Add support for `count_distinct` aggregate function

[rohitrastogi]

Add support for count_distinct aggregate function

Fenic currently supports the count() aggregate function and the drop_duplicates() DataFrame method, but does not yet support the count_distinct() aggregate function.

This issue is about adding support for count_distinct() to Fenic. Fenic represents expressions in a logical expression tree, which it then transpiles into Polars expressions for execution.

---

🛠️ What needs to be done

1. Add a new expression class

   • Create a CountDistinctExpr class that extends AggregateExpr (which in turn extends LogicalExpr).
   • Reference other aggregate expression definitions here:
     aggregate.py

2. Transpile the expression to Polars

   • Extend the expression transpiler switch to convert CountDistinctExpr to a Polars expression.
   • You'll find the switch here:
     expr_converter.py

3. Expose a user-facing API

   • Add a new count_distinct function to Fenic’s public API with a clean docstring and appropriate typing.
   • Follow the style and conventions in this file:
     builtin.py

4. Write unit tests

   • Add tests for count_distinct under group-by aggregations to:
     test_aggregations.py

---

Feel free to ask questions or open a draft PR early if you're unsure about anything. We're happy to help!

[tusharchou]

Hey how can I help?

[rohitrastogi]

Hi @tusharchou, thanks for your interest!

Here's a sample PR for a recent column function we added to Fenic: https://github.com/typedef-ai/fenic/pull/103/files. It demonstrates how we create a custom Polars plugin for execution.

Implementing the count_distinct aggregate function will follow a similar structure. However, instead of building a custom Polars plugin, you can simply wrap Polars' built-in n_unique expression in the expression transpiler.

Let me know if you have any questions!