Summary
Once a session has been populated via /load or /load_expr, there's no way to wipe it back to the empty state short of restarting the server process. The session's df / ldf / pl_df / dataflow / xorq_dataflow / cached state persists for the session's TTL, and any subsequent /load against the same session ID merges into the existing state rather than starting clean.
Scope
Server only.
Impact
- Debugging session-state bugs requires killing and restarting the server.
- Switching engines (pandas → xorq → polars) on the same session ID leaves stale state from the previous backend hanging around (e.g.
session.dataflow from pandas + session.xorq_dataflow from xorq both present, with WS dispatch depending on session.backend to disambiguate).
- For an interactive demo / playground, the ability to "start over" is table stakes UX.
Summary
Once a session has been populated via
/loador/load_expr, there's no way to wipe it back to the empty state short of restarting the server process. The session'sdf/ldf/pl_df/dataflow/xorq_dataflow/ cached state persists for the session's TTL, and any subsequent/loadagainst the same session ID merges into the existing state rather than starting clean.Scope
Server only.
Impact
session.dataflowfrom pandas +session.xorq_dataflowfrom xorq both present, with WS dispatch depending onsession.backendto disambiguate).