Functions

A scalar function works at any position that accepts a value. That includes a projection, a filter, a key of a GROUP BY, a default, and a check constraint. A function also nests inside another function:

SELECT upper(trim(name)) AS display_name
FROM robots
WHERE abs(year - 2000) <= 5;

The name of a function is not case-sensitive.

Categories

• Session Functions
• String Functions
• Math Functions
• Hash Functions
• Date/Time Functions
• JSON Functions
• Regex Functions
• Conversion Functions
• Null Functions
• UUID Functions
• Object Id Functions
• Array Functions, which are cardinality, array_length, and array_contains
• Sequence functions, which are nextval, currval, lastval, and setval
• Vector functions, which are octet_length, vector_dims, l2_distance, cosine_distance, and inner_product

Two rules worth your attention

A null propagates. Most functions return NULL when one argument is NULL. Null Functions covers the exceptions, such as coalesce and nullif. Each page of a category marks any other exception.

Some functions are volatile. Eight of them can return a different value at every evaluation: gen_id(), gen_uuid_v4(), gen_uuid_v7(), current_timestamp(), now(), current_date(), unix_timestamp(), and random(). That property makes them useful as the default of a column. It also makes them unsafe in a predicate that must stay stable across a scan.

The session functions are volatile for another reason. They vary by caller, not by row. They therefore bypass the result cache. You cannot use one as a default.

CamusDB evaluates the arguments before the call. It validates the number of the arguments and their types at the execution. It does not validate them at the parse.