wboayue / rust-ibapi

92%

push

github

Add collect_for/collect_until terminals and snapshot_once builder (#686) (#692)

* Add collect_for/collect_until terminals and snapshot_once builder (#686)

Offer a convenience terminal on Subscription<T> that collects data items
into a Vec bounded by a total timeout, an optional early-break predicate,
the snapshot-end sentinel, stream end, or a terminal error. Adds
MarketDataBuilder::snapshot_once(timeout) sugar so one-shot market-data
snapshots no longer require a hand-written collect-with-timeout loop.

- Subscription::collect_for / collect_until (sync + async)
- MarketDataBuilder::snapshot_once (sync + async)
- TickSnapshotEnd testdata builder; collect + snapshot_once tests
- CHANGELOG + README snapshot idiom

* Skip redundant cancel for completed async snapshot subscriptions (#686)

Review finding #2: the async Subscription<T> lacked the snapshot_ended
tracking the sync side has, so dropping an async snapshot subscription
after SnapshotEnd always spawned a cancelMktData for an already-completed
request. Port snapshot_ended to the async subscription:

- Add snapshot_ended (shared) + snapshot_end_fn fields; detect the
  snapshot-end sentinel in poll_next and skip cancel in cancel()/drop().
- Capture the detector via detect_snapshot_end() at the builder-send layer
  where T: StreamDecoder<T> holds, leaving new_from_internal usable for
  decoder types where D != T (test path).
- Regression test asserting no cancel lands after a completed snapshot.

* simplify: intrinsic async snapshot detection + fn-pointer + macro fixture (#686)

/simplify pass on the snapshot work:

- Centralize snapshot-end detection in new_from_internal (require
  T: StreamDecoder<T>, the self-decoding pattern every production decoder
  already satisfies) instead of wiring detect_snapshot_end() into three
  builder send sites. Removes the fragility where a future creation path
  would silently lose cancel suppression; mirrors the sync side's intrinsic
  tracking. (altitude)
- sna... (continued)

85 of 87 new or added lines in 5 files covered. (97.7%)

18196 of 19821 relevant lines covered (91.8%)

28.79 hits per line