datalog/functions

built-in function reference for datalog (CozoScript). all functions are available in rule bodies, expressions, and aggregation contexts

math

function	description
`add` / `+`	addition
`sub` / `-`	subtraction
`mul` / `*`	multiplication
`div` / `/`	division
`mod` / `%`	modulo
`pow` / `^`	exponentiation
`minus`	unary negation
`abs`	absolute value
`signum`	sign (-1, 0, 1)
`sqrt`	square root
`floor`, `ceil`, `round`	rounding
`exp`, `exp2`	exponential
`ln`, `log2`, `log10`	logarithmic
`sin`, `cos`, `tan`	trigonometric
`asin`, `acos`, `atan`, `atan2`	inverse trigonometric
`sinh`, `cosh`, `tanh`	hyperbolic
`asinh`, `acosh`, `atanh`	inverse hyperbolic
`deg_to_rad`, `rad_to_deg`	angle conversion
`haversine`, `haversine_deg_input`	geographic distance

comparison and boolean

function	description
`eq` / `==`	equality
`neq` / `!=`	inequality
`gt` / `>`, `ge` / `>=`	greater than / or equal
`lt` / `<`, `le` / `<=`	less than / or equal
`max`, `min`	extrema
`and` / `&&`	logical conjunction
`or` / `\|\|`	logical disjunction
`negate` / `!`	logical negation
`assert`	returns true or raises error

string

function	description
`length`	character count
`concat` / `++`	concatenation
`str_includes`	substring check
`lowercase`, `uppercase`	case conversion
`trim`, `trim_start`, `trim_end`	whitespace removal
`starts_with`, `ends_with`	prefix/suffix check
`unicode_normalize`	normalization
`chars`	split into character list
`from_substrings`	join from list

list

function	description
`list`	construct list
`is_in`	membership test
`first`, `last`	endpoint access
`get`, `maybe_get` / `->`	indexed access
`length`	list size
`slice`	subsequence
`concat` / `++`	concatenation
`prepend`, `append`	add element
`reverse`	reverse order
`sorted`	sort
`chunks`, `chunks_exact`	partition
`windows`	sliding window
`union`, `intersection`, `difference`	set operations on lists

vector

relevant for embedding-based queries over particles

function	description
`vec`	construct vector from numbers
`rand_vec`	random vector generation
`l2_normalize`	L2 normalization
`l2_dist`	Euclidean distance
`ip_dist`	inner product distance
`cos_dist`	cosine distance

// find particles with embeddings similar to a query vector
?[particle, distance] := *particle_embeddings{particle, embedding},
                         distance = cos_dist(embedding, vec([0.1, 0.5, 0.3, ...]))
:sort distance
:limit 10

JSON

function	description
`json`	convert to JSON
`is_json`	type check
`json_object`	create object
`dump_json`	serialize to string
`parse_json`	parse from string
`get`, `maybe_get` / `->`	access element
`set_json_path`	modify value
`remove_json_path`	delete element
`json_to_scalar`	convert to scalar
`concat` / `++`	deep merge

regex

function	description
`regex_matches`	pattern match test
`regex_replace`	replace first match
`regex_replace_all`	replace all
`regex_extract`	extract all matches
`regex_extract_first`	extract first match

timestamp

function	description
`now`	current timestamp
`format_timestamp`	to RFC3339 string
`parse_timestamp`	from RFC3339 string
`validity`	create validity object for time-travel

type checking and conversion

function	description
`coalesce` / `~`	first non-null value
`to_string`, `to_float`, `to_int`	type conversion
`to_unity`, `to_bool`, `to_uuid`	specialized conversion
`uuid_timestamp`	extract UUID timestamp
`is_null`, `is_int`, `is_float`, `is_num`	type checks
`is_finite`, `is_infinite`, `is_nan`	number checks
`is_bytes`, `is_list`, `is_string`, `is_uuid`, `is_json`	type checks

random

function	description
`rand_float`	random float in [0,1]
`rand_bernoulli`	random boolean with probability
`rand_int`	random integer in range
`rand_choose`	random element from list
`rand_uuid_v1`, `rand_uuid_v4`	UUID generation
`rand_vec`	random vector

aggregation operators

used in rule heads for grouping and reduction

operator	description	semi-lattice
`count`	row count	no
`sum`	sum values	no
`min`	minimum	yes
`max`	maximum	yes
`mean`	arithmetic mean	no
`collect`	gather into list	no
`choice`	pick arbitrary value	yes
`min_cost`	minimum with associated cost	yes
`shortest`	shortest path collector	yes
`bit_and`, `bit_or`	bitwise aggregation	yes
`union_is_in`	set union	yes

semi-lattice aggregation allows self-recursion — required for recursive shortest-path and reachability queries over the cybergraph

see datalog/queries for how aggregation works in rule heads. see datalog/algorithms for fixed-rule graph algorithms

datalog/functions.md