The REINDEX command is used to delete and recreate indices from scratch. This is useful when the definition of a collation sequence has changed.
If the REINDEX keyword is not followed by a collation-sequence or database object identifier, then all indices in all attached databases are rebuilt.
If the REINDEX keyword is followed by a collation-sequence name, then all indices in all attached databases that use the named collation sequences are recreated.
Or, if the argument attached to the REINDEX identifies a specific database table, then all indices attached to the database table are rebuilt. If it identifies a specific database index, then just that index is recreated.
For a command of the form "REINDEX name", a match against collation-name takes precedence over a match against index-name or table-name. This ambiguity in the syntax may be avoided by always specifying a schema-name when reindexing a specific table or index.