aggregate's lambda function can take three or four arguments, depending on whether a finish function is provided
We don't infer types on nested lambda functions correctly, so for the time being we just introduce a workaround with CAST(... AS ...) around the nested lambda, allowing the type to be manually specified by the user
Summary
This PR adds the following Spark SQL functions:
transform_keys
transform_values
map_zip_with
It also fixes a few bugs:
concat
on maps is possibleaggregate
's lambda function can take three or four arguments, depending on whether afinish
function is providedCAST(... AS ...)
around the nested lambda, allowing the type to be manually specified by the userTest Plan
make check
passesmake test
shows 100% unit test coverageDeployment Plan