vt_create_text_index — Add text index to an existing table
||in table varchar ,|
|in col varchar ,|
|in id_col varchar ,|
|in is_xml integer ,|
|in defer_generation integer ,|
|in clustered_columns any ,|
|in use_hook_function integer ,|
|in language_name varchar ,|
|in encoding_name integer ,|
|in silent integer
The vt_create_text_index procedure adds a text index to an existing table. There can at most be one text index per table, including super tables and subtables.
The table argument is a string naming the table. The column is the name of the column to index. The id_col should be the name of a unique integer row identifier column. If null, the system will either add such a column or use an existing integer primary key if one is available. The is_xml argument, if non-0, specifies that the values of the indexed column should be checked for XML well formedness and that the XML structure should be taken into account in indexing the values.
Use the CREATE TEXT INDEX statement as an alternative to the vt_create_text_index function.
In order for a table to be referenced in a text index it must have a uniquely identifying integer key. If the table in question has such a key this can be used as the id column. If there is no such column this procedure makes one. Using a previously existing identifier column saves space and if that is the primary key of the table this also saves in retrieval time.
If the table being indexed has a single part integer primary key vt_create_text_index will automatically use this as the identifier. Note that the zero and negative numbers may not be used as identifier values.
Creating the index will read through the table's contents and
generate the index. When the table is changed the index can either
be updated after each change or periodically, depending on the
application needs. The rationale for background maintenance of the
text index is that it is up to several times more efficient to
maintain a text index in batches of several changed documents than
after each single document change. The default maintenance mode is
synchronous, meaning that each insert, delete or update of the
indexed column will be immediately reflected in the index. This
mode can be set using the
vt_batch_update() procedure. The mode should be
set to batch if there are any massive operations on the table.
It will create two additional tables:
and two procedures:
vt_index_<datatable_qualifier>_<data_table_owner>_<data_table_name> (in to_delete integer)
The table <datatable>_<datafield>_WORDS contains the full text index data.
The table VTLOG_<datatable>_<datafield> is an update tracking table, similar to the snapshot log table but using the key column instead of the primary key.
in the transaction semantics section that there is the sync mode for the purpose of creating a text index.
the table containing the data to index.
the column in the data table containing the data to index (a long varchar column).
an integer unique indexed column used by the free text index as a key. If Virtuoso tries to choose such a column among the existing in the table and if it doesn't find a suitable column it adds such a column with the name <datacolumn>_ID, fills that up and makes an unique index on it.
if greater than 0 installs two additional triggers before insert and before update on the data table to ensure the data being inserted into it are valid XML documents.
if nonzero then the free-text index will not be filled by actual data immediately after the creation. It will remain empty until explicit request for "incremental indexing".
a vector of names of "clustered columns" or NULL to not store such data in the index.
if nonzero, two user-defined Virtuoso/PL functions will be called when free-text data are updated, not the default system routine. These functions are recognized by their special names: <datatable>_<datafield>_INDEX_HOOK will be called to index new documents and <datatable>_<datafield>_UNINDEX_HOOK will be called to remove obsolete index information about deleted documents.
the name of the language that is used for building the index. If the parameter is omitted or is equal to '*ini*' string, indexing routines will use the language specified in server's configuration.
the name of the encoding that is used by default to index source texts. If the parameter is omitted or is equal to '*ini*' string, indexing routines will use the encoding specified by charset of the RDBMS connection that is in use when the index is created.
Flag with values 1/0. Default is 0. When set to 0, the function signals error if the text index exists. If set to 1, the function signals no error if the text index exists.
Example24.448.Creating A Text Index Using FT functions
create table xml_text (xt_id integer, xt_file varchar, xt_text long varchar, primary key (xt_id)); create index xt_file on xml_text (xt_file); vt_create_text_index ('xml_text', 'xt_text', null, 1);
-- create a text index on xt_text with XML well formedness checking on
-- populate the text index based in the data in the table.
-- refresh the index to reflect changes to the data since the last call
-- to this function or the initial call to VT_INDEX_DB_DBA_XML_TEXT.
vt_batch_update ('DB.DBA.XML_TEXT', 'on', 10);
-- sets the index maintenance to be asynchronous, refreshed at maximum intervals of 10 minutes or at every scheduler cycle if the latter is longer than 10 minutes.
vt_batch_update ('DB.DBA.XML_TEXT', 'off', null);
-- sets the index maintenance to be synchronous, within the same transaction
-- as the operation changing the indexed data.