Display RecordSets as a flat non-editable list.

editor/core/data_types.py

@@ -2,16 +2,18 @@
 
 import numpy as np
 
+import mlcroissant as mlc
+
 
 def convert_dtype(dtype: Any):
     """Converts from NumPy/Pandas to Croissant data types."""
-    if dtype == np.int64:
-        return "https://schema.org/Integer"
-    elif dtype == np.float64:
-        return "https://schema.org/Float"
+    if dtype == np.int64 or dtype == np.int32:
+        return mlc.DataType.INTEGER
+    elif dtype == np.float64 or dtype == np.float32:
+        return mlc.DataType.FLOAT
     elif dtype == np.bool_:
-        return "https://schema.org/Boolean"
+        return mlc.DataType.BOOL
     elif dtype == np.str_ or dtype == object:
-        return "https://schema.org/Text"
+        return mlc.DataType.TEXT
     else:
         raise NotImplementedError(dtype)

editor/core/record_sets.py

@@ -24,7 +24,7 @@ def infer_record_sets(file: FileObject | FileSet, names: set[str]) -> list[Recor
         )
         field = Field(
             name=column,
-            data_types=convert_dtype(value),
+            data_types=[convert_dtype(value)],
             source=source,
             references=mlc.Source(),
         )

editor/cypress/e2e/uploadCsv.cy.js

@@ -45,5 +45,10 @@ describe('Editor loads a local CSV as a resource', () => {
     // On the record set page, we see the record set.
     cy.get('[data-testid="stMarkdownContainer"]').contains('Record sets').click()
     cy.contains('base.csv_record_set')
+    // We also see the fields with the proper types.
+    cy.get('[data-testid="stDataFrameResizable"]').contains("column1")
+    cy.get('[data-testid="stDataFrameResizable"]').contains("https://schema.org/Text")
+    cy.get('[data-testid="stDataFrameResizable"]').contains("column2")
+    cy.get('[data-testid="stDataFrameResizable"]').contains("https://schema.org/Integer")
   })
 })

editor/views/record_sets.py

@@ -2,38 +2,80 @@
 import streamlit as st
 
 from core.state import Metadata
+from core.state import RecordSet
+import mlcroissant as mlc
 from utils import DF_HEIGHT
+from utils import needed_field
 
 DATA_TYPES = [
-    "https://schema.org/Text",
-    "https://schema.org/Float",
-    "https://schema.org/Integer",
-    "https://schema.org/Boolean",
+    mlc.DataType.TEXT,
+    mlc.DataType.FLOAT,
+    mlc.DataType.INTEGER,
+    mlc.DataType.BOOL,
 ]
 
 
+class FieldDataFrame:
+    """Names of the columns in the pd.DataFrame for `fields`."""
+
+    NAME = "Name"
+    DESCRIPTION = "Description"
+    DATA_TYPE = "Data type"
+
+
 def render_record_sets():
-    if len(st.session_state[Metadata].record_sets) == 0:
-        st.markdown("Please add files first.")
-    else:
-        for record_set in st.session_state[Metadata].record_sets:
-            record_set_conv = pd.DataFrame(record_set.__dict__)
-            with st.container():
-                st.data_editor(
-                    record_set_conv,
-                    height=DF_HEIGHT,
-                    use_container_width=True,
-                    key=f"record_set_{record_set.name}",
-                    column_config={
-                        "name": st.column_config.TextColumn(
-                            "name",
-                            help="Name of the field",
-                            required=True,
-                        ),
-                        "description": st.column_config.TextColumn(
-                            "description",
-                            help="Description of the field",
-                            required=False,
-                        ),
-                    },
-                )
+    if not st.session_state[Metadata].distribution:
+        st.markdown("Please add resources first.")
+        return
+    record_set: RecordSet
+    for record_set in st.session_state[Metadata].record_sets:
+        with st.expander(f"**{record_set.name}**", expanded=True):
+            col1, col2 = st.columns([1, 3])
+            col1.text_input(
+                needed_field("Name"),
+                placeholder="Name without special character.",
+                key=f"{record_set.name}-name",
+                value=record_set.name,
+            )
+            col2.text_input(
+                "Description",
+                placeholder="Provide a clear description of the RecordSet.",
+                key=f"{record_set.name}-description",
+                value=record_set.description,
+            )
+            st.checkbox(
+                "Whether the RecordSet is an enumeration",
+                key=f"{record_set.name}-is-enumeration",
+                value=record_set.is_enumeration,
+            )
+            names = [field.name for field in record_set.fields]
+            descriptions = [field.description for field in record_set.fields]
+            data_types = [field.data_types[0] for field in record_set.fields]
+            fields = pd.DataFrame({
+                FieldDataFrame.NAME: names,
+                FieldDataFrame.DESCRIPTION: descriptions,
+                FieldDataFrame.DATA_TYPE: data_types,
+            })
+            st.data_editor(
+                fields,
+                height=DF_HEIGHT,
+                use_container_width=True,
+                column_config={
+                    FieldDataFrame.NAME: st.column_config.TextColumn(
+                        FieldDataFrame.NAME,
+                        help="Name of the field",
+                        required=True,
+                    ),
+                    FieldDataFrame.DESCRIPTION: st.column_config.TextColumn(
+                        FieldDataFrame.DESCRIPTION,
+                        help="Description of the field",
+                        required=False,
+                    ),
+                    FieldDataFrame.DATA_TYPE: st.column_config.SelectboxColumn(
+                        FieldDataFrame.DATA_TYPE,
+                        help="The Croissant type",
+                        options=DATA_TYPES,
+                        required=True,
+                    ),
+                },
+            )