Ternary logic¶

As specified in the SQL standard, ternary logic, or three-valued logic (3VL), is a logic system with three truth values: TRUE, FALSE, and UNKNOWN. In Snowflake, UNKNOWN is represented by NULL. Ternary logic applies to the evaluation of Boolean expressions, as well as predicates, and affects the results of logical operations such as AND, OR, and NOT:

When used in expressions (e.g. SELECT list), UNKNOWN results are returned as NULL values.
When used as a predicate (e.g. WHERE clause), UNKNOWN results evaluate to FALSE.

Truth tables¶

This section describes the truth tables for the comparison and logical operators.

Comparison operators¶

If any operand for a comparison operator is NULL, the result is NULL. Comparison operators are:

= , != , <>

< , <= , > , >=

GREATEST , LEAST

Logical operators¶

Given a BOOLEAN column C:

If `C` is:	`C AND NULL` evaluates to:	`C OR NULL` evaluates to:	`NOT C` evaluates to:
TRUE	NULL	TRUE	FALSE
FALSE	FALSE	NULL	TRUE
NULL	NULL	NULL	NULL

In addition:

If `C` is:	`C AND (NOT C)` evaluates to:	`C OR (NOT C)` evaluates to:	`NOT (C OR NULL)` evaluates to:
TRUE	FALSE	TRUE	FALSE
FALSE	FALSE	TRUE	NULL
NULL	NULL	NULL	NULL

Usage notes for conditional expressions¶

This section describes behavior specific to the following conditional expressions.

IFF behavior¶

The IFF function returns the following results for ternary logic. Given a BOOLEAN column C:

If `C` is:	`IFF(C, e1, e2)` evaluates to:
TRUE	`e1`
FALSE	`e2`
NULL	`e2`

[ NOT ] IN behavior¶

The [ NOT ] IN functions return the following results for ternary logic. Given 3 numeric columns c1, c2, and c3:

c1 IN (c2, c3, ...) is syntactically equivalent to (c1 = c2 or c1 = c3 or ...).

As a result, when the value of c1 is NULL, the expression c1 IN (c2, c3, NULL) always evaluates to FALSE.
c1 NOT IN (c2, c3, ... ) is syntactically equivalent to (c1 <> c2 AND c1 <> c3 AND ...).

Therefore, even if c1 NOT IN (c2, c3) is TRUE, c1 NOT IN (c2, c3, NULL) evaluates to NULL.