Functions

Built-in Functions

This section provides documentation for all built-in functions available with the reference implementation. For each function the documentation provides

A one sentence explanation of the functions intent.
The function's signature that specifies data types and names for each input argument, and, the expected data type for the function's return value. A function signature consists of the function's name followed by a colon : then a space separated list of data types--one for each formal argument of the function--followed by an arrow -> followed by the function's return type, e.g., add: Integer Integer -> Integer is the signature for add which accepts 2 inputs both Integer's and returns one value of type Integer.
The function's header that specifies names for each of the function's formal arguments. Any documentation following the header can refer to the function's formal arguments by name.
The function's purpose statement that further expands on how the function behaves and specifies any pre- and/or post-conditions.
A list of examples calling the function and their expected results.

Unknown (`null` and `missing`) propagation

Unless otherwise stated all functions listed below propagate null and missing argument values. Propagating null and missing values is defined as: if any function argument is either null or missing the function will return null, e.g.,

CHAR_LENGTH(null)    -- `null`
CHAR_LENGTH(missing) -- `null` (also returns `null`)

CAST

Given a value and a target data type, attempt to coerce the value to the target data type.

Signature : CAST: Any DataType -> DataType

Where DataType is one of

missing
null
integer
boolean
float
decimal
timestamp
symbol
string
list
struct
bag

Note for handling CAST Operation on +inf, -inf, and nan, see [Special Float Value Handling](#Special Float Value Handling)

Header : CAST(exp AS dt)

Purpose : Given an expression, exp and the data type name, dt, evaluate expr to a value, v and alter the data type of v to DT(dt). If the conversion cannot be made the implementation signals an error.

The runtime support for casts is

Casting to null from
- null is a no-op
- missing returns null
- else error
Casting to missing from
- missing is a no-op
- null returns missing
- else error
Casting to integer from
- Integer: is a no-op
- Boolean: true returns 1, false returns 0
- String or Symbol: attempt to parse the content as an Integer and return the Integer, else error.
- Float or Decimal: gets narrowed to Integer
- else error
Casting to boolean from
- Boolean is a no-op
- Integer or Decimal or Float: if v is a representation of the number 0 (e.g., 0 or -0 or 0e0 or 0d0 et.) then false else true
- String or Symbol: true unless v matches--ignoring character case--the Ion string "false" or matches--ignoring character case-- the Ion symbol 'false' then return false
- else error
Casting to float from
- Float is a no-op
- Boolean: false return 0.0, true returns 1.0
- Integer or Decimal: convert to Float and return
- String or Symbol: attempt to parse as Float and return the Float value, else error
- else error
Casting to decimal from
- Decimal is a no-op
- Boolean: return 1d0 if true, 0d0 if false
- String or Symbol: attemp to parse as Decimal and return Decimal value, else error
- else error
Casting to timestamp from
- Timestamp is a no-op
- String or Symbol: attemp to parse as Timestamp and return the Timestamp value, else error
- else error
Casting to symbol from
- Symbol is a no-op
- Integer or Float or Decimal: narrow to Integer and return the value as a Symbol, i.e, a Symbol with the same sequence of digits as characters
- String: return the String as a Symbol, i.e., represent the same sequence of characters as a Symbol
- Boolean: return 'true' for true and 'false' for false
- Timestamp: return the Symbol representation of the Timestamp, i.e., represent the same sequence of digits and characters as a Symbol
- else error
Casting to string from
- String is a no-op
- Integer or Float or Decimal: narrow to Integer and return the value as a String, i.e, a String with the same sequence of digits as characters
- Symbol: return the String as a Symbol, i.e., represent the same sequence of characters as a String
- Boolean: return "true" for true and "false" for false
- Timestamp: return the String representation of the Timestamp, i.e., represent the same sequence of digits and characters as a String
- else error
Casting to list from
- List is a no-op
- Bag: return a list with the same elements. The order of the elements in the resulting list is unspecified.
- else error
Casting to struct from
- Struct is a no-op
- else error
Casting to bag from
- Bag is a no-op
- List: return a list with the same elements. The order of the elements in the resulting bag is unspecified.

Special Float Value Handling

The following table list explicitly on behavior of CAST Function when expression supplied is one of +inf, -inf, or nan.

target type	can cast	cast quality	expected result/error
missing	false	N/A	EVALUATOR_INVALID_CAST
null	false	N/A	EVALUATOR_INVALID_CAST
integer	false	N/A	EVALUATOR_CAST_FAILED
boolean	true	LOSSY	true
float	true	LOSSLESS	+inf, -inf, nan
decimal	false	N/A	EVALUATOR_CAST_FAILED
timestamp	false	N/A	EVALUATOR_INVALID_CAST
symbol	true	LOSSLESS	'Infinity', '-Infinity', 'NaN'
string	true	LOSSLESS	"Infinity", "-Infinity", "NaN"
list	false	N/A	EVALUATOR_INVALID_CAST
struct	false	N/A	EVALUATOR_INVALID_CAST
bag	false	N/A	EVALUATOR_INVALID_CAST

Examples :

-- Unknowns propagation 
CAST(null    AS null)    -- null 
CAST(missing AS null)    -- null 
CAST(missing AS missing) -- null
CAST(null    AS missing) -- null
CAST(null    AS boolean) -- null (null AS any data type name result to null)
CAST(missing AS boolean) -- null (missing AS any data type name result to null)

-- any value that is not an unknown cannot be cast to `null` or `missing`
CAST(true AS null)    -- error 
CAST(true AS missing) -- error
CAST(1    AS null)    -- error
CAST(1    AS missing) -- error

-- AS boolean 
CAST(true      AS boolean) -- true no-op
CAST(0         AS boolean) -- false
CAST(1         AS boolean) -- true
CAST(`1e0`     AS boolean) -- true (float)
CAST(`1d0`     AS boolean) -- true (decimal)
CAST('a'       AS boolean) -- false
CAST('true'    AS boolean) -- true (PartiQL string 'true')
CAST(`'true'`  AS boolean) -- true (Ion symbol `'true'`)
CAST(`'false'` AS boolean) -- false (Ion symbol `'false'`)

-- AS integer
CAST(true   AS integer) -- 1
CAST(false  AS integer) -- 0
CAST(1      AS integer) -- 1
CAST(`1d0`  AS integer) -- 1
CAST(`1d3`  AS integer) -- 1000
CAST(1.00   AS integer) -- 1
CAST('12'   AS integer) -- 12
CAST('aa'   AS integer) -- error
CAST(`'22'` AS integer) -- 22
CAST(`'x'`  AS integer) -- error

-- AS flaot
CAST(true   AS float) -- 1e0
CAST(false  AS float) -- 0e0
CAST(1      AS float) -- 1e0
CAST(`1d0`  AS float) -- 1e0
CAST(`1d3`  AS float) -- 1000e0
CAST(1.00   AS float) -- 1e0
CAST('12'   AS float) -- 12e0
CAST('aa'   AS float) -- error
CAST(`'22'` AS float) -- 22e0
CAST(`'x'`  AS float) -- error

-- AS decimal
CAST(true   AS decimal) -- 1.
CAST(false  AS decimal) -- 0.
CAST(1      AS decimal) -- 1.
CAST(`1d0`  AS decimal) -- 1. (REPL printer serialized to 1.)
CAST(`1d3`  AS decimal) -- 1d3
CAST(1.00   AS decimal) -- 1.00
CAST('12'   AS decimal) -- 12.
CAST('aa'   AS decimal) -- error
CAST(`'22'` AS decimal) -- 22.
CAST(`'x'`  AS decimal) -- error

-- AS timestamp
CAST(`2001T`                      AS timestamp) -- 2001T
CAST('2001-01-01T'                AS timestamp) -- 2001-01-01
CAST(`'2010-01-01T00:00:00.000Z'` AS timestamp) -- 2010-01-01T00:00:00.000Z
CAST(true                         AS timestamp) -- error
CAST(2001                         AS timestamp) -- error

-- AS symbol 
CAST(`'xx'`                     AS symbol) -- xx (`'xx'` is an Ion symbol)
CAST('xx'                       AS symbol) -- xx ('xx' is a string)
CAST(42                         AS symbol) -- '42'
CAST(`1e0`                      AS symbol) -- '1'
CAST(`1d0`                      AS symbol) -- '1'
CAST(true                       AS symbol) -- 'true'
CAST(false                      AS symbol) -- 'false'
CAST(`2001T`                    AS symbol) -- '2001T'
CAST(`2001-01-01T00:00:00.000Z` AS symbol) -- '2001-01-01T00:00:00.000Z`

-- AS string 
CAST(`'xx'`                     AS string) -- "xx" (`'xx'` is an Ion symbol)
CAST('xx'                       AS string) -- "xx" ('xx' is a string)
CAST(42                         AS string) -- "42"
CAST(`1e0`                      AS string) -- "1.0"
CAST(`1d0`                      AS string) -- "1"
CAST(true                       AS string) -- "true"
CAST(false                      AS string) -- "false"
CAST(`2001T`                    AS string) -- "2001T"
CAST(`2001-01-01T00:00:00.000Z` AS string) -- "2001-01-01T00:00:00.000Z"

-- AS struct 
CAST(`{ a: 1 }` AS struct) -- { a:1 }
CAST(true       AS struct) -- err

-- AS list
CAST(`[1, 2, 3]`        AS list) -- [ 1, 2, 3 ] (REPL does not diplay the parens and commas)
CAST(<<'a', { 'b':2 }>> AS list) -- [ a, { 'b':2 } ] (REPL does not diplay the parens and commas)
CAST({ 'b':2 }          AS list) -- error

-- AS bag
CAST([1,2,3]      AS bag) -- <<1,2,3>> (REPL does not display << >> and commas)
CAST([1,[2],3]    AS bag) -- <<1,[2],3>> (REPL does not display << >> and commas)
CAST(<<'a', 'b'>> AS bag) -- <<'a', 'b'>> (REPL does not display << >> and commas)

CHAR_LENGTH, CHARACTER_LENGTH

Counts the number of characters in the specified string, where 'character' is defined as a single unicode code point.

Note: CHAR_LENGTH and CHARACTER_LENGTH are synonyms.

Signature : CHAR_LENGTH: String -> Integer

CHARACTER_LENGTH: String -> Integer

Header : CHAR_LENGTH(str)

CHARACTER_LENGTH(str)

Purpose : Given a String value str return the number of characters (code points) in str.

Examples :

CHAR_LENGTH('')          -- 0
CHAR_LENGTH('abcdefg')   -- 7
CHAR_LENGTH('😁😞😸😸') -- 4 (non-BMP unicode characters)
CHAR_LENGTH('eࠫ')         -- 2 (because 'eࠫ' is two codepoints: the letter 'e' and combining character U+032B)

COALESCE

Evaluates the arguments in order and returns the first non unknown, i.e. first non-null or non-missing. This function does not propagate null and missing.

Signature : COALESCE: Any Any ... -> Any

Header : COALESCE(exp, [exp ...])

Purpose : Given a list of 1 or more arguments, evaluates the arguments left-to-right and returns the first value that is not an unknown (missing or null).

Examples :

COALESCE(1)                -- 1
COALESCE(null)             -- null
COALESCE(null, null)       -- null
COALESCE(missing)          -- null
COALESCE(missing, missing) -- null
COALESCE(1, null)          -- 1
COALESCE(null, null, 1)    -- 1
COALESCE(null, 'string')   -- 'string'
COALESCE(missing, 1)       -- 1

DATE_ADD

Given a data part, a quantity and a timestamp, returns an updated timestamp by altering datetime part by quantity

Signature : DATE_ADD: DateTimePart Integer Timestamp -> Timestamp

Where DateTimePart is one of

year
month
day
hour
minute
second

Header : DATE_ADD(dp, q, timestamp)

Purpose : Given a data part dp, a quantity q, and, an Ion timestamp timestamp returns an updated timestamp by applying the value for q to the dp component of timestamp. Positive values for q add to the timestamp's dp, negative values subtract.

The value for timestamp as well as the return value from DATE_ADD must be a valid Ion Timestamp

Examples :

DATE_ADD(year, 5, `2010-01-01T`)                -- 2015-01-01 (equivalent to 2015-01-01T)
DATE_ADD(month, 1, `2010T`)                     -- 2010-02T (result will add precision as necessary)
DATE_ADD(month, 13, `2010T`)                    -- 2011-02T
DATE_ADD(day, -1, `2017-01-10T`)                -- 2017-01-09 (equivalent to 2017-01-09T)
DATE_ADD(hour, 1, `2017T`)                      -- 2017-01-01T01:00-00:00
DATE_ADD(hour, 1, `2017-01-02T03:04Z`)          -- 2017-01-02T04:04Z
DATE_ADD(minute, 1, `2017-01-02T03:04:05.006Z`) -- 2017-01-02T03:05:05.006Z
DATE_ADD(second, 1, `2017-01-02T03:04:05.006Z`) -- 2017-01-02T03:04:06.006Z

DATE_DIFF

Given a datetime part and two valid timestamps returns the difference in datetime parts.

Signature : DATE_DIFF: DateTimePart Timestamp Timestamp -> Integer

See DATE_ADD for the definition of DateTimePart

Header : DATE_DIFF(dp, t1, t2)

Purpose : Given a datetime part dp and two timestamps t1 and t2 returns the difference in value for dp part of t1 with t2. The return value is a negative integer when the dp value of t1 is greater than the dp value of t2, and, a positive integer when the dp value of t1 is less than the dp value of t2.

Examples :

DATE_DIFF(year, `2010-01-01T`, `2011-01-01T`)            -- 1
DATE_DIFF(year, `2010T`, `2010-05T`)                     -- 4 (2010T is equivalent to 2010-01-01T00:00:00.000Z)
DATE_DIFF(month, `2010T`, `2011T`)                       -- 12
DATE_DIFF(month, `2011T`, `2010T`)                       -- -12
DATE_DIFF(day, `2010-01-01T23:00T`, `2010-01-02T01:00T`) -- 0 (need to be at least 24h apart to be 1 day apart)

EXISTS

Given a PartiQL value returns true if and only if the value is a non-empty container(bag, sexp, list or struct), returns false otherwise.

Signature : EXISTS: Container -> Boolean

Header : EXISTS(val)

Purpose : Given a PartiQL value val, if val is

a container with size > 0, EXISTS(val) returns true;
a container with size = 0, EXISTS(val) returns false;
not a container, EXISTS(val) throws an error.

Note:

This function does not propagate null and missing.

Examples :

EXISTS(`[]`)        -- false (empty list)
EXISTS(`[1, 2, 3]`) -- true (non-empty list)
EXISTS(`[missing]`) -- true (non-empty list)
EXISTS(`{}`)        -- false (empty struct)
EXISTS(`{ a: 1 }`)  -- true (non-empty struct)
EXISTS(`()`)        -- false (empty s-expression)
EXISTS(`(+ 1 2)`)   -- true (non-empty s-expression)
EXISTS(`<<>>`)      -- false (empty bag)
EXISTS(`<<null>>`)  -- true (non-empty bag)
EXISTS(1)           -- error
EXISTS(`2017T`)     -- error
EXISTS(null)        -- error
EXISTS(missing)     -- error

EXTRACT

Given a datetime part and a datetime type returns then datetime's datetime part value.

Signature : EXTRACT: ExtractDateTimePart DateTime -> Integer

where ExtractDateTimePart is one of

year
month
day
hour
minute
second
timezone_hour
timezone_minute

and DateTime type is one of

DATE
TIME
TIMESTAMP

Note that ExtractDateTimePart differs from DateTimePart in DATE_ADD.

Header : EXTRACT(edp FROM t)

Purpose : Given a datetime part, edp, and a datetime type t return t's value for edp. This function allows for t to be unknown (null or missing) but not edp. If t is unknown the function returns null.

Examples :

EXTRACT(YEAR FROM `2010-01-01T`)                           -- 2010
EXTRACT(MONTH FROM `2010T`)                                -- 1 (equivalent to 2010-01-01T00:00:00.000Z)
EXTRACT(MONTH FROM `2010-10T`)                             -- 10
EXTRACT(HOUR FROM `2017-01-02T03:04:05+07:08`)             -- 3
EXTRACT(MINUTE FROM `2017-01-02T03:04:05+07:08`)           -- 4
EXTRACT(TIMEZONE_HOUR FROM `2017-01-02T03:04:05+07:08`)    -- 7
EXTRACT(TIMEZONE_MINUTE FROM `2017-01-02T03:04:05+07:08`)  -- 8
EXTRACT(YEAR FROM DATE '2010-01-01')                       -- 2010
EXTRACT(MONTH FROM DATE '2010-01-01')                      -- 1
EXTRACT(DAY FROM DATE '2010-01-01')                        -- 1
EXTRACT(HOUR FROM DATE '2010-01-01')                       -- 0
EXTRACT(MINUTE FROM DATE '2010-01-01')                     -- 0
EXTRACT(SECOND FROM DATE '2010-01-01')                     -- 0
EXTRACT(HOUR FROM TIME '23:12:59')                         -- 23
EXTRACT(MINUTE FROM TIME '23:12:59')                       -- 12
EXTRACT(SECOND FROM TIME '23:12:59')                       -- 59
EXTRACT(SECOND FROM TIME (2) '23:12:59.128')               -- 59.13
EXTRACT(TIMEZONE_HOUR FROM TIME WITH TIME ZONE '23:12:59-08:30')    -- -8
EXTRACT(TIMEZONE_MINUTE FROM TIME WITH TIME ZONE '23:12:59-08:30')  -- -30

Note that timezone_hour and timezone_minute are not supported for DATE and TIME (without time zone) type.

`FILTER_DISTINCT`

Signature : FILTER_DISTINCT: Container -> Bag

Header : FILTER_DISTINCT(c)

Purpose : Returns a bag of distinct values contained within a bag, list, sexp, or struct. If the container is a struct, the field names are not considered.

Examples :

FILTER_DISTINCT([0, 0, 1])                  -- <<0, 1>>
FILTER_DISTINCT(<<0, 0, 1>>)                -- <<0, 1>>
FILTER_DISTINCT(SEXP(0, 0, 1))              -- <<0, 1>>
FILTER_DISTINCT({'a': 0, 'b': 0, 'c': 1})   -- <<0, 1>>

LOWER

Given a string convert all upper case characters to lower case characters.

Signature : LOWER: String -> String

Header : LOWER(s)

Purpose : Given a string, s, alter every upper case character in s to lower case. Any non-upper cased characters remain unchanged. This operation does rely on the locale specified by the runtime configuration. The implementation, currently, relies on Java's String.toLowerCase() documentation.

Examples :

LOWER('AbCdEfG!@#$') -- 'abcdefg!@#$'

MAKE_DATE

Given the integer values for the year, month and day returns the associated date.

Signature : MAKE_DATE: Year_Int Month_Int Day_Int -> Date

where Year_Int, Month_Int, Day_Int are the Integers representing year, month and day respectively for the date.

Header : MAKE_DATE(year, month, day)

Purpose : Given integer values for year, month and day, returns an associated Date. This function allows arguments to be unknowns i.e. (null or missing).

Examples :

MAKE_DATE(2021, 02, 28)                  -- 2021-02-28
MAKE_DATE(2020, 02, 29)                  -- 2020-02-29
MAKE_DATE(2021, 12, 31)                  -- 2021-12-31
MAKE_DATE(null, 02, 28)                  -- null
MAKE_DATE(2021, null, 28)                -- null
MAKE_DATE(2021, 02, null)                -- null
MAKE_DATE(missing, 02, 28)               -- null
MAKE_DATE(2021, missing, 28)             -- null
MAKE_DATE(2021, 02, missing)             -- null

MAKE_TIME

Given the values for the hour (int), minute (int), second (BigDecimal) and optionally for timezone minutes (int) returns the associated time.

Signature : MAKE_TIME: Hour_Int Minute_Int Second_BigDecimal (optional)TimezoneMinutes_Int -> Time

where Hour_Int, Minute_Int, TimezoneMinutes_Int are the Integers representing hour, minute and timezone minutes respectively for the time. Second_BigDecimal is the BigDecimal representing the second with fraction of the time.

Header : MAKE_TIME(hour, minute, second, timezoneMinutes?)

Purpose : Given values for hour (int), minute (int), second (BigDecimal) and optionally for timezone minutes (int), returns an associated Time. This function allows arguments to be unknowns i.e. (null or missing).

Examples :

MAKE_TIME(23, 59, 59.)                   -- 23:59:59
MAKE_TIME(23, 59, 59.12345, 330)         -- 23:59:59.12345+05:30
MAKE_TIME(23, 59, 59.12345, -330)        -- 23:59:59.12345-05:30
MAKE_TIME(null, 02, 28.)                 -- null
MAKE_TIME(12, null, 28.)                 -- null
MAKE_TIME(21, 02, null)                  -- null
MAKE_TIME(missing, 02, 28.)              -- null
MAKE_TIME(23, missing, 59.)              -- null
MAKE_TIME(21, 02, missing)               -- null
MAKE_TIME(21, 02, 28., null)             -- null
MAKE_TIME(21, 02, 28., missing)          -- null

SIZE

Given any container data type (i.e., list, structure or bag) return the number of elements in the container.

Signature : SIZE: Container -> Integer

Header : SIZE(c)

Purpose : Given a container, c, return the number of elements in the container. If the input to SIZE is not a container the implementation throws an error.

Examples :

SIZE(`[]`)                   -- 0
SIZE(`[null]`)               -- 1
SIZE(`[1,2,3]`)              -- 3
SIZE(<<'foo', 'bar'>>)       -- 2
SIZE(`{foo: bar}`)           -- 1 (number of key-value pairs)
SIZE(`[{foo: 1}, {foo: 2}]`) -- 2
SIZE(12)                     -- error

NULLIF

Given two expressions return null if the two expressions evaluate to the same value, else, returns the result of evaluating the first expression

Signature : NULLIF: Any Any -> Any

Header : NULLIF(e1, e2)

Purpose : Given two expression, e1 and e2, evaluate both expression to get v1 and v2 respectively, and return null if and only if v1 equals v2, else, return v1. The implementation of NULLIF uses = for equality, i.e., v1 and v2 are considered equal by NULLIF if and only if v1 = v2 is true.

Note, NULLIF does not propagate unknowns (null and missing).

Examples :

NULLIF(1, 1)             -- null
NULLIF(1, 2)             -- 1
NULLIF(1.0, 1)           -- null
NULLIF(1, '1')           -- 1
NULLIF([1], [1])         -- null
NULLIF(1, NULL)          -- 1
NULLIF(NULL, 1)          -- null
NULLIF(null, null)       -- null
NULLIF(missing, null)    -- null
NULLIF(missing, missing) -- null

SUBSTRING

Given a string, a start index and optionally a length, returns the substring from the start index up to the end of the string, or, up to the length provided.

Signature : SUBSTRING: String Integer [ NNegInteger ] -> String

Where NNegInteger is a non-negative integer, i.e., 0 or greater.

Header : SUBSTRING(str, start [ , length ])

SUBSTRING(str FROM start [ FOR length ])

Purpose : Given a string, str, a start position, start and optionally a length, length, extract the characters (code points) starting at index start and ending at (start + length) - 1. If length is omitted, then proceed till the end of str.

The first character of str has index 1.

Examples :

SUBSTRING("123456789", 0)      -- "123456789"
SUBSTRING("123456789", 1)      -- "123456789"
SUBSTRING("123456789", 2)      -- "23456789"
SUBSTRING("123456789", -4)     -- "123456789"
SUBSTRING("123456789", 0, 999) -- "123456789" 
SUBSTRING("123456789", 0, 2)   -- "1"
SUBSTRING("123456789", 1, 999) -- "123456789"
SUBSTRING("123456789", 1, 2)   -- "12"
SUBSTRING("1", 1, 0)           -- ""
SUBSTRING("1", 1, 0)           -- ""
SUBSTRING("1", -4, 0)          -- ""
SUBSTRING("1234", 10, 10)      -- ""

TO_STRING

Given a timestamp and a format pattern return a string representation of the timestamp in the given format.

Signature : TO_STRING: Timestamp TimeFormatPattern -> String

Where TimeFormatPattern is a String with the following special character interpretations

Format	Example	Description
`yy`	69	2-digit year
`y`	1969	4-digit year
`yyyy`	1969	Zero padded 4-digit year
`M`	1	Month of year
`MM`	01	Zero padded month of year
`MMM`	Jan	Abbreviated month year name
`MMMM`	January	Full month of year name
`MMMMM`	J	Month of year first letter (NOTE: not valid for use with to_timestamp function)
`d`	2	Day of month (1-31)
`dd`	02	Zero padded day of month (01-31)
`a`	AM	AM or PM of day
`h`	3	Hour of day (1-12)
`hh`	03	Zero padded hour of day (01-12)
`H`	3	Hour of day (0-23)
`HH`	03	Zero padded hour of day (00-23)
`m`	4	Minute of hour (0-59)
`mm`	04	Zero padded minute of hour (00-59)
`s`	5	Second of minute (0-59)
`ss`	05	Zero padded second of minute (00-59)
`S`	0	Fraction of second (precision: 0.1, range: 0.0-0.9)
`SS`	06	Fraction of second (precision: 0.01, range: 0.0-0.99)
`SSS`	060	Fraction of second (precision: 0.001, range: 0.0-0.999)
...	...	...
`SSSSSSSSS`	060000000	Fraction of second (maximum precision: 1 nanosecond, range: 0.0-0.999999999)
`n`	60000000	Nano of second
`X`	+07 or Z	Offset in hours or "Z" if the offset is 0
`XX` or `XXXX`	+0700 or Z	Offset in hours and minutes or "Z" if the offset is 0
`XXX` or `XXXXX`	+07:00 or Z	Offset in hours and minutes or "Z" if the offset is 0
`x`	+07	Offset in hours
`xx` or `xxxx`	+0700	Offset in hours and minutes
`xxx` or `xxxxx`	+07:00	Offset in hours and minutes

Header : TO_STRING(t,f)

Purpose : Given a timestamp, t, and a format pattern, f, as a String, return a string representation of t in format f.

Examples :

TO_STRING(`1969-07-20T20:18Z`,  'MMMM d, y')                    -- "July 20, 1969"
TO_STRING(`1969-07-20T20:18Z`, 'MMM d, yyyy')                   -- "Jul 20, 1969"
TO_STRING(`1969-07-20T20:18Z`, 'M-d-yy')                        -- "7-20-69"
TO_STRING(`1969-07-20T20:18Z`, 'MM-d-y')                        -- "07-20-1969"
TO_STRING(`1969-07-20T20:18Z`, 'MMMM d, y h:m a')               -- "July 20, 1969 8:18 PM"
TO_STRING(`1969-07-20T20:18Z`, 'y-MM-dd''T''H:m:ssX')           -- "1969-07-20T20:18:00Z"
TO_STRING(`1969-07-20T20:18+08:00Z`, 'y-MM-dd''T''H:m:ssX')     -- "1969-07-20T20:18:00Z"
TO_STRING(`1969-07-20T20:18+08:00`, 'y-MM-dd''T''H:m:ssXXXX')   -- "1969-07-20T20:18:00+0800"
TO_STRING(`1969-07-20T20:18+08:00`, 'y-MM-dd''T''H:m:ssXXXXX')  -- "1969-07-20T20:18:00+08:00"

TO_TIMESTAMP

Given a string convert it to a timestamp. This is the inverse operation of TO_STRING

Signature : TO_TIMESTAMP: String [ TimeFormatPattern ] -> Timestamp

See definition of TimeFormatPattern in TO_STRING.

Header : TO_TIMESTAMP(str[ , f ])

Purpose : Given a string, str, and an optional format pattern, f, as a String return a timestamp whose values are extracted from str using f.

If the <format pattern> argument is omitted, <string> is assumed to be in the format of a standard Ion timestamp. This is the only recommended way to parse an Ion timestamp using this function.

Zero padding is optional when using a single format symbol (e.g. y, M, d, H, h, m, s) but required for their zero padded variants (e.g. yyyy, MM, dd, HH, hh, mm, ss).

Special treatment is given to 2-digit years (format symbol yy). 1900 is added to values greater than or equal to 70 and 2000 is added to values less than 70.

Month names and AM/PM specifiers are case-insensitive.

Examples :

Single argument parsing an Ion timestamp:

TO_TIMESTAMP('2007T')                         -- `2007T`
TO_TIMESTAMP('2007-02-23T12:14:33.079-08:00') -- `2007-02-23T12:14:33.079-08:00`
TO_TIMESTAMP('2016', 'y')                     -- `2016T`
TO_TIMESTAMP('2016', 'yyyy')                  -- `2016T`
TO_TIMESTAMP('02-2016', 'MM-yyyy')            -- `2016-02T`
TO_TIMESTAMP('Feb 2016', 'MMM yyyy')          -- `2016-02T`
TO_TIMESTAMP('Febrary 2016', 'MMMM yyyy')     -- `2016-02T`

Notes:

All issues for PartiQL's TO_TIMESTAMP function.

Internally, this is implemented with Java 8's java.time package. There are a few differences between Ion's timestamp and the java.time package that create a few hypothetically infrequently encountered caveats that do not really have good workarounds at this time.

The Ion specification allows for explicitly signifying an unknown timestamp with a negative zero offset (i.e. the -00:00 at the end of 2007-02-23T20:14:33.079-00:00) but Java 8's DateTimeFormatter doesn't recognize this. Hence, unknown offsets specified in this manner will be parsed as if they had an offset of +00:00, i.e. UTC. To avoid this issue when parsing Ion formatted timestamps, use the single argument variant of TO_TIMESTAMP. There is no workaround for custom format patterns at this time.
DateTimeFormatter is capable of parsing UTC offsets to the precision of seconds, but Ion Timestamp's precision for offsets is minutes. TimestampParser currently handles this by throwing an exception when an attempt is made to parse a timestamp with an offset that does does not land on a minute boundary. For example, parsing this timestamp would throw an exception: May 5, 2017 8:52pm +08:00:01 while May 5, 2017 8:52pm +08:00:00 would not.
Ion Java's Timestamp allows specification of offsets up to +/- 23:59, while an exception is thrown by DateTimeFormatter for any attempt to parse an offset greater than +/- 18:00. For example, attempting to parse: May 5, 2017 8:52pm +18:01 would cause and exception to be thrown. (Note: the Ion specification does indicate minimum and maximum allowable values for offsets.) In practice this will not be an issue for systems that do not abuse the offset portion of Timestamp because real-life offsets do not exceed +/- 12h.

TRIM

Trims leading and/or trailing characters from a String.

Signature : TRIM: [ String ] String -> String

Header : TRIM([[LEADING|TRAILING|BOTH r] FROM] str)

Purpose : Given a string, str, and an optional set of characters to remove, r, specified as a string, return the string with any character from r found at the beginning or end of str removed.

If r is not provided it defaults to ' '.

Examples :

TRIM('       foobar         ')               -- 'foobar'
TRIM('      \tfoobar\t         ')            -- '\tfoobar\t'
TRIM(LEADING FROM '       foobar         ')  -- 'foobar         '
TRIM(TRAILING FROM '       foobar         ') -- '       foobar'
TRIM(BOTH FROM '       foobar         ')     -- 'foobar'
TRIM(BOTH '😁' FROM '😁😁foobar😁😁')         -- 'foobar'
TRIM(BOTH '12' FROM '1112211foobar22211122') -- 'foobar'

UPPER

Given a string convert all lower case characters to upper case characters.

Signature : UPPER: String -> String

Header : UPPER(str)

Purpose : Given a string, str, alter every upper case character is str to lower case. Any non-lower cases characters remain unchanged. This operation does rely on the locale specified by the runtime configuration. The implementation, currently, relies on Java's String.toLowerCase() documentation.

Examples :

UPPER('AbCdEfG!@#$') -- 'ABCDEFG!@#$'

UTCNOW

Returns the current time in UTC as a timestamp.

Signature : UTCNOW: -> Timestamp

Header : UTCNOW()

Purpose : Return the current time in UTC as a timestamp

Examples :

UTCNOW() -- 2017-10-13T16:02:11.123Z

UNIX_TIMESTAMP

With no timestamp argument, returns the number of seconds since the last epoch ('1970-01-01 00:00:00' UTC).

With a timestamp argument, returns the number of seconds from the last epoch to the given timestamp (possibly negative).

Signature : UNIX_TIMESTAMP: [Timestamp] -> Integer|Decimal

Header : UNIX_TIMESTAMP([timestamp])

Purpose : UNIX_TIMESTAMP() called without a timestamp argument returns the number of whole seconds since the last epoch ('1970-01-01 00:00:00' UTC) as an Integer using UTCNOW.

UNIX_TIMESTAMP() called with a timestamp argument returns returns the number of seconds from the last epoch to the timestamp argument. If given a timestamp before the last epoch, UNIX_TIMESTAMP will return the number of seconds before the last epoch as a negative number. The return value will be a Decimal if and only if the given timestamp has a fractional seconds part.

Examples :

UNIX_TIMESTAMP()                            -- 1507910531 (if current time is `2017-10-13T16:02:11Z`; # of seconds since last epoch as an Integer)
UNIX_TIMESTAMP(`2020T`)                     -- 1577836800 (seconds from 2020 to the last epoch as an Integer)
UNIX_TIMESTAMP(`2020-01T`)                  -- ''
UNIX_TIMESTAMP(`2020-01-01T`)               -- ''
UNIX_TIMESTAMP(`2020-01-01T00:00Z`)         -- ''
UNIX_TIMESTAMP(`2020-01-01T00:00:00Z`)      -- ''
UNIX_TIMESTAMP(`2020-01-01T00:00:00.0Z`)    -- 1577836800. (seconds from 2020 to the last epoch as a Decimal)
UNIX_TIMESTAMP(`2020-01-01T00:00:00.00Z`)   -- ''
UNIX_TIMESTAMP(`2020-01-01T00:00:00.000Z`)  -- ''
UNIX_TIMESTAMP(`2020-01-01T00:00:00.100Z`)  -- 1577836800.1
UNIX_TIMESTAMP(`1969T`)                     -- -31536000 (timestamp is before last epoch)

FROM_UNIXTIME

Converts the given unix epoch into a timestamp.

Signature : FROM_UNIXTIME: Integer|Decimal -> Timestamp

Header : FROM_UNIXTIME(unix_timestamp)

Purpose : When given a non-negative numeric value, returns a timestamp after the last epoch. When given a negative numeric value, returns a timestamp before the last epoch. The returned timestamp has fractional seconds depending on if the value is a decimal.

Examples :

FROM_UNIXTIME(-1)           -- `1969-12-31T23:59:59-00:00`      (negative unix_timestamp; returns timestamp before last epoch)
FROM_UNIXTIME(-0.1)         -- `1969-12-31T23:59:59.9-00:00`    (unix_timestamp is decimal so timestamp has fractional seconds)
FROM_UNIXTIME(0)            -- `1970-01-01T00:00:00.000-00:00`
FROM_UNIXTIME(0.001)        -- `1970-01-01T00:00:00.001-00:00`  (decimal precision to fractional second precision) 
FROM_UNIXTIME(0.01)         -- `1970-01-01T00:00:00.01-00:00`
FROM_UNIXTIME(0.1)          -- `1970-01-01T00:00:00.1-00:00`
FROM_UNIXTIME(1)            -- `1970-01-01T00:00:01-00:00`
FROM_UNIXTIME(1577836800)   -- `2020-01-01T00:00:00-00:00`      (unix_timestamp is Integer so no fractional seconds)

Math Functions

CEIL / CEILING

Returns the nearest integer greater than or equal to the input.

Signature : CEIL/CEILING: Numeric -> Numeric

Examples:

CEIL(1.1) = 2
CEIL(-42.8) = -42
CEIL(`+inf`) = `+inf` -- Float
CEIL(`-inf`) = `-inf` -- Float
CEIL(`nan`) = `nan` -- Float

FLOOR

Returns the nearest integer less than or equal to the input.

Signature : FLOOR: Numeric -> Numeric

Examples:

FLOOR(1.1) = 1
FLOOR(-42.8) = -43
FLOOR(`+inf`) = `+inf` -- Float
FLOOR(`-inf`) = `-inf` -- Float
FLOOR(`nan`) = `nan` -- Float

This is the template for writing documentations for an PartiQL built-in function.

There are 5 parts to a function's documentation

One sentence statement -- much like the first sentence of a Java method's Javadoc
Signature -- the type signature of the function should use data types already defined on define a local data type using a WHERE clause
Header -- function with formal parameters only that we will use in the next step
Purpose -- english explanation of what the function does referring to formal argument names from Header. State any pre- and post-conditions as well as any exceptions to default behaviour e.g., propagation of unknowns null and missing
Examples -- examples and their expected results. Make sure the examples cover the preceding explanations and always include examples for uncommon behaviour

Here is an example for the imaginary function add

ADD

Given 1 or more values return ther sum

Signature : ADD: PosInt PosInt -> PosInt

where PosInt is a positive Integer

Header : ADD(v1, v2 ... vn)

Purpose : Given 1 or more values v1 .. vn return their sum. The summation proceeds from left-to-right. If any of the values passed is not a PosInt the function returns the current sum up to that point.

Examples :

ADD(1)         -- 1 (wrap extra explanations with parens)
ADD(1,2)       -- 3
ADD(1,2,"a",3) -- 3
ADD()          -- 0
ADD("a")       -- 0

Design & Development Documents

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Functions

Built-in Functions

Unknown (`null` and `missing`) propagation

CAST

Special Float Value Handling

CHAR_LENGTH, CHARACTER_LENGTH

COALESCE

DATE_ADD

DATE_DIFF

EXISTS

EXTRACT

`FILTER_DISTINCT`

LOWER

MAKE_DATE

MAKE_TIME

SIZE

NULLIF

SUBSTRING

TO_STRING

TO_TIMESTAMP

TRIM

UPPER

UTCNOW

UNIX_TIMESTAMP

FROM_UNIXTIME

Math Functions

CEIL / CEILING

FLOOR

ADD

Clone this wiki locally

Functions

Built-in Functions

Unknown (null and missing) propagation

CAST

Special Float Value Handling

CHAR_LENGTH, CHARACTER_LENGTH

COALESCE

DATE_ADD

DATE_DIFF

EXISTS

EXTRACT

FILTER_DISTINCT

LOWER

MAKE_DATE

MAKE_TIME

SIZE

NULLIF

SUBSTRING

TO_STRING

TO_TIMESTAMP

TRIM

UPPER

UTCNOW

UNIX_TIMESTAMP

FROM_UNIXTIME

Math Functions

CEIL / CEILING

FLOOR

ADD

Clone this wiki locally

Unknown (`null` and `missing`) propagation

`FILTER_DISTINCT`