DEVELOPING FOR OTHER DATABASES

Users of DBIx::TextSearch only use the generic methods provided in TextSearch.pm (and documented above) to interact with the index. Some of these methods then use database specific subroutines internally. These subroutines are provided by modules written separately for each database (such as DBIx::TextSearch::Pg).

See the code and comments in TextSearch/Pg.pm for further details.

*Danger Will Robinson* The code itself is probably more recent than this documentation

Requirements

• A database engine supported by the perl DBI which is capable of matching queries across several tables.

• The database engine must also have a corresponding Text::Query::BuildSQL package (unless you want to write your own text query->SQL parser).

Generic methods with a database specific component

Non-specific methods are all in lowercase like_this, database specific subroutines are all in studly caps LikeThis, private methods are prefixed with an underscore _like_this.

• new. Passes the index object (database handle and index name) to CreateIndex which creates the database tables for this index.

• find_document. Passes a single query (scalar) to GetQuery which runs the query through a Text::Query::ParseAdvanced parser and returns an SQL statement representing that query.

• _store_plain and _store_html. These are both invoked from index_file, they pass the URI, title, description and document contents (all scalars) into IndexFile. IndexFile doesn't return anything.

Database specific methods

• CreateIndex - see above for parameters. Creates database tables of the following structure (data types are all Postgres ones)

  Assume the index is called shme

   Table: shme_docID
          |-> URI          varchar(255)
          |-> title        varchar(100)
          \-> d_ID         int4
  
   Table: shme_words
          |-> w_ID         int4
          \-> word         text
  
   Table: shme_mets
          |-> m_ID         int4
          \-> description  text

  Returns nothing, croaks if unable to create table.

• FlushIndex - takes a DBIx::TextSearch object, and deletes all data stored within that index.

• GetQuery - see above for description

• IndexFile - Takes the index object, document URI, title, description and content as input.

  Finds a unique document ID number to use as d_ID, w_ID and m_ID by querying the database for the highest one used. Executes the following storage statements

   # URI, title and doc_ID
   "insert into $self->{name}_doc_ID (URI, title, d_ID) values 
   ($uri, $title, $docid)"
  
   # words
   "insert into $self->{name}_words (w_ID, word) values ($docid,
   $content)"
  
   # meta description
   "insert into $self->{name}_meta (m_ID, description) 
   values ($docid, $description)"