Translate invalid content types in HCA files (#7193, #7629)

Author: nadove-ucsc

src/azul/plugins/metadata/hca/__init__.py

@@ -545,6 +545,7 @@ def from_metadata(cls,
         else:
             assert ';' not in content_type, R(
                 'Unexpected MIME parameter in content type', content_type)
+        content_type = cls._content_type_corrections.get(content_type, content_type)
         return cls(uuid=json_str(descriptor['file_id']),
                    name=json_str(json_mapping(metadata['file_core'])['file_name']),
                    version=json_str(descriptor['file_version']),
@@ -566,3 +567,15 @@ def to_manifest_entry(self) -> JSON:
         entry['content-type'] = entry.pop('content_type')
         entry['indexed'] = False
         return entry
+
+    _content_type_corrections = {
+        '.gz': 'application/gzip',
+        'csv': 'text/csv',
+        'fastq.gz': 'application/gzip',
+        'h5': 'application/octet-stream',
+        'h5ad': 'application/octet-stream',
+        'txt': 'text/plain',
+        # Only correct for files created with Excel 2007 or later. Hopefully we
+        # don't have any files older than that.
+        'xlsx': 'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet',
+    }