[BUG]App hangs when filtering by large custom metadata fields

[mnk-nasir]

When attempting to filter a dataset in the FiftyOne App using a high-cardinality custom metadata field, the UI becomes unresponsive and eventually throws a "Script Snippet" timeout error in the browser.

Steps to Reproduce
Load a dataset with >100k samples: dataset = fo.load_dataset("my_large_dataset")

Open the App: session = fo.launch_app(dataset)

In the sidebar, navigate to metadata -> custom_field.

Select a specific value to filter.

The App freezes for 10+ seconds.

Expected Behavior
The App should apply the view stage and refresh the grid within a reasonable timeframe (under 2 seconds) or show a loading state without locking the browser thread.

Environment Information
FiftyOne Version: [e.g., 0.23.8]

OS: [e.g., macOS Sonoma / Ubuntu 22.04]

Python Version: [e.g., 3.10.12]

Browser: [e.g., Chrome 122]

Code Snippet / Logs
Python
import fiftyone as fo

Replace with your specific logic

dataset = fo.Dataset("test-issue")

... logic to add samples ...

session = fo.launch_app(dataset)

[AdonaiVera]

Hi @mnk-nasir, thanks for the report!

Actually, FiftyOne has a optimizing-query-performance feature that speeds up sidebar filtering on large datasets by leveraging database indexes. You can add an index on the field you're filtering by:

import fiftyone as fo

dataset = fo.load_dataset("my_large_dataset")

# See existing indexes
print(dataset.list_indexes())

# Index the field you filter on
dataset.create_index("metadata.custom_field")

session = fo.launch_app(dataset)

Once the index is created, you’ll see a ⚡ icon next to the field in the sidebar, that means filters on that field will be fast. After indexing, filtering should be near-instant, even on large datasets (You can see this in action with bdd100k in the Docs)

Let me know if this helps 👍