fasthtml websocket failure

[dennixin]

The data on the frontend, sent via WebSocket, is pako-compressed data. How does fasthtml ws handle the compressed data?

[78wesley]

The data on the frontend, sent via WebSocket, is pako-compressed data. How does fasthtml ws handle the compressed data?

I've ask the GPT within this repository. And it came with this answer.

Based on the code from the fasthtml repository, here's how the backend handles WebSocket (ws) data, including pako-compressed data from the frontend:

1. Receiving WebSocket Data:

   • The function _ws_endp is responsible for setting up the WebSocket endpoint. When data is received via WebSocket, it is passed as the data parameter to the _recv handler.

2. Deserialization:

   • Inside _ws_endp, the received data is processed with loads(data) (i.e., JSON deserialization) before being passed on:

     wd = _wrap_ws(ws, loads(data) if data else {}, _params(handler))

   • This means the server expects the ws payload to be JSON (or at least JSON-deserializable) after potential decompression.

3. Decompression Handling:

   • There is no explicit decompression code (e.g., zlib, gzip, or pako) in the backend Python source.
   • If the frontend sends pako-compressed data, the backend must receive it as a decompressed UTF-8 string for the JSON loader to work.
   • This strongly suggests that decompression is handled in the frontend (possibly in the JavaScript client), before the data is sent to the Python backend. The backend expects already-decompressed JSON strings.

4. Related Client Code:

   • The repository references the use of htmx-ext-ws (htmx websocket extension) and a custom fasthtml.js client script, which is likely where the pako decompression would happen in the browser before sending data to the backend.
   • The backend just processes the (decompressed) JSON as normal Python objects.

Summary:
The fasthtml backend does not handle pako decompression itself. It expects the frontend to decompress pako-compressed data before sending via WebSocket. The backend then receives the decompressed JSON string, deserializes it, and processes it as normal.

If you need to handle raw compressed data directly in the backend, you would have to add explicit decompression logic (e.g., using zlib for pako/deflate) before the loads(data) call. But as written, the backend expects plain JSON text.