Full text search in BLOB columns NOT Working

Question

Hi, I am using Oracle 9i.
Presently I use following code to create index.
begin
Ctx_Ddl.Drop_Preference ('text_only');
end;
begin
Ctx_Ddl.Create_Preference ( 'text_only', 'basic_lexer');
Ctx_Ddl.Set_Attribute ( 'text_only', 'index_text', 'true' );
Ctx_Ddl.Set_Attribute ( 'text_only', 'index_themes', 'false' );
end;
CREATE TABLE MY_XML_DOCS(
XML_RES_ID NUMBER (16) NOT NULL,
XML_RES BLOB NOT NULL,
XML_RES_PATH VARCHAR2 (128),
XML_DOCS_CREATED_BY VARCHAR2 (16) NOT NULL,
XML_DOCS_CREATED_DT DATE NOT NULL,
XML_DOCS_MODIFIED_BY VARCHAR2 (16),
XML_DOCS_MODIFIED_DT DATE ) ;
drop index FULL_TEXT_INDEX ;
create INDEX FULL_TEXT_INDEX ON MY_XML_DOCS(XML_RES)
INDEXTYPE IS ctxsys.CONTEXT
parameters ('lexer text_only stoplist ctxsys.empty_stoplist');
BEGIN
CTX_DDL.sync_index('FULL_TEXT_INDEX ','10M');
END;
And I've some XML documents stored in blob columns in the above table. But my query with 'contains' param does not return expected results.(it presently returns no results). Do I miss anything? Do I need to specify any thing else while I create the Index? Kindly advice.

Answer 1

I would agree the CLOB storage would be preferable for XML. If you are using BLOB storage, by default the documents are going to be FILTERed by the INSO_FILTERS. A BLOB column will give you reduced storage in a unicode environment though.
If you wish to store your data in a BLOB though, you should specify the NULL_FILTER in your create index statement to avoid having the INSO_FILTERS invoked.
Further, you may also require a CHARSET column in your table specifying the character set of your data for each row if it can vary. The Column should then also be specified in the 'PARAMETERS' clause of the create index statement, this way Oracle knows the correct character set to use when converting the binary data to character data.
My guess that you are not getting any query results because none of the documents are actually being indexed. Be sure to query the ctx_user_index_errors view to determine if you are getting errors at indexing time.