Struct IntermediateQuery 

pub struct IntermediateQuery {

    pub name: String,
    pub return_type: String,
    pub returns_list: bool,
    pub nullable: bool,
    pub arguments: Vec<IntermediateArgument>,
    pub description: Option<String>,
    pub sql_source: Option<String>,
    pub function: Option<String>,
    pub auto_params: Option<IntermediateAutoParams>,
    pub deprecated: Option<IntermediateDeprecation>,
    pub jsonb_column: Option<String>,
    pub relay: bool,
    pub count: bool,
    pub inject: IndexMap<String, String>,
    pub read_routing: ReadRouting,
    pub cache_ttl_seconds: Option<u64>,
    pub additional_views: Vec<String>,
    pub requires_role: Option<String>,
    pub requires_actor: Vec<String>,
    pub relay_cursor_type: Option<String>,
    pub rest: Option<IntermediateRest>,
    pub rest_stream: bool,
    pub pagination_order: Option<String>,
}

Query definition in intermediate format

Fields

name: String

Query name (e.g., “users”)

return_type: String

Return type name (e.g., “User”)

returns_list: bool

Returns a list?

nullable: bool

Result is nullable?

arguments: Vec<IntermediateArgument>

Query arguments

description: Option<String>

Query description (from docstring)

sql_source: Option<String>

SQL source (table/view name)

function: Option<String>

Back this root field with a declared function instead of a relation (#1329).

Names an entry in the functions section whose trigger is request:query. Mutually exclusive with sql_source — a field resolves one way, and a precedence rule would be one more thing to remember wrong.

Everything that lowers into SQL is refused beside it rather than ignored: relay, count, pagination_order, inject_params, rest_stream, jsonb_column, a non-default read_routing, and an explicitly declared auto_params flag. Each of those is a setting the compiler would otherwise carry into an artifact where nothing reads it, and one of them (inject_params) is a tenant-scoping control — a silently dropped security setting is the failure this seam exists to refuse.

Project-wide [query_defaults] are not refused, only defaulted away: an author who set limit = true for the project did not declare it for this query, and failing their compile over a default they never wrote here would be refusing the wrong statement.

auto_params: Option<IntermediateAutoParams>

Auto-generated parameters config

deprecated: Option<IntermediateDeprecation>

Deprecation info (from @deprecated directive)

jsonb_column: Option<String>

JSONB column name for extracting data (e.g., “data”) Used for tv_* (denormalized JSONB tables) pattern

relay: bool

Whether this is a Relay connection query. When true, the compiler wraps results in { edges: [{ node, cursor }], pageInfo } and generates first/after/last/before arguments instead of limit/offset. Requires returns_list = true and sql_source to be set.

count: bool

Emit a <name>Count(where): Int! sibling for this list query (#938).

Opt-in rather than automatic: the count is a second SELECT COUNT(*) that scans the whole filtered set, which is a real cost on a hot list endpoint and is wasted on every list that is not rendered with page numbers.

Requires returns_list = true and sql_source. Refused on relay = true — a connection already carries totalCount.

inject: IndexMap<String, String>

Server-injected parameters: SQL column name → source expression (e.g. "jwt:org_id"). Not exposed as GraphQL arguments.

Wire key is inject_params, matching the compiled schema. The two names were unified because the split is what let three SDKs emit a key that bound to nothing (#806); see the inject_params_serde module for the full reasoning.

read_routing: ReadRouting

Where this query’s reads may be served from (#957).

any (the default) follows server policy; primary refuses replicas — and the result cache — for reads where staleness is a correctness problem; replica opts out of the read-your-writes pin and may carry its own max_lag_ms.

FraiseQL defines the shape and enforces it; an authoring language emits it — a @reads_from(...) directive is one spelling.

cache_ttl_seconds: Option<u64>

Per-query result cache TTL in seconds. Overrides the global cache TTL for this query.

additional_views: Vec<String>

Additional database views this query reads beyond the primary sql_source.

Used for correct cache invalidation when a query JOINs or reads multiple views. Each entry is validated as a safe SQL identifier at schema compile time.

requires_role: Option<String>

Role required to execute this query and see it in introspection.

requires_actor: Vec<String>

Actor classes permitted to execute this query — an allow-list (#966).

Empty means unrestricted. Tokens are the snake_case ActorType names (human_user, service_account, ai_agent, system_job); an unrecognised one is a compile error rather than a silently-ignored restriction, which is the difference between a typo and an open operation.

relay_cursor_type: Option<String>

Relay cursor column type: "uuid" for UUID PKs, "int64" (or absent) for bigint PKs. Only meaningful when relay = true.

rest: Option<IntermediateRest>

REST route override for this query. See IntermediateRest.

rest_stream: bool

Allow this query to be exported as a stream over REST (#958).

Separate from rest rather than a key inside it, because that block’s path is mandatory: nesting the flag there would make every author who wants an export also override the route.

pagination_order: Option<String>

Override the order this query’s LIMIT/OFFSET pages are cut in (#1303).

Two spellings, and absence is the third:

• a column name — order pages by this column, which must be unique over sql_source or the pages still overlap;
• the literal "none" — emit no default ordering. This is how a view that carries its own ORDER BY keeps it: the compiler cannot read a view’s body, so the opt-out is declared rather than inferred, and declaring it also silences the compile warning about non-deterministic pages, because the author answered it.
• absent — the compiler derives the entity identity, which is what almost every query wants.

"none" is therefore not usable as a column name here. A relation column actually called none must be ordered by through the SQL view instead.

Trait Implementations

impl Clone for IntermediateQuery

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for IntermediateQuery

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for IntermediateQuery

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for IntermediateQuery

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for IntermediateQuery

impl PartialEq for IntermediateQuery

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for IntermediateQuery

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for IntermediateQuery