keyword

The keyword template emits the entire input as a single verbatim token — it performs no splitting, normalization or stemming. It is the right choice for values that must match exactly and as a whole: tags, status codes, enum values, identifiers and other atomic strings.

Because the token is the raw input, a query matches only when it is byte-for-byte identical, including case and spacing. For an exact-match column that should still be case- or accent-insensitive, use norm, which normalizes the single token; for per-word search, use text.

Options

The keyword template takes no options.

Tokenization

The template performs no analysis at all: whatever string it receives becomes a single token, byte for byte. Spaces, punctuation and case are all preserved, so a query matches only when it is identical to the indexed value as a whole.

Input	Tokens
New York City	{"New York City"}
Hello World 42	{"Hello World 42"}
ERR_TIMEOUT	{ERR_TIMEOUT}

The token is quoted in the output whenever it contains spaces or other characters that would otherwise be ambiguous in the array literal. Preview it with ts_lexize:

Query

CREATE TEXT SEARCH DICTIONARY tok_keyword (    template = 'keyword');
SELECT ts_lexize('tok_keyword', 'New York City');

Result

 ts_lexize------------------- {"New York City"}

Examples

Create a verbatim dictionary:

Query

CREATE TEXT SEARCH DICTIONARY verbatim (    template = 'keyword');

Preview how it tokenizes — the whole string, spaces and all, becomes one token:

Query

SELECT ts_lexize('verbatim', 'Hello World 42');

Result

 ts_lexize-------------------- {"Hello World 42"}

See also

• norm — a single token, normalized for case and accents
• text — split into words
• CREATE TEXT SEARCH DICTIONARY