All changes

src/pydiverse/transform/core/table_impl.py

@@ -181,6 +181,9 @@ def alias(self, name=None) -> AbstractTableImpl:
     def collect(self):
         ...
 
+    def collect_scalar(self):
+        ...
+
     def build_query(self):
         ...
 

src/pydiverse/transform/core/verbs.py

@@ -116,6 +116,12 @@ def collect(tbl: AbstractTableImpl):
     return tbl.collect()
 
 
+@builtin_verb()
+def collect_scalar(tbl: AbstractTableImpl):
+    validate_table_args(tbl)
+    return tbl.collect_scalar()
+
+
 @builtin_verb()
 def build_query(tbl: AbstractTableImpl):
     return tbl.build_query()

src/pydiverse/transform/eager/pandas_table.py

@@ -44,8 +44,11 @@ class PandasTableImpl(EagerTableImpl):
             computed at some point. This allows for a more lazy style API.
     """
 
-    def __init__(self, name: str, df: pd.DataFrame):
-        self.df = fast_pd_convert_dtypes(df)
+    def __init__(self, name: str, df: pd.DataFrame, convert_dtypes_to_ensure_semantics=True):
+        if convert_dtypes_to_ensure_semantics:
+            self.df = fast_pd_convert_dtypes(df)
+        else:
+            self.df = df
         self.join_translator = self.JoinTranslator()
 
         columns = {
@@ -133,6 +136,11 @@ def collect(self) -> pd.DataFrame:
         result.attrs["name"] = self.name
         return result
 
+    def collect_scalar(self) -> int | float | str | bool:  # TODO-in-this-pr: maybe-more?
+        collect = self.collect()
+        assert collect.shape == (1, 1)  # TODO in this PR: assert with if
+        return self.collect().iloc[0, 0]
+
     def mutate(self, **kwargs):
         uuid_kwargs = {self.named_cols.fwd[k]: (k, v) for k, v in kwargs.items()}
         self.df_name_mapping.update(