Report unsupported SPSS dictionary fidelity gaps

Author: TonisOrmisson

src/openstatspec/spss/sav.py

@@ -73,8 +73,9 @@ def import_sav_dataset(*, source: str | Path, database_url: str, dataset_id: str
         variables=variables, file_label=getattr(meta, "file_label", "") or "",
         source_encoding=getattr(meta, "file_encoding", None),
         documents=json.dumps(list(getattr(meta, "notes", []) or [])),
+        multiple_response_sets=json.dumps(dict(getattr(meta, "mr_sets", {}) or {}), default=str),
     )
-    return {**result, "loss_report": LossReport().events}
+    return {**result, "loss_report": _import_loss_report(meta)}
 
 
 def export_sav_dataset(*, database_url: str, dataset_id: str, destination: str | Path) -> dict[str, Any]:
@@ -107,10 +108,26 @@ def export_sav_dataset(*, database_url: str, dataset_id: str, destination: str |
     )
     return {
         "dataset_id": dataset_id, "destination": str(destination_path),
-        "loss_report": LossReport().events,
+        "loss_report": _export_loss_report(dataset, variables),
     }
 
 
+
+
+def _import_loss_report(meta: Any) -> tuple[dict[str, str], ...]:
+    if getattr(meta, "mr_sets", {}) or {}:
+        return ({"code": "multiple-response-sets-not-exportable", "detail": "Multiple-response sets are catalogued but pyreadstat cannot write them back to SAV."},)
+    return LossReport().events
+
+
+def _export_loss_report(dataset: dict[str, Any], variables: list[dict[str, Any]]) -> tuple[dict[str, str], ...]:
+    events = []
+    if dataset["multiple_response_sets"] != "{}":
+        events.append({"code": "multiple-response-sets-not-exported", "detail": "The SAV writer has no multiple-response-set output capability."})
+    if any(item.get("alignment") not in (None, "unknown") for item in variables):
+        events.append({"code": "variable-alignment-not-exported", "detail": "The SAV writer has no variable-alignment output capability."})
+    return tuple(events)
+
 def _require_source(source_path: Path) -> None:
     if source_path.suffix.lower() not in {".sav", ".zsav"}:
         raise UnsupportedOperationError("Only unencrypted SAV and ZSAV sources are supported.")

src/openstatspec/sql/wide.py

@@ -21,6 +21,7 @@ def catalog(metadata: MetaData) -> tuple[Table, Table]:
         Column("case_count", BigInteger, nullable=False),
         Column("file_label", Text, nullable=False, default=""),
         Column("documents", Text, nullable=False, default="[]"),
+        Column("multiple_response_sets", Text, nullable=False, default="{}"),
     )
     variables = Table(
         "variable_catalog", metadata,
@@ -61,6 +62,7 @@ def create_wide_dataset(
     *, database_url: str, dataset_id: str, source_name: str, source_format: str,
     rows: Iterable[Mapping[str, Any]], variables: list[dict[str, Any]], file_label: str = "",
     source_encoding: str | None = None, documents: str = "[]",
+    multiple_response_sets: str = "{}",
 ) -> dict[str, Any]:
     profile = profile_for_url(database_url)
     preflight(profile, len(variables))
@@ -83,6 +85,7 @@ def create_wide_dataset(
             dataset_id=dataset_id, data_table=data_table.name, source_format=source_format,
             source_name=source_name, source_encoding=source_encoding, case_count=len(materialized),
             file_label=file_label, documents=documents,
+            multiple_response_sets=multiple_response_sets,
         ))
         connection.execute(insert(variable_catalog), [dict(dataset_id=dataset_id, **item) for item in variables])
         if materialized:

tests/test_loss_reports.py

@@ -0,0 +1,27 @@
+from openstatspec import export_sav
+from openstatspec.sql.wide import create_wide_dataset
+
+
+def test_export_reports_unwritable_spss_dictionary_features(tmp_path) -> None:
+    database = f"sqlite:///{tmp_path / 'dataset.sqlite'}"
+    create_wide_dataset(
+        database_url=database,
+        dataset_id="dictionary-gap",
+        source_name="fixture.sav",
+        source_format="SAV",
+        rows=[{"answer": 1.0}],
+        variables=[{
+            "ordinal": 1, "source_name": "answer", "physical_name": "answer",
+            "storage_kind": "numeric", "string_width": None, "label": "",
+            "format": "F8.0", "measure": "nominal", "alignment": "left",
+            "display_width": 8, "value_labels": "{}", "missing_ranges": "[]",
+        }],
+        multiple_response_sets='{"set1": {"label": "example"}}',
+    )
+
+    result = export_sav(
+        database_url=database, dataset_id="dictionary-gap", destination=tmp_path / "output.sav"
+    )
+    assert {event["code"] for event in result["loss_report"]} == {
+        "multiple-response-sets-not-exported", "variable-alignment-not-exported"
+    }