Supported PostgreSQL functions

This page defines the functions supported for PostgreSQL-dialect databases in Spanner.

Mathematical functions

Unless otherwise specified, functions return the same data type as provided in the argument.

Function	Example/Notes	Description
`abs(float4 \| float8 \| int8 \| numeric)`	`abs(-17) → 17`	Absolute value.
`acos(float8)`	`acos(1) → 0`	Inverse cosine, result in radians.
`asin(float8)`	`asin(1) → 1.5707963267948966`	Inverse sine, result in radians.
`atan(float8)`	`atan(1) → 0.7853981633974483`	Inverse tangent, result in radians.
`atan2(x float8, y float8)`	`atan2(1,0) → 1.5707963267948966`	Inverse tangent of `x/y` , result in radians.
`ceil(float8 \| numeric)`	`ceil(42.2::FLOAT8) → 43 ceil(-42.8::FLOAT8) → -42`	Nearest integer greater than or equal to argument.
`cos(float8)`	`cos(0) → 1`	Cosine, argument in radians.
`dexp(float8)`	`dexp(3) → 20.085536923187668`	Raise e to the specified exponent (e^x).
`div(x numeric, y numeric)`	`div(9, 4) → 2`	Integer quotient of x/y (truncates towards zero).
`dlog10(float8)`		Returns the base 10 logarithm of the provided value.
`dlog1(float8)`		Returns the value's natural logarithm.
`dpow(float8, float8)`		Returns the value of the first number raised to the power of the second number.
`dsqrt(float8)`		Returns the argument's square root.
`exp(float8)`	`exp(1.0::FLOAT8) → 2.7182818284590452`	Exponential (e raised to the given power).
`floor(float8 \| numeric)`	`floor(42.8::FLOAT8) → 42 floor(-42.8) → -43`	Nearest integer less than or equal to argument.
`ln(float8)`	`ln(2.0::FLOAT8) → 0.6931471805599453`	Natural logarithm.
`log(float8)`	`log(100.0::FLOAT8) → 2`	Base 10 logarithm.
`mod(x int8, y int8) mod(x numeric, y numeric)`	`mod(9,4) → 1`	Remainder of `x/y` .
`power(float8, float8)`	`power(9.0::FLOAT8, 3.0::FLOAT8) → 729`	`a` raised to the power of `b` . `pow` is an alias of `power` .
`round(float8)`	`round(42.4::FLOAT8) → 42`	Rounds to nearest integer.
`sign(float8)`	`sign(-8.4::FLOAT8) → -1`	Sign of the argument (-1, 0, or +1).
`sin(float8)`	`sin(1) → 0.8414709848078965`	Sine, argument in radians.
`spanner.bit_reverse(bigint_value, /preserve_sign=/bool)`	`spanner.bit_reverse(1, true); --> returns 4611686018427387904 spanner.bit_reverse(10, false); --> returns 5764607523034234880`	Returns a bit-reversed value for a `bigint` value. When `preserve_sign` is true, this function provides the same bit-reversal algorithm used in bit-reversed sequence. See Bit-reversed sequence .
`spanner.cosine_distance(float4[], float4[]) spanner.cosine_distance(float8[], float8[])`	Returns float8. `spanner.cosine_distance('{1.0, 2.0}'::float4[], '{3.0, 4.0}'::float4[]) → 0.016130` `spanner.cosine_distance('{2.0, 1.0}'::float8[], '{4.0, 3.0}'::float8[]) → 0.016130`	Computes the cosine distance between two vectors. Each vector represents a quantity that includes magnitude and direction. Vectors are represented as `float4[]` or `float8[]` . A vector can have one or more dimensions. Both vectors in this function must share these same dimensions, and if they don't, an error is produced. The ordering of numeric values in a vector doesn't impact the results produced by this function. An error is produced if an element or field in a vector is `null` . A vector can't be a zero vector. A vector is a zero vector if all elements in the vector are 0. For example, `'{0.0, 0.0}'::float4` . If a zero vector is encountered, an error is produced. If either of the arguments is `null` , `null` is returned.
`spanner.dot_product(int8[], int8[]) spanner.dot_product(float4[], float4[]) spanner.dot_product(float8[], float8[])`	Returns float8. `spanner.dot_product('{100}'::int8[], '{200}'::int8[]) → 20000` `spanner.dot_product('{100, 10}'::float4[], '{200, 6}'::float4[]) → 20060`	Computes the dot product of two vectors. The dot product is computed by summing the product of corresponding vector elements. Each vector represents a quantity that includes magnitude and direction. Vectors are represented as `int8[]` , `float4[]` , or `float8[]` . A vector can have one or more dimensions. Both vectors in this function must share these same dimensions, and if they don't, an error is produced. The ordering of numeric values in a vector doesn't impact the results produced by this function. An error is produced if an element or field in a vector is `null` . A vector can be a zero vector. A vector is a zero vector if it has no dimensions or if all elements in the vector are 0. For example, `'{0.0, 0.0}'::float4` . If a zero vector is encountered, an error is produced. If either of the arguments is `null` , `null` is returned.
`spanner.euclidean_distance(float4[], float4[]) spanner.euclidean_distance(float8[], float8[])`	Returns float8. `spanner.euclidean_distance('{1.0, 2.0}'::float4[], '{3.0, 4.0}'::float4[]) → 2.828` `spanner.euclidean_distance('{2.0, 1.0}'::float8[], '{4.0, 3.0}'::float8[]) → 2.828`	Computes the Euclidean distance between two vectors. Each vector represents a quantity that includes magnitude and direction. Vectors are represented as `float4[]` or `float8[]` . A vector can have one or more dimensions. Both vectors in this function must share these same dimensions, and if they don't, an error is produced. The ordering of numeric values in a vector doesn't impact the results produced by this function. An error is produced if an element or field in a vector is `null` . A vector can be a zero vector. A vector is a zero vector if all elements in the vector are 0. For example, `'{0.0, 0.0}'::float4` . If either of the arguments is `null` , `null` is returned.
`sqrt(float8)`	`sqrt(2::FLOAT8) → 1.4142135623730951`	Square root.
`tan(float8)`	`tan(1) → 1.5574077246549023`	Tangent, argument in radians.
`trunc(float8)`	`trunc(42.8::FLOAT8) → 42 trunc(-42.8::FLOAT8) → -42`	Truncates to integer (towards zero).
`trunc(x numeric, y integer)`	`trunc(42.4382, 2) → 42.43`	Truncates x to y decimal places.

Machine learning functions

Function	Example / Notes	Description
`spanner.ML_PREDICT_ROW(model_endpoint text\|jsonb , args jsonb )`	Returns JSONB.	`spanner.ML_PREDICT_ROW` is a scalar function that allows predictions on a per row basis and can appear anywhere a scalar expression is allowed in SQL statements. You can get online predictions in your SQL code by calling this function. For more information about this function, see Use ML Predict for ML serving .

Array functions, operators, and comparisons

Array operators

Operator	Example/Notes	Description
`@>`	`array[1, 2, 3] @> array[1, 2, 1] → true`	Array contains operator. Returns `true` if the first array contains the second, that is, if every element in the second array equals some element in the first array.
`<@`	`array[1, 1, 3] <@ array[1, 2, 3, 4] → true`	Array contained operator. Returns `true` if the second array contains the first array. That is, if every element in the first array equals some element in the second array.
`&&`	`array[1, 2, 3] && array[1, 5] → true`	Array overlap operator. Returns `true` if the elements in the arrays overlap, that is, if they have any element in common.
`\|\|`	`array[1, 2] \|\| array[3, 4] → {1, 2, 3, 4}`	Concatenation operator. Concatenates two arrays.

Array functions

Function	Example / Notes	Description
`array_cat(anyarray, anyarray)`	`array_cat(ARRAY['cat', 'dog'], ARRAY['bird', 'turtle']) → {"cat", "dog", "bird", "turtle"}`	Concatenates two arrays.
`array_to_string ( array text_array, delimiter text [, null_string text ] )`	`array_to_string(ARRAY['a', NULL, 'c', 'd', 'e'], ',', '') → a,,c,d,e array_to_string(ARRAY['a', NULL, 'c', 'd', 'e'], ',') → a,c,d,e array_to_string(ARRAY['a', NULL, 'c', 'd', 'e'], ',', NULL) → NULL array_to_string(ARRAY['a', NULL, 'c', 'd', 'e'], NULL, '*') → NULL`	Converts the values of the elements in a text array to their string representations. The first argument is the array which must be a text array. The second argument is a user-specified delimiter . The third (optional) argument is a user-specified null_string that the function substitutes for NULL values. If you don't pass a null_string , and the function encounters a NULL value, the NULL value is not included in the results of the function. If you pass NULL for either the delimiter argument or the null_string argument, then the entire array_to_string function returns `NULL` .
`array_upper(anyarray, dimension int8)`	`array_upper(ARRAY[1, 2, 3, 4], 1) → 4`	Returns the upper bound of the requested array dimension. Note that Spanner does not support multidimensional arrays. The only dimension supported is `1` . For more information, see Working with arrays in PostgreSQL-dialect databases .
`array_length(anyarray, dimension int8)`	`array_length(ARRAY[1, 2, 3, 4], 1) → 4`	Returns the size of the array. Returns `NULL` for an empty or `NULL` array, or if the dimension is `NULL` . multidimensional arrays are not supported. The only dimension supported is `1` . For more information, see Working with arrays in PostgreSQL-dialect databases .
`array(subquery)`		Returns an `ARRAY` with one element for each row in the subquery. For more information, see Working with arrays in PostgreSQL-dialect databases .
`unnest(anyarray)`		Expands an `ARRAY` into a set of rows. For more information, see Working with arrays in PostgreSQL-dialect databases .

For details about the array aggregate function, see aggregate functions .

Array comparisons

Comparison syntax	Example	Description
`expression = ANY (anyarray)` `expression = SOME (anyarray)`	`2 = ANY(array[1, 2]) → true`	`ANY` / `SOME` array comparison construct. Returns `true` if the evaluated value of the expression on the left is equal to any of the array elements. There are no differences between `ANY` and `SOME` . `ANY` / `SOME` only supports the `=` operator.
`expression operator ALL (anyarray)`	`(1+2) >= ALL(array[1, 2, 3]) → true`	`ALL` array comparison construct. Returns `true` if the expression on the left evaluates to `true` when compared against all elements of the array with the specified operator. `=` , `<>` , `>` , `>=` , `<` , and `<=` operators are supported with `ALL` .

String functions

Function	Example / Notes	Description
`btrim(text)`	`btrim(' xyxyyx ') → xyxyyx`	Removes leading and trailing whitespace from the given string.
`btrim( string text, characters text)`	`btrim('xyxtrimyyx', 'xyz') → trim`	Removes the longest string containing only characters in characters from the start and end of string .
`concat(text, text[, ...])`	`concat('abcde', 2::text, '', 22::text) → abcde222` `concat('abcde', 2::text, NULL, 22::text) → NULL`	Concatenates the provided text arguments. Non-text arguments must first be explicitly cast to `text` . Any SQL NULL argument results in a SQL NULL result. `textcat` also concatenates text.
`length(text)`	Returns int8. `length('mike') → 4`	Returns the number of characters in the string.
`lower(text)`	`lower('PostgreSQL') → postgresql`	Converts the string to all lower case.
`lpad(text, int8)`	`lpad('hi', 7) → &blank;&blank;&blank;&blank;&blank;hi`	Extends the string to the specified length by prepending spaces. If the string is already longer than length then it is truncated on the right.
`lpad( string text, length int8, fill text)`	`lpad('hi', 7, 'xy') → xyxyxhi`	Extends the string to length length by prepending the characters fill , repeated. If the string is already longer than length then it is truncated on the right.
`ltrim(text)`	`ltrim(' test') → test`	Removes leading spaces from a string.
`ltrim( string text, characters text)`	`ltrim('zzzytest', 'xyz') → test`	Removes the longest string containing only characters in characters from the start of string .
`quote_ident( string text)`	`quote_ident('Example') → "Example"`	Given a string argument, returns a quoted identifier suitable for inclusion in SQL statements.
`regexp_replace( string text, pattern text, replacement text)`	`regexp_replace('Thomas', '.[mN]a.', 'M') → ThM`	Replaces substrings resulting from the first match of a POSIX regular expression. For more information, see the open source PostgreSQL POSIX Regular Expressions documentation .
`repeat(text, int8)`	Returns text. `repeat('Pg', 4) → PgPgPgPg`	Repeats a string the specified number of times.
`replace( string text, from text, to text)`	`replace('abcdefabcdef', 'cd', 'XX') → abXXefabXXef`	Replaces all occurrences in string of substring from with substring to .
`reverse(text)`	`reverse('abcde') → edcba`	Reverses the order of the characters in the string.
`rpad(text, int8)`	Returns text. In the following example, the result includes three trailing spaces. `rpad('hi', 5) → hi&blank;&blank;&blank;`	Extends the string to the specified length by appending spaces. If the string is already longer than the specified length then it is truncated.
`rpad( string text, length int8, fill text)`	Returns text. `rpad('hi', 5, 'xy') → hixyx`	Extends the string to length length by appending the characters fill , repeated if necessary. If the string is already longer than length then it is truncated.
`rtrim(text)`	`rtrim('test ') → test`	Removes trailing spaces from a string.
`rtrim( string text, characters text)`	`rtrim('testxxzx', 'xyz') → test`	Removes the longest string containing only characters in characters from the end of the string .
`spanner.soundex(text)`	`spanner.soundex('Ashcraft') -> A261`	Returns a `STRING` that represents the Soundex code for value. Soundex produces a phonetic representation of a string. It indexes words by sound, as pronounced in English. It's typically used to help determine whether two strings have similar English-language pronunciations, such as the family names Levine and Lavine , or the words to and too , The result of the Soundex consists of a letter followed by 3 digits. Non-latin characters are ignored. If the remaining string is empty after removing non-Latin characters, an empty string is returned.
`starts_with( string text, prefix text)`	Returns Boolean. `starts_with('alphabet', 'alph') → true`	Returns true if string starts with prefix .
`strpos( string text, substring text)`	Returns int8. `strpos('high', 'ig') → 2`	Returns first starting index of the specified substring within string , or zero if it's not present.
`substr( string text, start int8)`	`substr('alphabet', 3) → phabet`	Extracts the substring of the provided text starting at the specified character.
`substr( string text, start int8, count int8)`	`substr('alphabet', 3, 2) → ph`	Extracts the substring of string starting at the start character, and extending for count characters.
`substring( string text, pattern text)`	`substring('exampletext', 'tex.') → 'text'`	Extracts the substring that matches a POSIX regular expression. For more information, see the open source PostgreSQL POSIX Regular Expressions documentation .
`textcat( string text, string text)`	`textcat('abcde', '222') → abcde222` `textcat('abcde', NULL) → NULL`	Concatenates the text representations of the two arguments. Any SQL NULL argument results in a SQL NULL result.
`upper( string text)`	`upper('hello') → HELLO`	Converts the string to all upper case.

Binary string functions

Function	Example / Notes	Description
`btrim( bytes bytea, bytesremoved bytea)`	`btrim('\x1234567890'::bytea, '\x9012'::bytea) → \x345678`	Removes the longest string containing only bytes appearing in bytesremoved from the start and end of bytes .
`length(bytea)`	Returns int8. `length('\x1234567890'::bytea) → 5`	Returns the number of bytes in the binary string.
`sha256(bytea)`	`sha256('abc'::bytea) → ungWv48Bz+pBQUDeXa4iI7ADYaOWF3qctBD/YfIAFa0=`	Computes the SHA-256 hash of the binary string.
`sha512(bytea)`	`sha512('abc'::bytea) → 3a81oZNherrMQXNJriBBMRLm+k6JqX6iCp7u5ktV05ohkpkqJ0/BqDa6PCOj/uu9RU1EI2Q86A4qmslPpUyknw==`	Computes the SHA-512 hash of the binary string.
`substr( bytes bytea, start int8)`	`substr('\x1234567890'::bytea, 3) → \x567890`	Extracts the substring of bytes starting at the start byte.
`substr( bytes bytea, start int8, count int8)`	`substr('\x1234567890'::bytea, 3, 2) → \x5678`	Extracts the substring of bytes starting at the start byte, and extending for count bytes.

Hash functions

Function	Example / Notes	Description
`spanner.farm_fingerprint( value bytea \| text)`	`spanner.farm_fingerprint('abc') → 2640714258260161385`	Computes the fingerprint of value using the FarmHash Fingerprint64 algorithm.

Date and time functions and operators

This section describes the date and time functions and operators that are available in Spanner.

Date and time functions

Function	Example / Notes	Description
`current_date`	`SELECT CURRENT_DATE; Result: 2022-05-13`	Returns current `date` .
`current_timestamp`	`SELECT CURRENT_TIMESTAMP; Result: 2022-05-13T16:30:29.880850967Z`	Returns current date and time in `timestamptz` format.
`date_trunc(text, timestamptz)`	`date_trunc('day', timestamptz '2020-01-02 13:14:15+0') -> 2020-01-02 00:00:00-08`	Truncates a timestamp to the precision of the provided field. The truncation is done with respect to the default time zone (America/Los_Angeles)
`date_trunc(text, timestamptz, text)`	`date_trunc('day', timestamptz '2001-02-16 20:38:40+00', 'Australia/Sydney') -> 2001-02-16 08:00:00-05`	Truncates a timestamp to the precision of the provided field. The trunctation is done with respect to the provided time zone
`extract(field FROM source)`	`extract(decade from timestamptz '2001-01-01 01:00:00+00') -> 200`	Retrieves subfields from date and time values and returns values of type numeric. Source can use the date or timestamptz data type.
`make_date(int8, int8, int8)`	`make_date(2013, 7, 15) → 2013-07-15`	Creates date from year, month, and day fields (negative years signify BCE).
`now()`	`now() → 2022-05-02T19:17:45.145511221Z`	Returns current date and time in `timestamptz` format.
`to_timestamp(int8)`	`to_timestamp(1284352323) → 2010-09-13T04:32:03Z`	Converts Unix epoch (seconds since 1970-01-01 00:00:00+00) to `timestamptz` format.

Spanner specific date and time functions

Spanner has several functions that perform date or time math that accept INTERVAL values in TEXT form. You must use the spanner namespace to call these functions.

Function	Example / Notes	Description
`spanner.date(timestamptz, text)`	`spanner.date('2025-04-14 03:38:40+00'::timestamptz, 'America/New_York') -> 2025-04-13`	Extracts date from a timestamptz in a specified time zone. If a time zone value is not provided in the first parameter, the time zone value defaults to America/Los_Angeles . For example, `spanner.date('2025-04-14 23:38:40'::timestamptz, 'America/New_York') -> 2025-04-15`
`spanner.date_bin(text, timestamptz, timestamptz)`	`spanner.date_bin('15 minutes', timestamptz '2001-02-16 20:38:40Z', timestamptz '2001-02-16 20:05:00Z') -> 2001-02-16 20:35:00Z`	Bins input into a specified interval aligned with a specified origin.
`spanner.timestamptz_add(timestamptz, text)`	`spanner.timestamptz_add(timestamptz '2001-02-16 20:38:40Z', '1 day 3min') -> 2001-02-17 20:41:40Z`	Adds an interval to a timestamptz.
`spanner.timestamptz_subtract(timestamptz, text)`	`spanner.timestamptz_subtract(timestamptz '2001-02-16 20:38:40Z', '1 month 2 hours') -> 2001-01-16 18:38:40Z`	Subtracts an interval from a timestamptz.

Date and time operators

Operator	Example / Notes	Description
`date - date`	`date '2001-10-01' - date '2001-09-28' → 3`	Subtracts dates, returning the number of days that have elapsed.
`date - integer`	`date '2001-10-01' - 7 → 2001-09-24`	Subtracts a number of days from a date, returning the new date.
`date + integer`	`date '2001-09-28' + 7 → 2001-10-05`	Adds a number of days to a date, returning the new date.

Search functions

Spanner has several functions that perform full-text search operations. For more information, see Full-text search . For more information on search functions, see the GoogleSQL Search functions section.

Indexing

Functions that you can use to create search indexes.

Function	Example / Notes	Description
`spanner.token(value text/bytea)`	Returns a `spanner.tokenlist`	Constructs an exact match `tokenlist` value by tokenizing a text value verbatim to accelerate exact match expressions.
`spanner.tokenize_bool( value bool)`	Returns a `spanner.tokenlist` .	Constructs a boolean `tokenlist` value by tokenizing a `BOOL` value to accelerate boolean match expressions.
`spanner.tokenize_fulltext( value text/text[] [, language_tag text] [, content_type text] [, token_category text])`	Returns a `spanner.tokenlist` .	Constructs a full-text `tokenlist` value by tokenizing text for full-text matching.
`spanner.tokenize_jsonb( value jsonb)`	Returns a `spanner.tokenlist` .	Constructs a JSON `tokenlist` value by tokenizing a `JSONB` value to accelerate JSON predicates.
`spanner.tokenize_ngrams( value text [, ngram_size_min int8] [, ngram_size_max int8] [, remove_diacritics bool])`	Returns a `spanner.tokenlist` .	Constructs an n-gram `tokenlist` value by tokenizing a text value for matching n-grams.
`spanner.tokenize_number( value int8 [, comparison_type text] [, algorithm text] [, min int8] [, max int8] [, granularity int8] [, tree_base int8] [, ieee_precision int8])`	Returns a `spanner.tokenlist` .	Constructs a numeric `tokenlist` value by tokenizing numeric values to accelerate numeric comparison expressions.
`spanner.tokenize_substring( value text/text[] [, language_tag text] [, ngram_size_min int8] [, ngram_size_max int8] [, relative_search_types text[]] [, content_type text] [, remove_diacritics bool] [, short_tokens_only_for_anchors bool])`	Returns a `spanner.tokenlist` .	Constructs a substring `tokenlist` value by tokenizing text for substring matching.
`spanner.tokenlist_concat(tokens spanner.tokenlist[])`	Returns a `spanner.tokenlist` .	Displays a human-readable representation of tokens present in a `tokenlist` value for debugging purposes.

Retrieval and presentation

Functions that you can use to search for data, score the search result, or format the search result.

Function	Example / Notes	Description
`spanner.score( tokens spanner.tokenlist, query text [, dialect text] [, language_tag text] [, enhance_query bool] [, options jsonb])`	Returns a `float8` .	Calculates a relevance score of a `tokenlist` for a full-text search query. The higher the score, the stronger the match.
`spanner.score_ngrams( tokens spanner.tokenlist, ngrams_query text [, language_tag text] [, algorithm text])`		Calculates the relevance score of a `tokenlist` for a fuzzy search. The higher the score, the stronger the match.
`spanner.search( tokens spanner.tokenlist, query text [, dialect text] [, language_tag text] [, enhance_query bool])`	Returns a `bool` .	Returns `TRUE` if a full-text search query matches tokens.
`spanner.search_ngrams( tokens spanner.tokenlist, ngrams_query text [, language_tag text] [, min_ngrams int8] [, min_ngrams_percent float8])`	Returns a `bool` .	Checks whether enough n-grams match the tokens in a fuzzy search.
`spanner.search_substring( tokens spanner.tokenlist [, query text] [, language_tag text] [, relative_search_type text])`	Returns a `bool` .	Returns `TRUE` if a substring query matches tokens.
`spanner.snippet( value text, query text [, language_tag text] [, enhance_query bool] [, max_snippet_width int8] [, max_snippets int8] [, content_type text])`	Returns `jsonb` .	Gets a list of snippets that match a full-text search query.

Debugging

Functions that you can use for debugging.

Function	Example / Notes	Description
`spanner.debug_tokenlist( spanner.tokenlist)`	Returns `text` .	Displays a human-readable representation of tokens present in the `tokenlist` value for debugging purposes.

JSONB functions and operators

Spanner supports several JSONB functions and operators.

For more information, see the PostgreSQL JSONB documentation .

JSONB functions

Function	Example / Notes	Description
`jsonb_array_elements(JSONB)`	`jsonb_array_elements('[1, "abc", {"k": "v"}]'::jsonb)` /--------------------- \| jsonb_array_elements \| +----------------------+ \| '1' \| \| '"abc"' \| \| '{"k": "v"}' \| ---------------------/	Expands a `jsonb` array to a set of `jsonb` values. Returns multiple rows, with one element per row. Unlike open source PostgreSQL, this can only be called as a table valued function in the `FROM` clause.
`jsonb_build_array(ANY[, ...])`	`jsonb_build_array(1, 'abc') → [1, "abc"]`	Builds a `jsonb` array out of a variadic argument list.
`jsonb_build_object(TEXT, ANY[, ...])`	`jsonb_build_object('key1', 1, 'key2', 'abc') → {"key": 1, "key2": "abc"}`	Builds a `jsonb` object out of a variadic argument list. The argument list consists of alternating keys and values. The keys are of type `text` .
`jsonb_insert(target jsonb, path text[], new_value jsonb [, insert_after bool]) → jsonb`	`jsonb_insert('{"a": [0,1,2]}', '{a, 1}', '"new_value"') → {"a": [0, "new_value", 1, 2]}` `jsonb_insert('{"a": [0,1,2]}', '{a, 1}', '"new_value"', true) → {"a": [0, 1, "new_value", 2]}`	Returns `target` with `new_value` inserted as specified by `path` . If the item designated by the path is an array element, `new_value` is inserted before that item if `insert_after` is `false` (which is the default behavior), or after it if `insert_after` is `true` . If the item designated by the path is an object field, `new_value` is inserted only if the object does not already contain that key. All earlier steps in the path must exist, or the `target` is returned unchanged. As with the path-oriented operators, negative integers that appear in the path count from the end of JSON arrays. If the last path step is an array index that is out of range, the new value is added at the beginning of the array if the index is negative or at the end of the array if it is positive.
`jsonb_set(target jsonb, path text[], new_value jsonb [, create_if_missing bool]) → jsonb`	`jsonb_set('[{"f1":1,"f2":null},2,null,3]', '{0,f1}', '[2,3,4]', false) → [{"f1": [2, 3, 4], "f2": null}, 2, null, 3]` `jsonb_set('[{"f1":1,"f2":null},2]', '{0,f3}', '[2,3,4]') → [{"f1": 1, "f2": null, "f3": [2, 3, 4]}, 2]`	Returns `target` with the item designated by `path` replaced by `new_value` , or with `new_value` added if `create_if_missing` is true (which is the default behavior) and the item designated by the path does not exist. All earlier steps in the path must exist, or the `target` is returns unchanged. As with the path-oriented operators, negative integers that appear in the path count from the end of JSON arrays. If the last path step is an array index that is out of range. If `create_if_missing` is true and the last path step is an out-of-range array index, the new value is added to the beginning of the array (if the index is negative) or the end of the array (if the index is positive).
`jsonb_set_lax(target jsonb, path text[], new_value jsonb [, create_if_missing bool [, null_value_treatment text]]) → jsonb`	`jsonb_set_lax('[{"f1":1,"f2":null},2,null,3]', '{0,f1}', null) → [{"f1": null, "f2": null}, 2, null, 3]` `jsonb_set_lax('[{"f1":99,"f2":null},2]', '{0,f3}', null, true, 'return_target') → [{"f1": 99, "f2": null}, 2]`	If `new_value` is not `NULL` , this function behaves identically to `jsonb_set` . Otherwise, this function behaves according to the value of `null_value_treatment` which must be one of `raise_exception` , `use_json_null` , `delete_key` , or `return_target` . The default is `use_json_null` . `jsonb_set_lax` has the same behavior as `jsonb_set` unless the `null_value_treatment` parameter is a SQL null, then this function returns an error. If the `new_value` parameter is a SQL null then the `jsonb_set_lax` function returns behavior based on the `null_value_treatment` parameter.
`jsonb_strip_nulls(jsonb) → jsonb`	`jsonb_strip_nulls('[{"f1":1, "f2":null}, 2, null, 3]') → [{"f1": 1}, 2, null, 3]`	Deletes all object fields that have null values from a defined JSON array, recursively. Does not affect null values outside of object fields.
`to_jsonb(ANY)`	`to_jsonb(1.2334000) → 1.2334000` `to_jsonb(true) → true` `to_jsonb('abc'::varchar) → abc`	Converts the given value to `jsonb` .
`jsonb_object_keys(jsonb)`	`jsonb_object_keys('{"a":1, "b":{"c":1, "d":4}}'::jsonb) → {"a", "b"}`	Returns a set of keys in the top-level `jsonb` object as a set of `string` values. Unlike in open source PostgreSQL, this function can only be called in the `FROM` clause. The keys are de-duplicated and returned in length-first lexicographic order.
`jsonb_typeof(jsonb)`	`jsonb_typeof('-123.4') → number` `jsonb_typeof('{"a":1, "b":2}') → object` `jsonb_typeof('["a", "b", "c"]') → array` `jsonb_typeof('null'::jsonb) → null` `jsonb_typeof(NULL) IS NULL → true`	Returns the type of the top-level `jsonb` value as a text string. The possible types are `object` , `array` , `string` , `number` , `boolean` , and `NULL` . The `null` result shouldn't be confused with a SQL NULL, as the examples illustrate.

Spanner specific JSONB functions

Spanner has several JSONB functions that are not available in open source PostgreSQL. You must use the spanner namespace to call these functions.

Function	Example / Notes	Description
`spanner.bool_array(jsonb)`	`spanner.bool_array('[true, false]'::jsonb) → [true, false]` `spanner.bool_array('["true"]'::jsonb) → ERROR`	Returns an array of `boolean` values from a `jsonb` array. Raises an error if the argument is not an array of boolean values.
`spanner.float32_array(jsonb)`	`spanner.float32_array('[1, -2, 3.0]'::jsonb) → [1.0, -2.0, 3.0]` `spanner.float32_array('[1e100]'::jsonb) → ERROR`	Returns an array of `real` values from a `jsonb` array. Raises an error if the argument is not an array of number values in `real` domain.
`spanner.float64_array(jsonb)`	`spanner.float64_array('[1, -2, 3.0]'::jsonb) → [1.0, -2.0, 3.0]` `spanner.float64_array('[1e100]'::jsonb) → ERROR`	Returns an array of `real` values from a `jsonb` array. Raises an error if the argument is not an array of number values in `double precision` domain.
`spanner.int64_array(jsonb)`	`spanner.int64_array('[1, -2, 3.0]'::jsonb) → [1, -2, 3]` `spanner.int64_array('[1.1]'::jsonb) → ERROR`	Returns an array of `int8` values from a `jsonb` array. Raises an error if the argument is not an array of number values in `int8` domain.
`spanner.jsonb_query_array(jsonb)`	`spanner.jsonb_query_array('[1, "abc", {"k": "v"}]'::jsonb) → [1, "abc", {"k": "v"}]`	Returns an array of `jsonb` values from a `jsonb` array. Similar to jsonb_array_elements in PostgreSQL, except that it returns an array of values rather than a set of values.
`spanner.string_array(jsonb)`	`spanner.string_array('["a", "b", "c"]'::jsonb) → ['a', 'b', 'c']` `spanner.string_array('[null]'::jsonb) → ERROR`	Returns an array of `text` values from a `jsonb` array. Raises an error if the argument is not an array of string values.

JSONB operators

Operator	Example / Notes	Description
`->`	`'{"a": {"b":"bear"}}'::jsonb -> 'a' → {"b": "bear"}` `'[{"a":"apple"},{"b":"bear"},{"c":"cat"}]'::jsonb -> 2 → {"c": "cat"}` `'{"a": {"b":"bear"}}'::jsonb -> 'a' -> 'b' → bear` `'[{"a":"apple"},{"b":"bear"},{"c":"cat"}]'::jsonb -> -1 IS NULL → true`	Takes text or an integer as an argument and returns a `jsonb` object. When the argument is text, a `jsonb` object field is extracted with the given key. When the argument is an integer n , the n th element of a `jsonb` array is returned. The operator can be chained to extract nested values. See the third example provided. Negative indexes are not supported. If they're used, SQL NULL is returned. See the last example provided.
`->>`	`'{"a": {"b":"bear"}}'::jsonb ->> 'a' → {"b": "bear"}` `'[{"a":"apple"},{"b":"bear"},{"c":"cat"}]'::jsonb ->> 2 → {"c": "cat"}` `'[{"a":"apple"},{"b":"bear"},{"c":"cat"}]'::jsonb ->> -1 IS NULL → true`	Takes text or an integer as an argument and returns text. When the argument is text, a `jsonb` object field is extracted with the given key. When the argument is an integer n , the n th element of a `jsonb` array is returned. Negative indexes are not supported. If they're used, SQL NULL is returned. See the last example provided.
`@>`	`'{"a":1, "b":2}'::jsonb @> '{"b":2}'::jsonb → true`	Tests whether the left JSONB value contains the right JSONB value.
`<@`	`'{"b":2}'::jsonb <@ '{"a":1, "b":2}'::jsonb → true`	Tests whether the left JSONB value is contained in the right JSONB value.
`?`	`'{"a":1, "b":2}'::jsonb ? 'b' → true` `'["a", "b", "c"]'::jsonb ? 'b' → true`	Tests whether a text string exists as a top-level key or array element within a JSONB value.
`?\|`	`'{"a":1, "b":2, "c":3}'::jsonb ?\| array['b', 'd'] → true`	Tests whether any of the strings in a text array exist as top-level keys or array elements.
`?&`	`'["a", "b", "c"]'::jsonb ?& array['a', 'b'] → true`	Tests whether all of the strings in a text array exist as top-level keys or array elements.
`jsonb \|\| jsonb → jsonb`	`'["a", "b"]'::jsonb \|\| '["a", "d"]'::jsonb → ["a", "b", "a", "d"]` `'{"a": "b"}'::jsonb \|\| '{"c": "d"}'::jsonb → {"a": "b", "c": "d"}` `'[1, 2]'::jsonb \|\| '3'::jsonb → [1, 2, 3]` `'{"a": "b"}'::jsonb \|\| '42'::jsonb → [{"a": "b"}, 42]` To append an array to another array as a single entry, wrap it in an additional array layer: `'[1, 2]'::jsonb \|\| jsonb_build_array('[3, 4]'::jsonb) → [1, 2, [3, 4]]`	Concatenates two `jsonb` values. Concatenating two arrays generates an array containing all the elements of each input. Concatenating two objects generates an object containing the union of their keys, taking the second object's value when there are duplicate keys. All other cases are treated by converting a non-array input into a single-element array, and then processing them as two separate arrays. Does not operate recursively; only merges the top-level array or object structure.
`jsonb - text → jsonb`	`'{"a": "b", "c": "d"}'::jsonb - 'a' → {"c": "d"}` `'["a", "b", "c", "b"]'::jsonb - 'b' → ["a", "c"]`	Deletes a key and its value from a `jsonb` object, or matching string values from a `jsonb` array.
`jsonb - integer → jsonb`	`'["a", "b"]'::jsonb - 1 → ["a"]`	Deletes the array element with the specified index. Negative integers are counted from the end. This function expects an array value.
`jsonb #- text[] → jsonb`	`'["a", {"b":1}]'::jsonb #- '{1,b}' → ["a", {}]`	Deletes the field or array element at the specified path, where path elements can be either field keys or array indexes.

Interval functions

Function	Example / Notes	Description
`make_interval ( [ years int8 [, months int8 [, weeks int8 [, days int8 [, hours int8 [, mins int8 [, secs double precision ]]]]]]] ) → interval`	`SELECT make_interval(years => 1, months => 2, weeks => 3, days => 15, hours => 10, mins => 30, secs => 15.1) -> P1Y2M36DT10H30M15.1S`	Creates an interval from years, months, weeks, days, hours, minutes, and seconds fields. The default value for each of the fields is 0.
`-interval`	`SELECT -INTERVAL '1 year 2 months 15 days 45 seconds 500 microseconds';` `Result: P-1Y-2M-15DT-45.0005S`	Negate an interval.
`interval + interval → interval`	`SELECT INTERVAL '1 year 2 months 15 days' + INTERVAL '1 hour 15 minutes 45 seconds 500 milliseconds';` `Result: P1Y2M15DT1H15M45.5S`	Add intervals.
`interval - interval → interval`	`SELECT INTERVAL '1 year 2 months 10 hours 30 minutes' - INTERVAL '15 days 45 seconds 500 microseconds';` `Result: P1Y2M-15DT10H29M14.9995S`	Subtract intervals.
`interval * double precision → interval`	`SELECT INTERVAL '4 months 12 days 20 seconds' * 4.0;` `Result: P1Y4M48DT1M20S`	Multiply an interval by a scalar.
`interval / double precision → interval`	`SELECT INTERVAL '1 hour' / 2.5;` `Result: PT24M`	Divide an interval by a scalar.
`timestamptz + interval → timestamptz`	`SELECT TIMESTAMP WITH TIME ZONE '2021-12-18T10:00:00+00' + INTERVAL '2 months 15 days 40 minutes';` `Result: 2022-03-05T10:40:00Z`	Add an interval to a timestamp with a time zone.
`timestamptz - interval → timestamptz`	`SELECT TIMESTAMP WITH TIME ZONE '2024-12-18T10:00:00+00' - INTERVAL '2 months 15 days 40 minutes';` `Result: 2024-10-03T08:20:00Z`	Subtract an interval from a timestamp with a time zone.
`timestamptz - timestamptz → interval`	`SELECT TIMESTAMPTZ '2024-12-18T10:00:00+00' - TIMESTAMPTZ '2024-10-03T08:20:00Z'` `Result: PT1825H40M`	Subtract timestamps with a time zone. Unlike open source PostgreSQL, Spanner doesn't convert 24 hour time periods into days.
`justify_hours(interval) → interval`	`SELECT justify_hours(interval '50 hours 10 minutes')` `Result: P2DT2H10M` `SELECT justify_hours(interval '-12 day 50 hours 10 minutes')` `Result: P-9DT-21H-50M`	Normalizes 24-hour time periods into full days. Adjusts time and days to have the same sign.
`justify_days(interval) → interval`	`SELECT justify_days(interval '45 days 50 hours 10 minutes')` `Result: P1M15DT50H10M` `SELECT justify_days(interval '-1 year 45 days')` `Result: P-10M-15D`	Normalizes 30-day time periods into full months. Adjusts days and months to have the same sign.
`justify_interval(interval) → interval`	`SELECT justify_interval(INTERVAL '29 days 60 hours')` `Result: P1M1DT12H` `SELECT justify_interval(INTERVAL '-34 days 60 hours')` `Result: P-1M-1DT-12H`	Normalizes 24-hour time periods into full days, then 30-day time periods into full months. Adjusts all parts to have the same sign.
`extract(field FROM source) → numeric`	`SELECT extract(SECOND FROM INTERVAL '1 year 2 months 15 days 10 hours 30 minutes 15 seconds 100 milliseconds')` `Result: 15.100000`	Retrieves subfield from an interval value and returns a value of type numeric.

Aggregate functions

Function	Example / Notes	Description
`array_agg(anynonarray [ORDER BY input_sort_columns])`		Inserts the given values, including nulls, into an array. `input_sort_columns` , if specified, must have the same syntax as a query-level ORDER BY clause and is used to sort the inputs.
`avg(float4 \| float8 \| interval \| int8 \| numeric)`		Computes the average (arithmetic mean) of all the non-null input values.
`bit_and(int8)`		Computes the bitwise AND of all non-null input values.
`bit_or(int8)`		Computes the bitwise OR of all non-null input values.
`bool_and(bool)`		Returns true if all non-null input values are true, otherwise false.
`bool_or(bool)`		Returns true if any non-null input value is true, otherwise false.
`count(*)`	Returns int8.	Computes the number of input rows.
`count(bool \| bytea \| float4 \| float8 \| interval \| int8 \| text \| timestamptz)`	Returns int8.	Computes the number of input rows in which the input value is not null.
`every(bool)`		Equivalent to `bool_and()` .
`max(float4 \| float8 \| interval \| int8 \| numeric \| text \| timestamptz)`	Returns same type as input type.	Computes the maximum of the non-null input values.
`min(float4 \| float8 \| interval \| int8 \| numeric \| text \| timestamptz)`		Computes the minimum of the non-null input values.
`string_agg( value bytea, delimiter bytea)`		Concatenates the non-null input values into a string. Each value after the first is preceded by the corresponding delimiter (if it's not null).
`string_agg( value text, delimiter text [ORDER BY input_sort_columns])`		Concatenates the non-null input values into a string. Each value after the first is preceded by the corresponding delimiter (if it's not null). `input_sort_columns` , if specified, must have the same syntax as a query-level ORDER BY clause and is used to sort the inputs.
`sum(float4 \| float8 \| interval \| int8 \| numeric)`		Computes the sum of the non-null input values.

Conditional functions

Function	Example / Notes	Description
`coalesce(ANY REPEATED)`	`coalesce(NULL, 'abc', 'def') → 'abc'`	Returns the first of its arguments that is not null. Null is returned only if all arguments are null. It is often used to substitute a default value for null values when data is retrieved for display. The arguments must all use the same data type. The result has the same data type.
`greatest(ANY REPEATED)`	`greatest(6, 10, 3, 14, 2) → 14`	Returns the largest value from a list of any number of expressions. The expressions must all use the same data type. The result has the same data type. NULL values in the list are ignored. The result is NULL only if all the expressions evaluate to NULL.
`least(ANY REPEATED)`	`least(6, 10, 3, 14, 2) → 2`	Returns the smallest value from a list of any number of expressions. The expressions must all use the same data type. The result will have the same data type. NULL values in the list are ignored. The result is NULL only if all the expressions evaluate to NULL.
`nullif( value1 ANY, value2 ANY)`		Returns a null value if `value1` equals `value2` ; otherwise it returns `value1` . The two arguments must use comparable types. Specifically, they are compared exactly as if you had written `value1 = value2` , so there must be a suitable = operator available. The result has the same type as the first argument, but there is a subtle difference. What is actually returned is the first argument of the implied = operator, and in some cases that is promoted to match the second argument's type. For example, `NULLIF(1, 2.2)` yields a numeric, because there is no integer = numeric operator, only numeric = numeric.
`ISNULL`	`datatype ISNULL → boolean`	Tests whether value is null (non-standard syntax).
`NOTNULL`	`datatype NOTNULL → boolean`	Tests whether value is not null (non-standard syntax).
`IS TRUE`	`boolean IS TRUE → boolean true IS TRUE → true NULL::boolean IS TRUE → false (rather than NULL)`	Tests whether boolean expression yields true.
`IS NOT TRUE`	`boolean IS NOT TRUE → boolean true IS NOT TRUE → false NULL::boolean IS NOT TRUE → true (rather than NULL)`	Tests whether boolean expression yields false or unknown.
`IS FALSE`	`boolean IS FALSE → boolean true IS FALSE → false NULL::boolean IS FALSE → false (rather than NULL)`	Tests whether boolean expression yields false.
`IS NOT FALSE`	`boolean IS NOT FALSE → boolean true IS NOT FALSE → true NULL::boolean IS NOT FALSE → true (rather than NULL)`	Tests whether boolean expression yields true or unknown.
`IS UNKNOWN`	`boolean IS UNKNOWN → boolean true IS UNKNOWN → false NULL::boolean IS UNKNOWN → true (rather than NULL)`	Tests whether boolean expression yields unknown.
`IS NOT UNKNOWN`	`boolean IS NOT UNKNOWN → boolean true IS NOT UNKNOWN → true NULL::boolean IS NOT UNKNOWN → false (rather than NULL)`	Tests whether boolean expression yields true or false.

Pattern matching functions and operators

This section describes the pattern matching functions and operators that are available in Spanner.

Pattern matching functions

Function	Example / Notes	Description
`like( string bytea, pattern bytea)`	Returns Boolean.	Returns true if the string matches the supplied pattern. For more information about the `LIKE` expression, see the postgresql.org documentation .
`like( string text, pattern text)`	Returns Boolean.	Returns true if the string matches the supplied pattern. For more information about the `LIKE` expression, see the postgresql.org documentation .
`regexp_match( string text, pattern text)`	`regexp_match('exambarbequeple','(bar)(beque)') → {'bar', 'beque'}`	Returns an array of matching substrings within the first match of a POSIX regular expression pattern to a string. If there is no match, then the result is `NULL` . If there is a match, and the pattern contains parenthesized subexpressions, then the result is a text array whose n th element is the substring matching the n th parenthesized subexpression of the pattern (not counting non-capturing parentheses).
`regexp_match( string text, pattern text, flags text)`	`regexp_match('examBaRBeQUEple','(bar)(beque)', 'i') → {'BaR', 'BeQUE'}`	Returns an array of matching substrings within the first match of a POSIX regular expression pattern to a string. If there is no match, the result is `NULL` . If a match is found and the pattern contains parenthesized subexpressions, then the result is a text array whose n th element is the substring matching the n th parenthesized subexpression of the pattern (not counting non-capturing parentheses). The flags parameter contains zero or more single-letter flags that change the function's behavior. For more information about using flags, see the open source PostgreSQL Embedded-option Table documentation .
`regexp_split_to_array( string text, pattern text)`	`regexp_split_to_array('the quick brown fox jumps over the lazy dog','\s+') → {'the','quick',''brown','fox','jumps','over','the','lazy','dog'}`	Splits a string using a POSIX regular expression pattern as a delimiter. If there is no match to the pattern, the function returns the string. If there is at least one match, then for each match, the function returns the text from the end of the last match (or the beginning of the string) to the beginning of the match. When there are no more matches, the function returns the text from the end of the last match to the end of the string.
`regexp_split_to_array( string text, pattern text, flags text)`	`regexp_split_to_array('thE QUick bROWn FOx jUMPs ovEr The lazy dOG','e', 'i') → {'th',' QUick bROWn FOx jUMPs ov','r Th',' lazy dOG'}`	Splits a string using a POSIX regular expression pattern as a delimiter. If there is no match to the pattern, then the function returns the string. If there is at least one match, then for each match, the function returns the text from the end of the last match (or the beginning of the string) to the beginning of the match. When there are no more matches, the function returns the text from the end of the last match to the end of the string. The flags parameter contains zero or more single-letter flags that change the function's behavior. For more information about using flags, see the open source PostgreSQL Embedded-option Table documentation .

Pattern matching operators

Operator	Example / Notes	Description
`string text !~ pattern text`	`'thomas' !~ 't.*max' → true`	Tests whether a string text does not match a regular expression. Case sensitive.

Formatting functions

Function	Example / Notes	Description
`to_char( interval_value interval, format text)`	`SELECT to_char(INTERVAL '1 year 2 months 15 days 10 hours 30 minutes 15 seconds 100 milliseconds', 'YYYY-MM-DD HH24:MI:SS.MS');`	Converts interval to string according to the given date format. ^[1]
`to_char( number int8, format text)`	`to_char(125, '999') → 125`	Converts int8 to string according to the given format. ^[2]
`to_char( number numeric, format text)`	`to_char(-125.8, '999D99S') → 125.8-`	Converts numeric to string according to the given format. ^[2]
`to_char( number float4, format text)`	`to_char(125.8::float4, '999D9') → 125.8`	Converts float4 to string according to the given format. ^[2]
`to_char( number float8, format text)`	`to_char(125.8::float8, '999D9') → 125.8`	Converts float8 to string according to the given format. ^[2]
`to_char( timestamp timestamptz, format text)`	`to_char(timestamp '2002-04-20 17:31:12.66', 'HH12:MI:SS') → 05:31:12`	Converts timestamptz to string according to the given date format. ^[3]
`to_number( number text, format text)`	`to_number('12,454.8-', '99G999D9S') → -12454.8`	Converts string to numeric according to the given format. ^[2]
`to_date( date text, format text)`	`to_date('05 Dec 2000', 'DD Mon YYYY') → 2000-12-05`	Converts string to date according to the given date format. ^[3]
`to_timestamp( timestamp timestamptz, format text)`	`to_timestamp('05 Dec 2000', 'DD Mon YYYY') → 2000-12-05 00:00:00-05`	Converts string to timestamptz format according to the given date time format. ^[3]

^[1] For a list of supported numeric formatting, see Template Patterns for Numeric Formatting .

^[2] For a list of supported numeric formatting, see Template Patterns for Numeric Formatting .

^[3] For a list of supported date/time formatting, see Supported formats for date data type and Supported formats for timestamptz data type .

Sequence functions

Function	Example / Notes	Description
`nextval (varchar) → bigint`	`nextval ('MySequence')`	Takes a sequence name string and returns the next sequence value in the `bigint` data type. This function is only allowed in read-write transactions
`spanner.get_internal_sequence_state(varchar)`	`spanner.get_internal_sequence_state('MySequence')`	Gets the current sequence internal counter before bit reversal. As the sequence generates values, its internal counter changes. This function is useful when using import or export, and for migrations. If `nextval ('MySequence')` is never called on the sequence, then this function returns NULL.

Utility functions

Function	Example / Notes	Description
`spanner.generate_uuid()`	`SELECT spanner.generate_uuid() AS uuid → 4192bff0-e1e0-43ce-a4db-912808c32493`	Returns a random universally unique identifier (UUID) (Version 4) as a string. that Spanner can use for primary key columns. The returned string consists of 32 hexadecimal digits in five groups separated by hyphens in the form 8-4-4-4-12. The hexadecimal digits represent 122 random bits and 6 fixed bits, in compliance with RFC 4122 section 4.4. The returned string is lowercase.