notfilippo opened a new issue, #22563: URL: https://github.com/apache/datafusion/issues/22563
### Is your feature request related to a problem or challenge? DataFusion has a default runtime inventory of scalar, higher-order, aggregate, and window functions, but there is no generated Substrait extension YAML (https://github.com/substrait-io/substrait/blob/main/text/simple_extensions_schema.yaml) that describes that function set. Without generated extension YAML, Substrait function declarations can drift from the functions registered by a default DataFusion session, and downstream projects have no reusable DataFusion API for generating equivalent YAML for custom UDF registries. ### Describe the solution you'd like Add a generator that emits Substrait function YAML from the same default `SessionStateDefaults` inventory used by DataFusion's function docs generator. The generator should: * expose a public API that accepts an explicit function inventory and generator config * allow custom URNs, output paths, and overrides for signatures or return types that cannot be inferred from DataFusion `Signature` metadata * include a binary with stdout, `--check`, and `--write` modes * commit generated scalar, aggregate, and window YAML files under `datafusion/substrait/extensions/` * fail if a default function cannot be emitted or explicitly overridden ### Describe alternatives you've considered The main alternative is to maintain the Substrait YAML by hand. That would be more error-prone because the source of truth for available DataFusion functions is the runtime registry, not static YAML files. ### Additional context _No response_ -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
