sdf-labs / sql-functions

8 stars 1 forks source link

Regexp UDFs and beginnings of generic vectorization #70

Closed vgapeyev closed 3 months ago

vgapeyev commented 3 months ago

This brings in all the remaining regexp-involving Trino UDFs (except the overload of regexp_replace that has a lambda argument).

They are implemented with the help of a nascent framework for implementing UDFs by lifting / mapping relatively simple row functions to vectorized operations on columns. The framework is in its very beginning and still relies on excessive cut-paste-modify activity, but already provides some effort savings. Improvement advice or discussions from those more experienced with Rust and Arrow will be gladly appreciated and incorporated!