pg_sql (equery v0.22.0)
View SourceSQL expression builders.
Operators, scalar functions, aggregates, CASE, type casts. These produce
qast:ast_node() values that go inside q:where/q:select/etc.
closures.
Inside DSL closures, native Erlang operators (=:=, >, andalso,
+, …) are rewritten by equery_pt to corresponding pg_sql:
functions, so you rarely call these directly for primitive operations.
Use the explicit pg_sql: form for:
- Functions (
pg_sql:count/1,pg_sql:lower/1,pg_sql:date_trunc/2, …); - Operators that aren't Erlang operators (
pg_sql:like/2,pg_sql:'~'/2); - Aggregates and
pg_sql:filter/2; pg_sql:case_when/1,2;- Type casts (
pg_sql:as/2).
JSON/JSONB live in qjson.
Summary
Aggregates
array_agg(A) — collect values into an array. NULLs are included.
array_agg(A ORDER BY ...) — collect into array with explicit ordering.
avg(A) — arithmetic mean. Returns numeric or double precision.
bool_and(A) — true iff every non-NULL input is true.
bool_or(A) — true iff any non-NULL input is true.
count(A) — non-NULL row count. Returns bigint.
DISTINCT A qualifier — typically used inside another aggregate, e.g. count(distinct(x)).
every(A) — SQL standard alias for bool_and/1.
Wrap any aggregate with FILTER (WHERE Cond) — limits which rows the
aggregate sees, without affecting the rest of the query. Essential for
dashboard-style queries with multiple conditional metrics.
json_agg(A) — aggregate into a JSON array. See JSON Functions.
json_agg(A ORDER BY ...).
json_object_agg(K, V) — aggregate into a JSON object.
jsonb_agg(A) — aggregate into a JSONB array.
jsonb_agg(A ORDER BY ...).
jsonb_object_agg(K, V) — aggregate into a JSONB object.
max(A) — aggregate maximum.
min(A) — aggregate minimum.
mode() WITHIN GROUP (ORDER BY x) — most frequent value (statistical mode).
percentile_cont(Frac) WITHIN GROUP (ORDER BY x) — interpolated continuous
percentile. Frac ∈ [0, 1].
percentile_disc(Frac) WITHIN GROUP (ORDER BY x) — discrete percentile (picks one of the input values).
string_agg(expr, sep) — concatenate string values with a separator. NULL inputs are ignored.
string_agg(expr, sep ORDER BY ...) — concatenate with explicit ordering.
sum(A). See Aggregate Functions.
Arithmetic
A * B.
A + B. See Mathematical Operators.
A - B.
A / B — division. For integers, truncates toward zero.
abs(A) — absolute value. See Math Functions.
div(A, B) — integer quotient, truncated toward zero.
Alias for mod/2. Erlang's rem operator inside DSL closures rewrites to this.
Arrays
A && B — arrays overlap (have any common element).
A <@ B — is contained by.
A @> B — contains. See Array Functions and Operators.
ARRAY[v1, v2, ...] constructor. Element type is inferred from the
first element's opts.
array_append(arr, elem) — append element.
array_cat(a, b) — concatenate two arrays. Equivalent to a || b.
array_length(arr, 1) — length of the first dimension.
array_length(arr, dim) — length of dimension dim.
array_position(arr, elem) — 1-based index of elem in arr, or NULL.
array_prepend(elem, arr) — prepend element.
array_remove(arr, elem) — remove all occurrences of elem.
array_replace(arr, from, to) — replace all from elements with to.
unnest(arr) — set-returning function that produces one row per array
element.
Comparison
A < B.
A <> B. See Comparison Operators.
A = B. See Comparison Operators.
A <= B.
A > B. See Comparison Operators.
A >= B.
A IS B. Low-level building block for IS NULL, IS TRUE, etc. See Comparison Predicates.
A IS DISTINCT FROM B — NULL-safe inequality (treats NULL = NULL as
equal). See Comparison Predicates.
A IS NOT DISTINCT FROM B — NULL-safe equality.
A IS NOT NULL.
A IS NULL. See Comparison Predicates.
Conditional
Searched CASE WHEN cond THEN val ... END without ELSE — NULL if no branch matches.
CASE WHEN cond THEN val ... ELSE default END.
COALESCE(a, b, c, ...) — returns the first non-NULL argument.
GREATEST(A, B, C, ...) — N-ary form. Returns max of all non-NULL inputs.
LEAST(A, B, C, ...) — N-ary form. Returns min of all non-NULL inputs.
GREATEST(A, B) — maximum of two values, ignoring NULLs.
LEAST(A, B) — minimum of two values, ignoring NULLs. See Conditional Expressions.
Date/time
age(t) — interval since current_date to t (or vice-versa).
age(t1, t2) — symbolic interval t1 - t2.
current_date — current date (date only).
current_time — current time of day with timezone.
current_timestamp — start-of-transaction timestamp with timezone. Standard SQL.
date_part('field', source) — same as extract, returns double precision.
date_trunc('field', source) — truncate a timestamp to the specified
precision. Field is a closed enum (validated at build time).
EXTRACT(field FROM source) — get a sub-field as numeric.
now() — current transaction timestamp (with timezone). PostgreSQL extension; equivalent to current_timestamp/0.
to_char(val, fmt) — format a date/timestamp/number as text. See Data Type Formatting Functions.
to_date(text, fmt) — parse a string into date.
to_timestamp(epoch) — Unix-epoch seconds to timestamptz.
to_timestamp(text, fmt) — parse a formatted string into timestamptz.
Logical
Logical AND. Short-circuits on boolean literals. See Logical Operators.
Logical NOT. See Logical Operators.
Logical OR. Short-circuits on boolean literals. See Logical Operators.
Misc
Build a function call AST: FunName(Args...).
A IN (...) — membership test.
ROW(...) constructor with anonymous record type. See Row Constructors.
ROW(...) constructor tagged with a model module (for type inference in row-typed projections).
Numeric
ceil(A) — ceiling (smallest integer not less than A).
exp(A) — exponential (e^A).
floor(A) — floor (largest integer not greater than A).
ln(A) — natural logarithm.
log(A) — base-10 logarithm.
log(B, A) — logarithm of A to base B.
mod(A, B) — remainder of A / B. See Math Functions.
power(A, B) — A raised to the power of B.
random() — pseudo-random double precision in [0.0, 1.0).
round(A) — round to nearest integer (banker's rounding for numeric).
round(A, N) — round to N decimal places.
sign(A) — -1, 0, or 1 according to sign.
sqrt(A) — square root.
trunc(V, N) — truncate to N decimal places.
Pattern matching
A ILIKE B — case-insensitive LIKE (PostgreSQL extension).
A LIKE B. See LIKE.
POSIX regex match (case-insensitive) A ~* B.
POSIX regex match A ~ B. See POSIX Regular Expressions.
Strings
char_length(s) — SQL-standard alias for length.
concat(a, b, c, ...) — concatenate. NULLs are skipped (unlike ||). See String Functions.
concat(a, b) — 2-arg form for convenience.
length(s) — string length in characters.
lower(s) — lowercase.
ltrim(s) — strip whitespace from the left.
ltrim(s, chars) — strip listed characters from the left.
regexp_match(s, pat) — returns a text[] of capture groups, or NULL if no match.
regexp_match(s, pat, flags) — with regex flags.
regexp_replace(s, pat, repl) — replace first match. See POSIX Regex Match.
regexp_replace(s, pat, repl, flags) — 'g' for global replace, 'i' case-insensitive, etc.
replace(s, from, to) — replace all occurrences of from with to.
rtrim(s) — strip whitespace from the right.
rtrim(s, chars) — strip listed characters from the right.
split_part(s, sep, n) — n-th field after splitting s by sep (1-indexed).
starts_with(s, prefix) — boolean prefix test. Faster than LIKE 'prefix%' for indexed lookup.
strpos(haystack, needle) — 1-based position of needle in haystack, or 0.
substring(s, from) — from the given 1-based start position to end.
substring(s, from, len) — len characters starting at from.
trim(s) — strip whitespace from both ends.
trim(s, chars) — strip any character listed in chars from both ends.
upper(s) — uppercase.
A || B — concatenation operator.
Type casts
(Ast)::Type — SQL type cast.
Annotate an AST node with a type, without emitting an SQL cast.
Types
-type agg_order_specs() :: [agg_order_spec(), ...].
-type datetime_field() ::
century | day | decade | dow | doy | epoch | hour | isodow | isoyear | julian | microseconds |
millennium | milliseconds | minute | month | quarter | second | timezone | timezone_hour |
timezone_minute | week | year.
-type expr() :: qast:ast_node() | term().
Aggregates
-spec array_agg(expr()) -> qast:ast_node().
array_agg(A) — collect values into an array. NULLs are included.
-spec array_agg(expr(), agg_order_specs()) -> qast:ast_node().
array_agg(A ORDER BY ...) — collect into array with explicit ordering.
-spec avg(expr()) -> qast:ast_node().
avg(A) — arithmetic mean. Returns numeric or double precision.
-spec bool_and(expr()) -> qast:ast_node().
bool_and(A) — true iff every non-NULL input is true.
-spec bool_or(expr()) -> qast:ast_node().
bool_or(A) — true iff any non-NULL input is true.
-spec count(qast:ast_node()) -> qast:ast_node().
count(A) — non-NULL row count. Returns bigint.
-spec distinct(expr()) -> qast:ast_node().
DISTINCT A qualifier — typically used inside another aggregate, e.g. count(distinct(x)).
-spec every(expr()) -> qast:ast_node().
every(A) — SQL standard alias for bool_and/1.
-spec filter(qast:ast_node(), expr()) -> qast:ast_node().
Wrap any aggregate with FILTER (WHERE Cond) — limits which rows the
aggregate sees, without affecting the rest of the query. Essential for
dashboard-style queries with multiple conditional metrics.
#{
total => pg_sql:count(Id),
paid => pg_sql:filter(pg_sql:count(Id), Status =:= <<"paid">>)
}.-spec json_agg(expr()) -> qast:ast_node().
json_agg(A) — aggregate into a JSON array. See JSON Functions.
-spec json_agg(expr(), agg_order_specs()) -> qast:ast_node().
json_agg(A ORDER BY ...).
-spec json_object_agg(expr(), expr()) -> qast:ast_node().
json_object_agg(K, V) — aggregate into a JSON object.
-spec jsonb_agg(expr()) -> qast:ast_node().
jsonb_agg(A) — aggregate into a JSONB array.
-spec jsonb_agg(expr(), agg_order_specs()) -> qast:ast_node().
jsonb_agg(A ORDER BY ...).
-spec jsonb_object_agg(expr(), expr()) -> qast:ast_node().
jsonb_object_agg(K, V) — aggregate into a JSONB object.
-spec max(expr()) -> qast:ast_node().
max(A) — aggregate maximum.
-spec min(expr()) -> qast:ast_node().
min(A) — aggregate minimum.
-spec mode(expr()) -> qast:ast_node().
mode() WITHIN GROUP (ORDER BY x) — most frequent value (statistical mode).
-spec percentile_cont(expr(), expr()) -> qast:ast_node().
percentile_cont(Frac) WITHIN GROUP (ORDER BY x) — interpolated continuous
percentile. Frac ∈ [0, 1].
-spec percentile_disc(expr(), expr()) -> qast:ast_node().
percentile_disc(Frac) WITHIN GROUP (ORDER BY x) — discrete percentile (picks one of the input values).
-spec string_agg(expr(), expr()) -> qast:ast_node().
string_agg(expr, sep) — concatenate string values with a separator. NULL inputs are ignored.
-spec string_agg(expr(), expr(), agg_order_specs()) -> qast:ast_node().
string_agg(expr, sep ORDER BY ...) — concatenate with explicit ordering.
Without ORDER BY, string_agg is non-deterministic. Use the 3-arg form
when result order matters.
-spec sum(qast:ast_node()) -> qast:ast_node().
sum(A). See Aggregate Functions.
Arithmetic
-spec '*'(expr(), expr()) -> qast:ast_node().
A * B.
-spec '+'(expr(), expr()) -> qast:ast_node().
A + B. See Mathematical Operators.
-spec '-'(expr(), expr()) -> qast:ast_node().
A - B.
-spec '/'(expr(), expr()) -> qast:ast_node().
A / B — division. For integers, truncates toward zero.
-spec abs(expr()) -> qast:ast_node().
abs(A) — absolute value. See Math Functions.
-spec 'div'(expr(), expr()) -> qast:ast_node().
div(A, B) — integer quotient, truncated toward zero.
-spec 'rem'(expr(), expr()) -> qast:ast_node().
Alias for mod/2. Erlang's rem operator inside DSL closures rewrites to this.
Arrays
-spec '&&'(expr(), expr()) -> qast:ast_node().
A && B — arrays overlap (have any common element).
-spec '<@'(expr(), expr()) -> qast:ast_node().
A <@ B — is contained by.
A @> B — contains. See Array Functions and Operators.
-spec array([expr()]) -> qast:ast_node().
ARRAY[v1, v2, ...] constructor. Element type is inferred from the
first element's opts.
Empty array (ARRAY[]) needs an explicit cast in PG; use
as/2: pg_sql:as(pg_sql:array([]), {array, int}).
-spec array_append(expr(), expr()) -> qast:ast_node().
array_append(arr, elem) — append element.
-spec array_cat(expr(), expr()) -> qast:ast_node().
array_cat(a, b) — concatenate two arrays. Equivalent to a || b.
-spec array_length(expr()) -> qast:ast_node().
array_length(arr, 1) — length of the first dimension.
-spec array_length(expr(), expr()) -> qast:ast_node().
array_length(arr, dim) — length of dimension dim.
-spec array_position(expr(), expr()) -> qast:ast_node().
array_position(arr, elem) — 1-based index of elem in arr, or NULL.
-spec array_prepend(expr(), expr()) -> qast:ast_node().
array_prepend(elem, arr) — prepend element.
-spec array_remove(expr(), expr()) -> qast:ast_node().
array_remove(arr, elem) — remove all occurrences of elem.
-spec array_replace(expr(), expr(), expr()) -> qast:ast_node().
array_replace(arr, from, to) — replace all from elements with to.
-spec unnest(expr()) -> qast:ast_node().
unnest(arr) — set-returning function that produces one row per array
element.
Integrates with q:from/1 and q:lateral_join/2,3,4: the resulting
table has a single column named unnest (PostgreSQL default).
q:pipe(q:from(?USER), [
q:lateral_join(inner, fun([#{tags := T}]) ->
q:from(pg_sql:unnest(T))
end),
q:select(fun([#{name := N}, #{unnest := Tag}]) ->
#{name => N, tag => Tag}
end)
]).Comparison
-spec '<'(expr(), expr()) -> qast:ast_node().
A < B.
-spec '=/='(expr(), expr()) -> qast:ast_node().
A <> B. See Comparison Operators.
-spec '=:='(expr(), expr()) -> qast:ast_node().
A = B. See Comparison Operators.
-spec '=<'(expr(), expr()) -> qast:ast_node().
A <= B.
-spec '>'(expr(), expr()) -> qast:ast_node().
A > B. See Comparison Operators.
-spec '>='(expr(), expr()) -> qast:ast_node().
A >= B.
-spec is(expr(), expr()) -> qast:ast_node().
A IS B. Low-level building block for IS NULL, IS TRUE, etc. See Comparison Predicates.
-spec is_distinct_from(expr(), expr()) -> qast:ast_node().
A IS DISTINCT FROM B — NULL-safe inequality (treats NULL = NULL as
equal). See Comparison Predicates.
-spec is_not_distinct_from(expr(), expr()) -> qast:ast_node().
A IS NOT DISTINCT FROM B — NULL-safe equality.
-spec is_not_null(expr()) -> qast:ast_node().
A IS NOT NULL.
-spec is_null(expr()) -> qast:ast_node().
A IS NULL. See Comparison Predicates.
Conditional
-spec case_when([{expr(), expr()}, ...]) -> qast:ast_node().
Searched CASE WHEN cond THEN val ... END without ELSE — NULL if no branch matches.
-spec case_when([{expr(), expr()}, ...], expr() | undefined) -> qast:ast_node().
CASE WHEN cond THEN val ... ELSE default END.
pg_sql:case_when([
{Id > 100, <<"big">>},
{Id > 10, <<"medium">>}
], <<"small">>).COALESCE(a, b, c, ...) — returns the first non-NULL argument.
-spec greatest([expr(), ...]) -> qast:ast_node().
GREATEST(A, B, C, ...) — N-ary form. Returns max of all non-NULL inputs.
-spec least([expr(), ...]) -> qast:ast_node().
LEAST(A, B, C, ...) — N-ary form. Returns min of all non-NULL inputs.
-spec max(expr(), expr()) -> qast:ast_node().
GREATEST(A, B) — maximum of two values, ignoring NULLs.
-spec min(expr(), expr()) -> qast:ast_node().
LEAST(A, B) — minimum of two values, ignoring NULLs. See Conditional Expressions.
Date/time
-spec age(expr()) -> qast:ast_node().
age(t) — interval since current_date to t (or vice-versa).
-spec age(expr(), expr()) -> qast:ast_node().
age(t1, t2) — symbolic interval t1 - t2.
-spec current_date() -> qast:ast_node().
current_date — current date (date only).
-spec current_time() -> qast:ast_node().
current_time — current time of day with timezone.
-spec current_timestamp() -> qast:ast_node().
current_timestamp — start-of-transaction timestamp with timezone. Standard SQL.
-spec date_part(datetime_field(), expr()) -> qast:ast_node().
date_part('field', source) — same as extract, returns double precision.
-spec date_trunc(datetime_field(), expr()) -> qast:ast_node().
date_trunc('field', source) — truncate a timestamp to the specified
precision. Field is a closed enum (validated at build time).
See Date/Time Functions.
-spec extract(datetime_field(), expr()) -> qast:ast_node().
EXTRACT(field FROM source) — get a sub-field as numeric.
See EXTRACT.
-spec now() -> qast:ast_node().
now() — current transaction timestamp (with timezone). PostgreSQL extension; equivalent to current_timestamp/0.
-spec to_char(expr(), expr()) -> qast:ast_node().
to_char(val, fmt) — format a date/timestamp/number as text. See Data Type Formatting Functions.
-spec to_date(expr(), expr()) -> qast:ast_node().
to_date(text, fmt) — parse a string into date.
-spec to_timestamp(expr()) -> qast:ast_node().
to_timestamp(epoch) — Unix-epoch seconds to timestamptz.
-spec to_timestamp(expr(), expr()) -> qast:ast_node().
to_timestamp(text, fmt) — parse a formatted string into timestamptz.
Logical
-spec 'andalso'(V, V) -> V when V :: boolean() | qast:ast_node().
Logical AND. Short-circuits on boolean literals. See Logical Operators.
-spec 'not'(V) -> V when V :: boolean() | qast:ast_node().
Logical NOT. See Logical Operators.
-spec 'orelse'(V, V) -> V when V :: boolean() | qast:ast_node().
Logical OR. Short-circuits on boolean literals. See Logical Operators.
Misc
-spec call(iodata(), [expr()], qast:opts()) -> qast:ast_node().
Build a function call AST: FunName(Args...).
Used internally by all pg_sql: function builders. Public for custom
PG functions not yet wrapped by the library:
pg_sql:call("my_extension_fn", [Arg1, Arg2], #{type => text}).Warning
FunName is inlined raw. Never pass user-controlled input here;
use a literal string.
EXISTS (subquery). See EXISTS.
A IN (...) — membership test.
- With a
query()argument, emitsA IN (subquery). See Subquery Expressions. - With a list, emits
A = ANY($1)(parameterized). - With a single-element list, optimized to
A = item.
-spec row(#{atom() => qast:ast_node()}) -> qast:ast_node().
ROW(...) constructor with anonymous record type. See Row Constructors.
-spec row(Model :: module(), #{atom() => qast:ast_node()}) -> qast:ast_node().
ROW(...) constructor tagged with a model module (for type inference in row-typed projections).
Numeric
-spec ceil(expr()) -> qast:ast_node().
ceil(A) — ceiling (smallest integer not less than A).
-spec exp(expr()) -> qast:ast_node().
exp(A) — exponential (e^A).
-spec floor(expr()) -> qast:ast_node().
floor(A) — floor (largest integer not greater than A).
-spec ln(expr()) -> qast:ast_node().
ln(A) — natural logarithm.
-spec log(expr()) -> qast:ast_node().
log(A) — base-10 logarithm.
-spec log(expr(), expr()) -> qast:ast_node().
log(B, A) — logarithm of A to base B.
-spec mod(expr(), expr()) -> qast:ast_node().
mod(A, B) — remainder of A / B. See Math Functions.
-spec power(expr(), expr()) -> qast:ast_node().
power(A, B) — A raised to the power of B.
-spec random() -> qast:ast_node().
random() — pseudo-random double precision in [0.0, 1.0).
-spec round(expr()) -> qast:ast_node().
round(A) — round to nearest integer (banker's rounding for numeric).
-spec round(expr(), expr()) -> qast:ast_node().
round(A, N) — round to N decimal places.
-spec sign(expr()) -> qast:ast_node().
sign(A) — -1, 0, or 1 according to sign.
-spec sqrt(expr()) -> qast:ast_node().
sqrt(A) — square root.
-spec trunc(expr(), qast:ast_node() | non_neg_integer()) -> qast:ast_node().
trunc(V, N) — truncate to N decimal places.
Pattern matching
A ILIKE B — case-insensitive LIKE (PostgreSQL extension).
A LIKE B. See LIKE.
-spec '~*'(expr(), expr()) -> qast:ast_node().
POSIX regex match (case-insensitive) A ~* B.
-spec '~'(expr(), expr()) -> qast:ast_node().
POSIX regex match A ~ B. See POSIX Regular Expressions.
Strings
-spec char_length(expr()) -> qast:ast_node().
char_length(s) — SQL-standard alias for length.
-spec concat([expr(), ...]) -> qast:ast_node().
concat(a, b, c, ...) — concatenate. NULLs are skipped (unlike ||). See String Functions.
-spec concat(expr(), expr()) -> qast:ast_node().
concat(a, b) — 2-arg form for convenience.
-spec length(expr()) -> qast:ast_node().
length(s) — string length in characters.
-spec lower(expr()) -> qast:ast_node().
lower(s) — lowercase.
-spec ltrim(expr()) -> qast:ast_node().
ltrim(s) — strip whitespace from the left.
-spec ltrim(expr(), expr()) -> qast:ast_node().
ltrim(s, chars) — strip listed characters from the left.
-spec regexp_match(expr(), expr()) -> qast:ast_node().
regexp_match(s, pat) — returns a text[] of capture groups, or NULL if no match.
-spec regexp_match(expr(), expr(), expr()) -> qast:ast_node().
regexp_match(s, pat, flags) — with regex flags.
-spec regexp_replace(expr(), expr(), expr()) -> qast:ast_node().
regexp_replace(s, pat, repl) — replace first match. See POSIX Regex Match.
-spec regexp_replace(expr(), expr(), expr(), expr()) -> qast:ast_node().
regexp_replace(s, pat, repl, flags) — 'g' for global replace, 'i' case-insensitive, etc.
-spec replace(expr(), expr(), expr()) -> qast:ast_node().
replace(s, from, to) — replace all occurrences of from with to.
-spec rtrim(expr()) -> qast:ast_node().
rtrim(s) — strip whitespace from the right.
-spec rtrim(expr(), expr()) -> qast:ast_node().
rtrim(s, chars) — strip listed characters from the right.
-spec split_part(expr(), expr(), expr()) -> qast:ast_node().
split_part(s, sep, n) — n-th field after splitting s by sep (1-indexed).
-spec starts_with(expr(), expr()) -> qast:ast_node().
starts_with(s, prefix) — boolean prefix test. Faster than LIKE 'prefix%' for indexed lookup.
-spec strpos(expr(), expr()) -> qast:ast_node().
strpos(haystack, needle) — 1-based position of needle in haystack, or 0.
-spec substring(expr(), expr()) -> qast:ast_node().
substring(s, from) — from the given 1-based start position to end.
-spec substring(expr(), expr(), expr()) -> qast:ast_node().
substring(s, from, len) — len characters starting at from.
-spec trim(expr()) -> qast:ast_node().
trim(s) — strip whitespace from both ends.
-spec trim(expr(), expr()) -> qast:ast_node().
trim(s, chars) — strip any character listed in chars from both ends.
-spec upper(expr()) -> qast:ast_node().
upper(s) — uppercase.
-spec '||'(expr(), expr()) -> qast:ast_node().
A || B — concatenation operator.
NULL-propagating: 'foo' || NULL → NULL. For NULL-skipping behavior
use concat/2. Same operator works for arrays.
See String Functions.
Type casts
(Ast)::Type — SQL type cast.
pg_sql:as(Id, text) %% (id)::text
pg_sql:as(V, {varchar, 60}) %% (v)::varchar(60)
pg_sql:as(V, {array, integer}) %% (v)::integer[]See Type Casts.
Warning
Type is rendered raw. Use only with literal atoms / tuples; never
with user-controlled values.
Annotate an AST node with a type, without emitting an SQL cast.
Useful when you know the inferred type is wrong (e.g. when wrapping a
call/3 to a function whose return type isn't auto-inferred)
and want correct downstream type propagation without paying for a
runtime cast.