Custom resolvers

When a generated field cannot express a query, write the resolver yourself and keep Strawchemy's data access by calling a repository from it.

Fetching one record

Called as a function, strawchemy.field() builds a repository-backed resolver for you. Decorate your own method with @strawchemy.field to build that repository yourself and add your own logic around it — a lookup by title, say, which strawchemy.field() alone has no argument for:

from sqlalchemy import select
from strawchemy import StrawchemySyncRepository


@strawberry.type
class Query:
    @strawchemy.field
    def get_post_by_title(self, info: strawberry.Info, title: str) -> PostType | None:
        repo = StrawchemySyncRepository(PostType, info, filter_statement=select(Post).where(Post.title == title))
        return repo.get_one_or_none().graphql_type_or_none()

filter_statement narrows the query before Strawchemy's own filtering, ordering and field-selection logic runs on top of it. A statement that only adds WHERE clauses to select(Post) has them copied into the query; any other statement is joined on the primary key, so its own ORDER BY doesn't order the result. The repository takes none of the field's arguments or the mapper's configuration: pass query_hook, execution_options or deterministic_ordering to it when you need them. get_one_or_none() runs it and returns a GraphQLResult; graphql_type_or_none() converts that into PostType | None.

Fetching by primary key needs no filter_statement: pass the key as a keyword argument to get_by_id():

@strawchemy.field
def get_post_by_id(self, info: strawberry.Info, id: int) -> PostType:
    repo = StrawchemySyncRepository(PostType, info)
    return repo.get_by_id(id=id).graphql_type()

get_by_id() takes the primary key's field name and value as keyword arguments — here id, matching Post.id. graphql_type() is the non-optional counterpart to graphql_type_or_none(): it raises instead of returning None when there's no match.

Returning a list

@strawchemy.field
def published_posts(self, info: strawberry.Info) -> list[PostType]:
    repo = StrawchemySyncRepository(PostType, info, filter_statement=select(Post).where(Post.published_at.is_not(None)))
    return repo.list().graphql_list()

The repository has four methods for fetching data, each paired with its own conversion call:

• get_one(), get_one_or_none() — return at most one result, and raise MultipleResultsFound if several rows match; graphql_type() then raises QueryResultError on no match, while graphql_type_or_none() returns None
• get_by_id() — returns a single result filtered on primary key
• list() — returns every matching result

instance and instances on the returned GraphQLResult give the model instances instead. Relationships the GraphQL query selected are not set on them: declare the ones your code reads in QueryHook(load=...).

See async sessions for the same resolvers written against StrawchemyAsyncRepository.

See query hooks for constraining every query against a type, and for loading data a custom field needs.