Edit

OData expression syntax reference for Azure AI Search

  • Article

Azure AI Search uses OData expressions as parameters throughout the API. Most commonly, OData expressions are used for the $orderby and $filter parameters. These expressions can be complex, containing multiple clauses, functions, and operators. However, even simple OData expressions like property paths are used in many parts of the Azure AI Search REST API. For example, path expressions are used to refer to subfields of complex fields everywhere in the API, such as when listing subfields in a suggester, a scoring function, the $select parameter, or even fielded search in Lucene queries.

This article describes all these forms of OData expressions using a formal grammar. There is also an interactive diagram to help visually explore the grammar.

Formal grammar

We can describe the subset of the OData language supported by Azure AI Search using an EBNF (Extended Backus-Naur Form) grammar. Rules are listed "top-down", starting with the most complex expressions, and breaking them down into more primitive expressions. At the top are the grammar rules that correspond to specific parameters of the Azure AI Search REST API:

  • $filter, defined by the filter_expression rule.
  • $orderby, defined by the order_by_expression rule.
  • $select, defined by the select_expression rule.
  • Field paths, defined by the field_path rule. Field paths are used throughout the API. They can refer to either top-level fields of an index, or subfields with one or more complex field ancestors.

After the EBNF is a browsable syntax diagram that allows you to interactively explore the grammar and the relationships between its rules.

/* Top-level rules */

filter_expression ::= boolean_expression

order_by_expression ::= order_by_clause(',' order_by_clause)*

select_expression ::= '*' | field_path(',' field_path)*

field_path ::= identifier('/'identifier)*


/* Shared base rules */

identifier ::= [a-zA-Z_][a-zA-Z_0-9]*


/* Rules for $orderby */

order_by_clause ::= (field_path | sortable_function) ('asc' | 'desc')?

sortable_function ::= geo_distance_call | 'search.score()'


/* Rules for $filter */

boolean_expression ::=
    collection_filter_expression
    | logical_expression
    | comparison_expression
    | boolean_literal
    | boolean_function_call
    | '(' boolean_expression ')'
    | variable

/* This can be a range variable in the case of a lambda, or a field path. */
variable ::= identifier | field_path

collection_filter_expression ::=
    field_path'/all(' lambda_expression ')'
    | field_path'/any(' lambda_expression ')'
    | field_path'/any()'

lambda_expression ::= identifier ':' boolean_expression

logical_expression ::=
    boolean_expression ('and' | 'or') boolean_expression
    | 'not' boolean_expression

comparison_expression ::= 
    variable_or_function comparison_operator constant | 
    constant comparison_operator variable_or_function

variable_or_function ::= variable | function_call

comparison_operator ::= 'gt' | 'lt' | 'ge' | 'le' | 'eq' | 'ne'


/* Rules for constants and literals */

constant ::=
    string_literal
    | date_time_offset_literal
    | integer_literal
    | float_literal
    | boolean_literal
    | 'null'

string_literal ::= "'"([^'] | "''")*"'"

date_time_offset_literal ::= date_part'T'time_part time_zone

date_part ::= year'-'month'-'day

time_part ::= hour':'minute(':'second('.'fractional_seconds)?)?

zero_to_fifty_nine ::= [0-5]digit

digit ::= [0-9]

year ::= digit digit digit digit

month ::= '0'[1-9] | '1'[0-2]

day ::= '0'[1-9] | [1-2]digit | '3'[0-1]

hour ::= [0-1]digit | '2'[0-3]

minute ::= zero_to_fifty_nine

second ::= zero_to_fifty_nine

fractional_seconds ::= integer_literal

time_zone ::= 'Z' | sign hour':'minute

sign ::= '+' | '-'

/* In practice integer literals are limited in length to the precision of
the corresponding EDM data type. */
integer_literal ::= sign? digit+

float_literal ::=
    sign? whole_part fractional_part? exponent?
    | 'NaN'
    | '-INF'
    | 'INF'

whole_part ::= integer_literal

fractional_part ::= '.'integer_literal

exponent ::= 'e' sign? integer_literal

boolean_literal ::= 'true' | 'false'


/* Rules for functions */

function_call ::=
    geo_distance_call |
    boolean_function_call

geo_distance_call ::=
    'geo.distance(' variable ',' geo_point ')'
    | 'geo.distance(' geo_point ',' variable ')'

geo_point ::= "geography'POINT(" lon_lat ")'"

lon_lat ::= float_literal ' ' float_literal

boolean_function_call ::=
    geo_intersects_call |
    search_in_call |
    search_is_match_call

geo_intersects_call ::=
    'geo.intersects(' variable ',' geo_polygon ')'

/* You need at least four points to form a polygon, where the first and
last points are the same. */
geo_polygon ::=
    "geography'POLYGON((" lon_lat ',' lon_lat ',' lon_lat ',' lon_lat_list "))'"

lon_lat_list ::= lon_lat(',' lon_lat)*

search_in_call ::=
    'search.in(' variable ',' string_literal(',' string_literal)? ')'

/* Note that it is illegal to call search.ismatch or search.ismatchscoring
from inside a lambda expression. */
search_is_match_call ::=
    'search.ismatch'('scoring')?'(' search_is_match_parameters ')'

search_is_match_parameters ::=
    string_literal(',' string_literal(',' query_type ',' search_mode)?)?

query_type ::= "'full'" | "'simple'"

search_mode ::= "'any'" | "'all'"

Syntax diagram

To visually explore the OData language grammar supported by Azure AI Search, try the interactive syntax diagram:

OData syntax diagram for Azure AI Search

See also