Unblocking [RUST-INT][TPCH]: Add a call to .combine_chunks() when cre…

…ating a Table from arrow (#700)

* When we create Tables from PyArrow tables, we can only handle one
batch at the moment
* This causes some issues when running our TPC-H unit tests, ostensibly
because the parquet files define multiple Parquet row groups
* This PR adds a temporary fix to unblock running our new Rust code on
TPC-H by calling to `.combine_chunks()` before creating the table so
that we only have one batch

daft/table/table.py

@@ -70,6 +70,9 @@ def _from_pytable(pyt: _PyTable) -> Table:
     @staticmethod
     def from_arrow(arrow_table: pa.Table) -> Table:
         assert isinstance(arrow_table, pa.Table)
+        # TODO: [RUST-INT][TPCH] _PyTable.from_arrow_record_batches only supports single-batch inputs at the moment
+        # so we hack around it by combining the chunks first. We should fix this and remove arrow_table.combine_chunks() here.
+        arrow_table = arrow_table.combine_chunks()
         pyt = _PyTable.from_arrow_record_batches(arrow_table.to_batches())
         return Table._from_pytable(pyt)