📝 document queued sink methods

2026-07-29 22:16:34 +00:00 · 2026-06-13 22:42:41 +08:00
parent 86b99fe004
commit 0df565d454
5 changed files with 304 additions and 0 deletions
@@ -0,0 +1,76 @@
+---
+name: queued-sink-dropped-count
+group: api
+category: sink
+update-time: 20260613
+description: Read the cumulative dropped-record count from a QueuedSink.
+key-word:
+    - sink
+    - queue
+    - dropped
+    - public
+---
+
+## Queued-sink-dropped-count
+
+Read the cumulative dropped-record count from a `QueuedSink[S]`. This helper exposes the direct sink-level loss metric when a bounded queue overflows.
+
+### Interface
+
+```moonbit
+pub fn[S] QueuedSink::dropped_count(self : QueuedSink[S]) -> Int {
+```
+
+#### input
+
+- `self : QueuedSink[S]` - Queued sink whose cumulative dropped-record metric should be inspected.
+
+#### output
+
+- `Int` - Number of dropped records reported by the queue wrapper.
+
+### Explanation
+
+Detailed rules explaining key parameters and behaviors
+
+- The counter value comes from `self.dropped_count.val`.
+- It increases when the queue is full and an incoming record cannot be kept without applying overflow handling.
+- The counter is cumulative for the lifetime of the concrete queued sink value.
+- This helper does not explain why drops happened beyond exposing the count.
+
+### How to Use
+
+Here are some specific examples provided.
+
+#### When Need Direct Queue Loss Visibility
+
+When a bounded queue may discard records under pressure:
+```moonbit
+if sink.dropped_count() > 0 {
+  println("queue dropped records")
+}
+```
+
+In this example, direct queue loss becomes visible without inspecting internal fields manually.
+
+#### When Compare Overflow Tuning Changes
+
+When queue capacity or overflow policy should be checked operationally:
+```moonbit
+ignore(sink.dropped_count())
+```
+
+In this example, callers can compare the cumulative drop metric across different queue settings.
+
+### Error Case
+
+e.g.:
+- If the queue never overflows, this counter remains `0`.
+
+- If callers need a higher-level wrapper around runtime-configured sinks, `RuntimeSink::dropped_count()` or `ConfiguredLogger::dropped_count()` may fit better.
+
+### Notes
+
+1. This helper reports cumulative loss, not the exact overflow moment.
+
+2. Pair it with `pending_count()`, `drain()`, and queue configuration when investigating pressure.