Classes · 2 min read

DuckDbStatementBuilder

A class used to build the SQL statements for DuckDB.

This class is the BaseStatementBuilder-derived implementation for DuckDB. It is automatically registered by DuckDbBootstrap — you do not need to instantiate it directly under normal use.

Constructors

public DuckDbStatementBuilder()
public DuckDbStatementBuilder(IDbSetting dbSetting,
    IResolver<Field, IDbSetting, string> convertFieldResolver = null,
    IResolver<Type, Type> averageableClientTypeResolver = null)

The parameterless constructor resolves dbSetting via DbSettingMapper.Get<DuckDBConnection>(), so it can only be used after DuckDbBootstrap has run.

DuckDB-specific generation notes

Paging uses LIMIT/LIMIT ... OFFSET for BatchQuery/skip-take queries and top-N Query.
Insert appends a RETURNING <key> AS "Result" clause onto the same statement, so the generated key comes back from the one round trip instead of a separate identity-lookup call.
InsertAll builds a single multi-row VALUES (...), (...), ... statement with one trailing RETURNING clause — DuckDB returns the generated keys in the same order the rows were listed, so the whole batch is inserted and its keys retrieved in a single round trip.
Merge/MergeAll are compiled as INSERT ... ON CONFLICT (<qualifiers>) DO UPDATE SET ... RETURNING ... (DuckDB’s native upsert).

MergeAll packs batchSize of those statements into a single command text (one round trip for the whole batch). DuckDB’s binder rejects any parameter reference inside a RETURNING projection outright, so the row’s ordinal position — normally passed as a bound __RepoDb_OrderColumn_{index} parameter — is instead embedded as an integer literal directly in each statement’s RETURNING clause; the unused parameter is simply left unbound.

Usability

Use StatementBuilderMapper to override it with a custom implementation.

StatementBuilderMapper.Add(typeof(DuckDBConnection), new MyCustomDuckDbStatementBuilder(dbSetting), true);