This is Thunderstone's name for an inverted (search-engine) index on a column of unstructured text, as distinguished from a b-tree (sorted-order) index normally used on numeric or string database fields. This following is somewhat technical and is intended for experienced database programmers.
A Metamorph index is set up and used in a manner analogous to any other database index:
create metamorph index descriptionindex on products(description);
As an illustration of the power this provides, consider a typical Texis query of this model:
SELECT id, name FROM products WHERE description LIKE 'big fancy gizmo'
ORDER BY price;
The Texis database optimizer uses the metamorph index on the description field, along with the B-tree index on the price field, to quickly and efficiently resolve this query. (From a more technical point of view, you'd probably create a single compound index combining the characteristics of those two indexes, for even better performance.)
Any other database, to accomplish something similar, would receive no help from the database optimizer. The text index is a black box to it. It can only hand off the gizmo query to a separate text index; then create a temporary table containing the text-search results; then do a join between that and the table containing the price information.
Texis is the only relational database that can resolve a query of this type without a join. This makes Texis many times more efficient.