end_tag() writes the Title back out after utf8 decode
197 # if($metadata_name eq "Title") { $metadata_value =~ s/[\n\r]*$//; }
198
199 (my $title = $1) =~ s@(\r|\n|\\n)*$@@; # get rid of trailing newlines/carriage returns
200 $line = "<Title>$title\n"; # add single newline
201 }
202 $tmp .= $line;
203 }
204 $$win_text = $tmp;
205
206 # slashes in windows metadata text need to be turned into linux style slashes.
207 # index\col.gdb uses double backslashes, and single for \n,\t
208 #$$win_text =~ s@\\\\@/@g;
209 }
210 else { # archiveinf gdb file
211
212 # slashes in windows metadata text need to be turned into linux style slashes.
213 # In the two archivesinf gdb files, filepaths may use single backslashes
214 $$win_text =~ s@\\@/@g; #$$win_text =~ s@\\([^n|r|\|"])@/$1@g; # filepath something\rtf remains something\rtf
215 }
216
217 # cut down absolute paths to files to just collect/colname/.../file, same as before
218 $$lin_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg; # $$lin_text =~ s@^([^\\\/]*(//)?).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
219 $$win_text =~ s@^([^\\/]*(//)*).*(\\|/)(collect(\\|/)$strColName)(.*)$@$1$4$6@mg;
220
221 # for the windows text, need to further get rid of the driveletter after [ or <meta>
222 $$win_text =~ s@^(\[|<[^>]*>)[a-zA-Z]:collect@$1collect@mg;
223
224 } # end of equalising differences between a windows collection's db file and linux coll's db file
225
226 # The following block of code is necessary to deal with tmp (html) source files generated when using PDFBox
227 # These tmpdirs are located inside the toplevel *greenstone* directory
228 (my $gsdlhome_re = $ENV{'GSDLHOME'}) =~ s@\\@\/@g;
229 $gsdlhome_re = ".*" unless $$ENV{'GSDLHOME'};
230 my $tmpfile_regex = "<URL>http://$gsdlhome_re/tmp/([^\.]*?)(\..{3,4})"; # $gsdlhome/tmp/randomfilename.html, file ext can be 3 or 4 chars long
231 if($test_text =~ m@$tmpfile_regex@g) {
232 # found a match, replace the tmp file name with "random", keeping the original file extension
233 # in <OrigSource|URL|UTF8URL|gsdlconvertedfilename>
234
235 # This code is slightly different from doc.xml because each document has its own doc.xml, so this needs to be done
236 # only once for doc.xml, but multiple times in index/col.gdb since it contains the random filenames of all docs in the col
237 #my ($old_tmp_filename, $ext) = ($1, $2);
238
239 my $new_tmp_filename = "random";
240
241
242 $tmpfile_regex = "(<(URL|UTF8URL|gsdlconvertedfilename|OrigSource)>(http://)?)($gsdlhome_re)?(/tmp/)?.*?(\..{3,4})";
243 if($5) {
244 $test_text =~ s@$tmpfile_regex@$1$5$new_tmp_filename$6@mg;
245 } else { # OrigSource contains only the filename
246 $test_text =~ s@$tmpfile_regex@$1$new_tmp_filename$6@mg;
247 }
248
249 # modelcol used a different gsdlhome, but also a tmp dir, so make the same changes to its random filename
250 $tmpfile_regex = "(<(URL|UTF8URL|gsdlconvertedfilename|OrigSource)>(http://)?)(.*)?(/tmp/)?.*?(\..{3,4})";
251 if($5) {
252 $model_text =~ s@$tmpfile_regex@$1$5$new_tmp_filename$6@mg;
253 } else { # OrigSource contains only the filename
254 $model_text =~ s@$tmpfile_regex@$1$new_tmp_filename$6@mg;
255 }
256
257 # index/col.gdb also has entries for the random tmp file names in the form: [http://research/ak19/GS2bin_5July2013/tmp/F639.html]
258 # need to equalise these also. Sadly, when there are multiple intermediate files, their random tmp filenames are not
259 # guaranteed to be generated in the same (alphabetical/numerical) order between model and test collection, so the
260 # HASH OIDs, although all of them accounted for, appear in a different order. So we have to remove the Hash OIDs.
261 #$test_text =~ s@\[http://.*/tmp/.*(\..{3,4})\]@tmp/random$1@mg; # HASH OIDs can appear in different order
262 #$model_text =~ s@\[http://.*/tmp/.*(\..{3,4})\]@tmp/random$1@mg;
263
264 $test_text =~ s@\[http://[^\n]*?/tmp/.*?(\..{3,4})\]\n<section>([^\n]*?)\n@[tmp/random$1\n<section>$2]\n@sg;
265 $model_text =~ s@\[http://[^\n]*?/tmp/.*?(\..{3,4})\]\n<section>([^\n]*?)\n@[tmp/random$1\n<section>$2]\n@sg;
266
267 # need to re- sort the keys, now that the absolute paths to tmp locations has been removed
268 # so that we get the tmp files in the same order in both model and test collections
269
270 $model_text = text_to_db_to_text($model_text, "model.gdb");
271 $test_text = text_to_db_to_text($test_text, "test.gdb");
272 }
273
274 # now can go back to using $model_text and $test_text
275
276 if($debugging) {
277 print_string_to_file($test_text, $savepath.$dbname."_test.out");
278 print_string_to_file($model_text, $savepath.$dbname."_model.out");
279 }
280
281 my $report_type = "OldStyle"; # Can not change this type.
282 my $diff_gdb = diff \$model_text, \$test_text, { STYLE => $report_type };
283
284 # leaving the ignore regex as it used to be in the following, in case it helps with single line comparisons
285 $diff_gdb = &diffutil::GenerateOutput($diff_gdb,"^<(lastmodified|lastmodifieddate|oailastmodified|oailastmodifieddate|ImageSize)>.*");
286
287 if($diff_gdb eq "")
288 {
289 return "";
290 }
291 else
292 {
293 return "Difference Report: Differences found in the Database file: \n$diff_gdb";
294 }
295 # Call diff?
296}
297
298# returns true if the contents are windows AND it matters for the diffing on the db that it's windows
299# 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
300sub isDBWindowsSensitive
301{
302 my ($dbtailname, $db_contents) = @_; # db filename without suffix
303
304# return ($db_contents =~ m/\\/) ? 1 : 0; # windows slashes detected. Better test would be: [Something\something] OR <tag>something\something
305
306 if($dbtailname =~ m/^archiveinf-doc/) {
307 return ($db_contents =~ m@<src-file>[a-zA-Z]:\\@) ? 1 : 0; # <src-file>C:\path
308 }
309 elsif($dbtailname =~ m/^archiveinf-src/) { # <src-file>C:\path
310 return ($db_contents =~ m@\[[a-zA-Z]:\\@) ? 1 : 0; # [C:\path]
311 }
312 else { # index/col.gdb file
313 if ($db_contents =~ m@<URL>http://[a-zA-Z]:/@) { # <URL>http://C:/path
314 return 1;
315 }
316 elsif ($db_contents =~ m@^(<URL>http://[a-zA-Z]:/)|(<null_file>[^\\]*\\)@m) { # <URL>http://C:/path OR <null_file>CMSwp-all.00000001\\00000035.nul
317 return 1;
318 }
319 elsif ($db_contents =~ m@^(<ex.File.Directory>[a-zA-Z]:\\\\)@m) { # <ex.File.Directory>C:\\path\\path for OAI collection
320 return 1;
321 }
322 return 0;
323 }
324}
325
3261;
Note: See TracBrowser for help on using the repository browser.