Spice MCP

import polars as pl from spice_mcp.adapters.dune.extract import _process_raw_table, infer_type def test_process_raw_table_basic_types_and_nulls(): # header + mixed types + nulls + ragged line raw = ( "id,name,amount,ts\n" "1,Alice,10.5,2024-10-01 12:34:56.789 UTC\n" "2,<nil>,,2024-10-01 00:00:00.000 UTC\n" "3,Bob,7.0,2024-10-02 01:02:03.004 UTC,extra_col_that_is_ragged\n" ) df = _process_raw_table(raw, types=None, all_types=None) assert df.shape[0] == 3 assert df.columns == ["id", "name", "amount", "ts"] # type inference: id -> Int64, name -> String, amount -> Float64, ts -> Datetime dtypes = [str(t) for t in df.dtypes] assert "Int64" in dtypes[0] assert "String" in dtypes[1] assert "Float64" in dtypes[2] assert "Datetime" in dtypes[3] # value checks with null handling assert df["name"][1] is None assert df["amount"][1] is None # datetime parsed to polars Datetime assert isinstance(df["ts"].dtype, pl.Datetime) or df["ts"].dtype == pl.Datetime def test_process_raw_table_all_types_must_cover_all_columns(): raw = "a,b\n1,2\n" all_types = {"a": pl.Int64, "b": pl.Int64} df = _process_raw_table(raw, types=None, all_types=all_types) assert df.shape == (1, 2) assert [str(t) for t in df.dtypes] == [str(pl.Int64), str(pl.Int64)] def test_infer_type_roundtrip_via_csv(): s = pl.Series("x", ["1", "2", "3"]) # numeric as strings t = infer_type(s) # Should infer Int64 for numeric-only strings assert str(t) == str(pl.Int64()) or str(t) == str(pl.Int64)