Author: Maxime

A new version, 4.50, of DataparkSearch Engine has been released. Changes since version 4.49 are:

• Default value for PopRankSkipSameSite command has been changed to "yes".
• Possible memory leak has been fixed for a sub-document indexed from stored database.
• The strict option has been added for Section command.
• A word break has been added for French-style contractions.
• Big lists of Russian and English synonyms have been added.
• MaxSiteLevel command accept now a negative argument to group URLs on subdirectory basis.
• The SkipUnreferred command has been extended to delete unreferred documents if necessary.
• Del log processing has been fixed in splitter for case when cache log is empty.
• Some German letters automatically replace by bi-letter combinations in accent-free search mode.
  ß -> ss, ä -> ae, ö -> oe, ü -> ue.
• SQLite3 support has been added. Use --with-sqlite3 option for configure to enable it.
• Indexing has been fixed for documents with several versions in different languages. You need to execute "indexer -Erehashstored" command when upgrade.
• HTML parser understands now <!-- google_ad_section_start -->, <!-- google_ad_section_start(weight=ignore) --> and
  <!-- google_ad_section_end --> comments as tags to include/exclude content for indexing.
• Relevance calculation has been improved for case when acronyms and abbreviations are used.

A comparison between public-domain search engines (PDF, 20 pages) as of March 2006. They compare Nutch, MnoGoSearch, DataparkSearch and ht//Dig.

UPDATE: There is an another comparison of DataparkSearch made in October 2007: Yury Vetrov's blog (in Russian). They compare Sphinx, Tsearch2, DataparkSearch, MnoGoSearch and Yandex.Server (free and commercial versions). Overall weighted rating for these search engines is:

1. Sphinx (4,48);
2. DataparkSearch (4,33);
3. Yandex.Server, free version (4,24);
4. Yadex.Server,commercial version (4,09);
5. MnoGoSearch (3,82);
6. Tsearch2 (3,45).

The "strict" option has been added for the Section command. By specifying this option you can set strict string tokenization for a section, which mean a word break at any non-character symbol despite the context. This is useful, for example, in indexing of URL where hyphen character uses as delimiter between words.

Section url  3 0 strict

There is a mistake in this shapshot which prevent compilation:

socket.c: In function 'socket_connect':
socket.c:93: error: 'struct sockaddr_in' has no member named 'sin_len'

To solve the problem, it need to comment in this, 93rd, line of socket.c. After this correction, the snapshot compiles and works smoothly.