Update.

core/state.py

@@ -20,6 +20,7 @@ from core.constants import OAUTH_CLIENT_SECRET
 from core.constants import PAST_PROJECTS_PATH
 from core.constants import PROJECT_FOLDER_PATTERN
 from core.constants import REDIRECT_URI
+from core.names import find_unique_name
 import mlcroissant as mlc
 
 
@@ -257,6 +258,8 @@ class Metadata:
         del self.distribution[key]
 
     def add_record_set(self, record_set: RecordSet) -> None:
+        name = find_unique_name(self.names(), record_set.name)
+        record_set.name = name
         self.record_sets.append(record_set)
 
     def remove_record_set(self, key: int) -> None:
@@ -323,3 +326,7 @@ class Metadata:
             distribution=distribution,
             record_sets=record_sets,
         )
+
+    def names(self) -> set[str]:
+        nodes = self.distribution + self.record_sets
+        return set([node.name for node in nodes])

cypress.config.js

@@ -3,5 +3,6 @@ const { defineConfig } = require("cypress");
 module.exports = defineConfig({
   // To access content within Streamlit iframes for custom components:
   chromeWebSecurity: false,
+  defaultCommandTimeout: 10000,
   e2e: {},
 });

cypress/e2e/createManually.cy.js

@@ -8,7 +8,7 @@ describe('Create a resource manually', () => {
   it('should allow adding a FileObject resource', () => {
     // Streamlit starts on :8501.
     cy.visit('http://localhost:8501')
-    cy.get('button', {timeout: 10000}).contains('Create', {timeout: 10000}).click()
+    cy.get('button').contains('Create').click()
     cy.get('input[aria-label="Name:red[*]"]').type('MyDataset').blur()
     cy.get('[data-testid="stMarkdownContainer"]')
     .contains('Metadata')

cypress/e2e/uploadCsv.cy.js

@@ -8,7 +8,7 @@ describe('Editor loads a local CSV as a resource', () => {
   it('should display the form: Overview, Metadata, Resources, & Record Sets', () => {
     // Streamlit starts on :8501.
     cy.visit('http://localhost:8501')
-    cy.get('button', {timeout: 10000}).contains('Create', {timeout: 10000}).click()
+    cy.get('button').contains('Create').click()
 
     cy.get('input[aria-label="Name:red[*]"]').type('MyDataset').blur()
     cy.get('[data-testid="stMarkdownContainer"]')

views/files.py

@@ -113,11 +113,8 @@ def _render_upload_panel():
             url = st.session_state[_DISTANT_URL_KEY]
             uploaded_file = st.session_state[_LOCAL_FILE_KEY]
             file_type = FILE_TYPES[file_type_name]
-            nodes = (
-                st.session_state[Metadata].distribution
-                + st.session_state[Metadata].record_sets
-            )
-            names = set([node.name for node in nodes])
+            metadata: Metadata = st.session_state[Metadata]
+            names = metadata.names()
             if url:
                 file = file_from_url(file_type, url, names)
             elif uploaded_file:

views/load.py

@@ -3,6 +3,7 @@ import os
 from etils import epath
 import streamlit as st
 
+from core.constants import EDITOR_CACHE
 from core.past_projects import save_current_project
 from core.state import CurrentStep
 from core.state import Metadata
@@ -10,25 +11,26 @@ import mlcroissant as mlc
 from utils import jump_to
 
 
+def _on_file_upload(key):
+    file = st.session_state[key]
+    file_cont = file.read()
+    # TODO(marcenacp): The Python library should support loading from an open file/dict.
+    newfile_name = EDITOR_CACHE / "loaded_croissant"
+    os.makedirs(os.path.dirname(newfile_name), exist_ok=True)
+    with open(newfile_name, mode="wb+") as outfile:
+        outfile.write(file_cont)
+    try:
+        dataset = mlc.Dataset(newfile_name)
+        st.session_state[Metadata] = Metadata.from_canonical(dataset.metadata)
+        jump_to(CurrentStep.editor)
+        save_current_project()
+    except mlc.ValidationError as e:
+        st.warning(e)
+        st.toast(body="Invalid Croissant File!", icon="🔥")
+
+
 def render_load():
-    file = st.file_uploader("Select a JSON-LD", type="json")
-    if file is not None:
-        try:
-            file_cont = file.read()
-            newfile_name = (
-                epath.Path("~").expanduser()
-                / ".cache"
-                / "croissant"
-                / "loaded_croissant"
-            )
-            os.makedirs(os.path.dirname(newfile_name), exist_ok=True)
-            with open(newfile_name, mode="wb+") as outfile:
-                outfile.write(file_cont)
-            dataset = mlc.Dataset(newfile_name)
-            st.session_state[Metadata] = Metadata.from_canonical(dataset.metadata)
-            jump_to(CurrentStep.editor)
-            save_current_project()
-            st.rerun()
-        except mlc.ValidationError as e:
-            st.warning(e)
-            st.toast(body="Invalid Croissant File!", icon="🔥")
+    key = "json-ld-file-upload"
+    st.file_uploader(
+        "Select a JSON-LD", type="json", key=key, on_change=_on_file_upload, args=(key,)
+    )

views/record_sets.py

@@ -83,6 +83,11 @@ def _find_joins(fields: list[Field]) -> set[Join]:
     return joins
 
 
+def _handle_create_record_set():
+    metadata: Metadata = st.session_state[Metadata]
+    metadata.add_record_set(RecordSet(name="new-record-set", description=""))
+
+
 def _handle_fields_change(record_set_key: int, record_set: RecordSet):
     data_editor_key = _data_editor_key(record_set_key, record_set)
     result = st.session_state[data_editor_key]
@@ -148,7 +153,7 @@ def _render_left_panel():
     record_sets = st.session_state[Metadata].record_sets
     record_set: RecordSet
     for record_set_key, record_set in enumerate(record_sets):
-        title = f"**{record_set.name}** ({len(record_set.fields)} fields)"
+        title = f"**{record_set.name or '-'}** ({len(record_set.fields)} fields)"
         prefix = f"record-set-{record_set_key}"
         with st.expander(title, expanded=False):
             col1, col2 = st.columns([1, 3])
@@ -238,7 +243,8 @@ def _render_left_panel():
             )
             st.data_editor(
                 fields,
-                use_container_width=True,
+                # There is a bug with `st.data_editor` when the df is empty.
+                use_container_width=not fields.empty,
                 num_rows="dynamic",
                 key=data_editor_key,
                 column_config={
@@ -269,6 +275,12 @@ def _render_left_panel():
                 on_click=_handle_on_click_field,
                 args=(record_set_key, record_set),
             )
+    st.button(
+        "Create a new RecordSet",
+        key=f"create-new-record-set",
+        type="primary",
+        on_click=_handle_create_record_set,
+    )
 
 
 def _render_right_panel():

views/splash.py

@@ -40,8 +40,8 @@ def render_splash():
             )
         with st.expander("**Try out an example!**", expanded=True):
 
-            def create_example():
-                url = "https://raw.githubusercontent.com/mlcommons/croissant/main/datasets/titanic/metadata.json"
+            def create_example(dataset: str):
+                url = f"https://raw.githubusercontent.com/mlcommons/croissant/main/datasets/{dataset.lower()}/metadata.json"
                 try:
                     json = requests.get(url).json()
                     metadata = mlc.Metadata.from_json(mlc.Issues(), json, None)
@@ -50,16 +50,29 @@ def render_splash():
                     jump_to(CurrentStep.editor)
                 except Exception as exception:
                     logging.error(exception)
-                    st.write(
+                    st.error(
                         "Sorry, it seems that the example is broken... Can you please"
                         " [open an issue on"
                         " GitHub](https://github.com/mlcommons/croissant/issues/new)?"
                     )
 
+            dataset = st.selectbox(
+                label="Dataset",
+                options=[
+                    "Titanic",
+                    "FLORES-200",
+                    "GPT-3",
+                    "COCO2014",
+                    "PASS",
+                    "MovieLens",
+                    "Bigcode-The-Stack",
+                ],
+            )
             st.button(
-                "Titanic dataset",
+                f"{dataset} dataset",
                 on_click=create_example,
                 type="primary",
+                args=(dataset,),
             )
     with col2:
         with st.expander("**Past projects**", expanded=True):