Protection against too long single-word titles

Titles consisting of a single word exceeding the max word length limit
are not indexed (similarly to how that long word would be handled if it
were not the only word in the title).

An alternative might be to index a truncated version of the title but
that would complicate the verification of the bug fix (because the
potentially dangerous title would remain in the expected output of the
unit test).

Author: veloman-yunkan

src/writer/xapianIndexer.cpp

@@ -130,9 +130,12 @@ size_t getTermCount(const Xapian::Document& d)
 
 void XapianIndexer::indexTitle(const std::string& path, const std::string& title, const std::string& targetPath)
 {
+  const size_t MAX_WORD_LENGTH = 64;
+
   assert(indexingMode == IndexingMode::TITLE);
   Xapian::Stem stemmer;
   Xapian::TermGenerator indexer;
+  indexer.set_max_word_length(MAX_WORD_LENGTH);
   indexer.set_flags(Xapian::TermGenerator::FLAG_CJK_NGRAM);
   try {
     stemmer = Xapian::Stem(stemmer_language);
@@ -162,7 +165,9 @@ void XapianIndexer::indexTitle(const std::string& path, const std::string& title
       // only ANCHOR_TERM was added, hence unaccentedTitle is made solely of
       // non-word characters. Then add entire title as a single term.
       currentDocument.remove_term(*currentDocument.termlist_begin());
-      currentDocument.add_term(unaccentedTitle);
+      if ( unaccentedTitle.size() <= MAX_WORD_LENGTH ) {
+        currentDocument.add_term(unaccentedTitle);
+      }
     }
   }
 

test/suggestion.cpp

@@ -753,11 +753,10 @@ TEST(Suggestion, titleEdgeCases) {
   );
 
   EXPECT_SUGGESTED_TITLES(archive, "awordthatis",
-      w65, // a very long word slips in when it is the only word of a title
       w64,
       "Is " + w64 + " too long?"
-      // "Is " + w65 + " too long?" isn't included because w65 has been ignored
-      //                            during indexing
+      // w65 and "Is " + w65 + " too long?" aren't included because w65 has
+      //                                    been ignored during indexing
   );
 }