Feedback

Author: fealho

sdgym/datasets.py

@@ -34,16 +34,18 @@ def _get_bucket_name(bucket):
     return bucket[len(S3_PREFIX) :] if bucket.startswith(S3_PREFIX) else bucket
 
 
-def _validate_dataset_availability(
+def _raise_dataset_not_found_error(
     s3_client,
     bucket_name,
     dataset_name,
     current_modality,
-    display_name,
     bucket,
     modality,
 ):
-    """Return modalities where the dataset exists in the bucket, excluding the current modality."""
+    display_name = dataset_name
+    if isinstance(dataset_name, Path):
+        display_name = dataset_name.name
+
     available_modalities = []
     for other_modality in MODALITIES:
         if other_modality == current_modality:
@@ -85,12 +87,8 @@ def _download_dataset(
 
     contents = _list_s3_bucket_contents(s3_client, bucket_name, prefix)
     if not contents:
-        display_name = dataset_name
-        if isinstance(dataset_name, Path):
-            display_name = dataset_name.name
-
-        _validate_dataset_availability(
-            s3_client, bucket_name, dataset_name, modality, display_name, bucket, modality
+        _raise_dataset_not_found_error(
+            s3_client, bucket_name, dataset_name, modality, bucket, modality
         )
 
     for obj in contents:

sdgym/result_explorer/result_explorer.py

@@ -26,18 +26,12 @@ def _validate_local_path(path):
 
 def _resolve_effective_path(path, modality):
     """Append the modality folder to the given base path if provided."""
-    if not modality:
-        return path
-
-    _validate_modality(modality)
-
     # Avoid double-appending if already included
     if str(path).rstrip('/').endswith(('/' + modality, modality)):
         return path
 
     if is_s3_path(path):
-        path = path.rstrip('/') + '/' + modality
-        return path
+        return path.rstrip('/') + '/' + modality
 
     return os.path.join(path, modality)
 
@@ -59,12 +53,15 @@ def _create_results_handler(self, original_path, effective_path):
         _validate_local_path(effective_path)
         return LocalResultsHandler(effective_path, baseline_synthesizer=baseline_synthesizer)
 
-    def __init__(self, path, modality, aws_access_key_id=None, aws_secret_access_key=None):
+    def __init__(
+        self, path, modality='single_table', aws_access_key_id=None, aws_secret_access_key=None
+    ):
         self.path = path
+        _validate_modality(modality)
         self.modality = modality.lower()
         self.aws_access_key_id = aws_access_key_id
         self.aws_secret_access_key = aws_secret_access_key
-        effective_path = _resolve_effective_path(path, modality)
+        effective_path = _resolve_effective_path(path, self.modality)
         self._handler = self._create_results_handler(path, effective_path)
 
     def list(self):

sdgym/result_explorer/result_handler.py

@@ -5,8 +5,6 @@
 import os
 from abc import ABC, abstractmethod
 from datetime import datetime
-from io import BytesIO
-from zipfile import ZipFile
 
 import cloudpickle
 import pandas as pd
@@ -27,7 +25,6 @@ class ResultsHandler(ABC):
     """Abstract base class for handling results storage and retrieval."""
 
     def __init__(self, baseline_synthesizer=SYNTHESIZER_BASELINE):
-        # Allow overrides per modality while maintaining the historical default.
         self.baseline_synthesizer = baseline_synthesizer or SYNTHESIZER_BASELINE
 
     @abstractmethod
@@ -395,15 +392,7 @@ def load_synthetic_data(self, file_path):
         response = self.s3_client.get_object(Bucket=self.bucket_name, Key=key)
         body = response['Body'].read()
         if file_path.endswith('.zip'):
-            tables = {}
-            with ZipFile(BytesIO(body)) as zf:
-                for name in zf.namelist():
-                    if name.endswith('.csv'):
-                        table_name = os.path.splitext(os.path.basename(name))[0]
-                        with zf.open(name) as csv_file:
-                            tables[table_name] = pd.read_csv(csv_file)
-
-            return tables
+            return _read_zipped_data(io.BytesIO(body), modality='multi_table')
 
         return pd.read_csv(io.BytesIO(body))
 

sdgym/run_benchmark/upload_benchmark_results.py

@@ -116,6 +116,7 @@ def upload_results(
     run_date = folder_infos['date']
     result_explorer = ResultsExplorer(
         OUTPUT_DESTINATION_AWS,
+        modality='single_table',
         aws_access_key_id=aws_access_key_id,
         aws_secret_access_key=aws_secret_access_key,
     )