Make runner onData() return a fresh runner

After the run-builder collapse, calling onData() twice on the same
VerificationSuite/AnalysisRunner/ColumnProfilerRunner/ConstraintSuggestionRunner
mutated state in place, so reusing a suite across tables leaked checks
and engine bindings between runs. Restore the pre-refactor semantics by
returning a fresh runner from onData() with the engine rebound.

Author: chenliu0831

pydeequ/v2/profiles.py

@@ -74,6 +74,9 @@ class ColumnProfilerRunner:
     """
     Run column profiling.
 
+    ``onData()`` returns a fresh runner bound to the chosen data; the
+    original instance is left untouched.
+
     Example:
         profiles = (ColumnProfilerRunner(engine)
             .onData(table="users")
@@ -92,16 +95,16 @@ def onData(
         table: Optional[str] = None,
         dataframe: "Optional[DataFrame]" = None,
     ) -> "ColumnProfilerRunner":
-        """Bind data for profiling (keyword-only)."""
+        """Return a fresh runner bound to the given data (keyword-only)."""
         if table is not None and dataframe is not None:
             raise ValueError("Provide either 'table' or 'dataframe', not both")
         if table is not None:
-            self._engine = self._engine.for_table(table)
+            bound = self._engine.for_table(table)
         elif dataframe is not None:
-            self._engine = self._engine.for_dataframe(dataframe)
+            bound = self._engine.for_dataframe(dataframe)
         else:
             raise ValueError("Must provide either 'table' or 'dataframe'")
-        return self
+        return ColumnProfilerRunner(bound)
 
     def restrictToColumns(self, columns: Sequence[str]) -> "ColumnProfilerRunner":
         self._restrict_to_columns = columns

pydeequ/v2/suggestions.py

@@ -70,6 +70,9 @@ class ConstraintSuggestionRunner:
     """
     Generate constraint suggestions.
 
+    ``onData()`` returns a fresh runner bound to the chosen data; the
+    original instance is left untouched.
+
     Example:
         suggestions = (ConstraintSuggestionRunner(engine)
             .onData(table="users")
@@ -88,16 +91,16 @@ def onData(
         table: Optional[str] = None,
         dataframe: "Optional[DataFrame]" = None,
     ) -> "ConstraintSuggestionRunner":
-        """Bind data for constraint suggestion (keyword-only)."""
+        """Return a fresh runner bound to the given data (keyword-only)."""
         if table is not None and dataframe is not None:
             raise ValueError("Provide either 'table' or 'dataframe', not both")
         if table is not None:
-            self._engine = self._engine.for_table(table)
+            bound = self._engine.for_table(table)
         elif dataframe is not None:
-            self._engine = self._engine.for_dataframe(dataframe)
+            bound = self._engine.for_dataframe(dataframe)
         else:
             raise ValueError("Must provide either 'table' or 'dataframe'")
-        return self
+        return ConstraintSuggestionRunner(bound)
 
     def addConstraintRules(self, rules: Rules) -> "ConstraintSuggestionRunner":
         self._rules.append(rules)

pydeequ/v2/verification.py

@@ -83,6 +83,11 @@ class VerificationSuite:
     """
     Run data-quality verification.
 
+    ``onData()`` returns a fresh suite bound to the chosen data; the
+    original instance is left untouched. This means a single suite
+    object can be reused across multiple tables or dataframes without
+    state bleeding between runs.
+
     Example:
         result = (VerificationSuite(engine)
             .onData(table="users")
@@ -100,9 +105,10 @@ def onData(
         table: Optional[str] = None,
         dataframe: "Optional[DataFrame]" = None,
     ) -> "VerificationSuite":
-        """Bind data for verification (keyword-only)."""
-        self._engine = _bind_engine(self._engine, table=table, dataframe=dataframe)
-        return self
+        """Return a fresh suite bound to the given data (keyword-only)."""
+        return VerificationSuite(
+            _bind_engine(self._engine, table=table, dataframe=dataframe)
+        )
 
     def addCheck(self, check: Check) -> "VerificationSuite":
         self._checks.append(check)
@@ -118,6 +124,9 @@ class AnalysisRunner:
     """
     Run analyzers without checks.
 
+    ``onData()`` returns a fresh runner bound to the chosen data; the
+    original instance is left untouched.
+
     Example:
         result = (AnalysisRunner(engine)
             .onData(table="users")
@@ -136,9 +145,10 @@ def onData(
         table: Optional[str] = None,
         dataframe: "Optional[DataFrame]" = None,
     ) -> "AnalysisRunner":
-        """Bind data for analysis (keyword-only)."""
-        self._engine = _bind_engine(self._engine, table=table, dataframe=dataframe)
-        return self
+        """Return a fresh runner bound to the given data (keyword-only)."""
+        return AnalysisRunner(
+            _bind_engine(self._engine, table=table, dataframe=dataframe)
+        )
 
     def addAnalyzer(self, analyzer: _ConnectAnalyzer) -> "AnalysisRunner":
         self._analyzers.append(analyzer)