Deploy (see actual commits on https://github.com/mlcommons/croissant).

core/state.py

@@ -33,6 +33,9 @@ def create_class(mlc_class: type, instance: Any, **kwargs) -> Any:
         name = field.name
         if hasattr(instance, name) and name not in kwargs:
             params[name] = getattr(instance, name)
+    if "uuid" in params and params.get("uuid") is None:
+        # Let mlcroissant handle the default value
+        del params["uuid"]
     return mlc_class(**params, **kwargs)
 
 
@@ -137,6 +140,7 @@ class FileObject:
     sha256: str | None = None
     df: pd.DataFrame | None = None
     folder: epath.PathLike | None = None
+    uuid: str | None = None
 
 
 @dataclasses.dataclass
@@ -149,6 +153,7 @@ class FileSet:
     encoding_format: str | None = ""
     includes: str | None = ""
     name: str = ""
+    uuid: str | None = None
 
 
 @dataclasses.dataclass
@@ -161,6 +166,7 @@ class Field:
     data_types: str | list[str] | None = None
     source: mlc.Source | None = None
     references: mlc.Source | None = None
+    uuid: str | None = None
 
 
 @dataclasses.dataclass
@@ -174,6 +180,7 @@ class RecordSet:
     is_enumeration: bool | None = None
     key: str | list[str] | None = None
     fields: list[Field] = dataclasses.field(default_factory=list)
+    uuid: str | None = None
 
 
 @dataclasses.dataclass
@@ -191,6 +198,7 @@ class Metadata:
     date_published: datetime.datetime | None = None
     license: str | None = ""
     personal_sensitive_information: str | None = None
+    uuid: str | None = None
     url: str = ""
     distribution: list[FileObject | FileSet] = dataclasses.field(default_factory=list)
     record_sets: list[RecordSet] = dataclasses.field(default_factory=list)

requirements.txt

@@ -1,5 +1,5 @@
 etils[epath]
-mlcroissant
+mlcroissant>=1.0.1
 numpy
 pandas
 pytest

views/overview.py

@@ -10,7 +10,7 @@ from utils import needed_field
 from views.metadata import handle_metadata_change
 from views.metadata import MetadataEvent
 
-_NON_RELEVANT_METADATA = ["ctx", "name", "distribution", "record_sets"]
+_NON_RELEVANT_METADATA = ["ctx", "name", "distribution", "record_sets", "uuid"]
 
 _INFO_TEXT = """Croissant files are composed of three layers:
 
@@ -38,8 +38,9 @@ def _relevant_fields(class_or_instance: type):
     else:
         return [
             field
-            for field, value in dataclasses.asdict(class_or_instance).items()
-            if value and field not in _NON_RELEVANT_METADATA
+            for field in dataclasses.fields(Metadata)
+            if hasattr(class_or_instance, field.name)
+            and field.name not in _NON_RELEVANT_METADATA
         ]