Skip to main content
In addition to indexing columns, Postgres expressions can also be indexed.

​
Indexing Text/JSON Expressions

The following statement indexes an expression which concatenates description and category, which are both text fields: 
CREATE INDEX search_idx ON mock_items
USING bm25 (id, ((description || ' ' || category)::pdb.simple('alias=description_concat')))
WITH (key_field='id');
To index a text/JSON expression:
  1. Add the expression to the column list. In this example, the expression is description || ' ' || category.
  2. Cast it to a tokenizer, in this example pdb.simple.
  3. ParadeDB will try and infer a field name based on the field used in the expression. However, if the field name cannot be inferred (e.g. because the expression involves more than one field), you will be required to add an alias=<alias_name> to the tokenizer.
Querying against the expression is the same as querying a regular field: 
SELECT description, rating, category
FROM mock_items
WHERE (description || ' ' || category) &&& 'running shoes';
The expression on the left-hand side of the operator must exactly match the expression that was indexed.

​
Indexing Non-Text Expressions

To index a non-text expression, cast the expression to pdb.alias. For example, the following statement indexes the expression rating + 1, which returns an integer: 
CREATE INDEX search_idx ON mock_items
USING bm25 (id, description, ((rating + 1)::pdb.alias('rating')))
WITH (key_field='id');
With the expression indexed, queries containing the expression can be pushed down to the ParadeDB index: 
SELECT description, rating, category
FROM mock_items
WHERE description &&& 'running shoes'
AND rating + 1 > 3;
A current limitation is that the alias name must be the same as the field inside the expression in order for the index to be used at query time. For example, the statement above uses rating as the alias name.