New ACM paper, free-tier cloud, and open-source license

Lesson 7: Understanding query patterns

Patterns as constraints

  • A pattern comprises a conjunction of any number of consecutive statements.

  • A pattern is equivalent to a list of constraints about the terms involved, which must be simultaneously satisfied. A term can be either a variable or a literal.

  • Type inference determines the type of each variable by solving the pattern as a constraint satisfaction problem.

  • If a variable in a query pattern has no possible types, the pattern fails validation and an exception is thrown.

Relation patterns

  • A relation with any number of roleplayers can be represented using a relation tuple with the same number of elements. This allows for unary, binary, ternary, and n-ary relations.

    (role-1: $a, role-2: $b, role-3: $c, role-4: $d, ...) isa n-ary-relation;
  • Roleplayers can be omitted from relation tuples, allowing partial tuple representations of relations. This will match those relations regardless of the omitted roleplayers.

    (role-1: $a, role-2: $b) isa ternary-relation;
  • Roles can be omitted from relation tuples, allowing TypeDB to use role inference to infer them. This will match the roleplayers in any combination of the relation’s roles.

    ($a, $b) isa example-relation;
  • A relation’s type can be omitted after a relation tuple, allowing TypeDB to use type inference to infer the possible types. This will match any relations with the supplied role names.

    (role-1: $a, role-2: $b);

Logical operators

  • A disjunction of patterns is made by placing those patterns in separate blocks interleaved with the or keyword, terminated with a ; delimiter.

        # branch 1
    } or {
        # branch 2
  • A negation of a pattern is made by placing it in a block preceded by the not keyword and followed by a ; delimiter.

    not {
        # negated pattern
  • An is statement is used to specify that two variables represent the same data object.

    $a is $b;

Value comparisons

  • The equality operators == and != are used to compare values of all types.

    $attribute-1 == $attribute-2;
  • The ordering operators >, >=, <, and <= are used to compare numeric values.

    $numeric-1 > $numeric-2;
  • The contains operator is used to specify that one string contains another string in a case-insensitive manner.

    $string contains "sub-string";
  • The like operator is used to specify that a string matches a Java regex pattern.

    $string like "^java regex pattern$";

Value expressions

  • Value variables are declared with a ? prefix. They exist only within the scope of the query and can be read out in the results of read queries or written to attribute instances with write queries.

  • The assignment operator = is used to specify the value of a value variable.

    ?value = $attribute;
  • Numeric values can be computed using arithmetic operators ^, *, /, %, +, and -, as well as arithmetic functions min, max, round, floor, ceil, and abs.

Solution set semantics

  • If a query pattern can be split into multiple independent sub-patterns with no shared variables, then those sub-patterns are resolved against the schema independently, and all possible combinations of data matching each sub-pattern are returned.

  • If the constraints on two variables are identical, then each pair of data instances matched by those variables will be returned twice.

  • These behaviours can lead to unintended query results if not accounted for.

Further learning

Learn how to structure query results in the optimal manner, covering result grouping, aggregations, sorting, and pagination.

Learn how to design schemas for TypeDB using the conceptual PERA model, including common design patterns and pitfalls.

Read the TypeQL patterns reference, covering conjunctions, disjunctions, negations, and pattern matching behaviour.

Provide Feedback