Return the character count of the string `'Alexandra'` and the character count of a missing value (a SQL `NULL`) in a single row

The shape

LENGTH(NULL) returns NULL, not 0. The query produces one row with two columns: name_length is the integer 9 for the nine characters in 'Alexandra', and null_length is NULL because the input is itself NULL. The result mirrors the universal rule for scalar functions in PostgreSQL: NULL in, NULL out.

Clause by clause

• SELECT LENGTH('Alexandra') AS name_length, LENGTH(NULL) AS null_length calls LENGTH twice in the same SELECT list, once on a real string and once on NULL. The first call counts characters and returns 9. The second call has no string to measure, so it returns NULL: there is no meaningful character count for a value that is absent. The two results are returned side by side in a single row.
• There is no FROM because both inputs are written into the SELECT list directly. The function takes one input per call and returns one output per call, regardless of whether the input is a literal string or a literal NULL.

Why LENGTH(NULL) is not 0

NULL means the value is absent, not that the value is the empty string. LENGTH('') returns 0 because the empty string is a real value with zero characters; LENGTH(NULL) returns NULL because there is no string to count at all. The two are different inputs and the function correctly distinguishes them. Confusing the empty string with NULL is the root of an entire class of downstream bugs, and the difference between 0 and NULL in this column is the signal that catches it.

The trap

Every scalar string function in PostgreSQL follows this same rule. UPPER(NULL) is NULL. LOWER(NULL) is NULL. TRIM(NULL) is NULL. SUBSTRING(NULL, 1, 5) is NULL. The NULL propagates through whatever expression it lands in, including arithmetic, comparisons, and further string functions. A downstream aggregation that does not account for this will silently drop the missing rows or, worse, fold a NULL into a column where a 0 was expected and skew the report. When a column can be missing and the downstream consumer needs a definite value, the conversion has to be explicit, not assumed from the function's behavior.