DatadogTracingAppender

An io.github.smyrgeorge.log4k.Appender that publishes finished spans to a local Datadog Agent over its APM trace intake API (PUT /v0.4/traces, JSON). Register it with RootLogger.Tracing.appenders.register(...).

Spans are shipped in batches (via BatchAppender): a batch is sent as soon as it reaches batchSize spans or when batchTimeout elapses since the first span of the batch arrived — so low traffic never holds spans back for long. Within a batch, spans are grouped by trace, as the intake API expects.

The HTTP transport is a Ktor client: pass an engine explicitly, or add exactly one Ktor engine artifact for your platform (CIO, Darwin, Curl, WinHttp, Js, ...) to the runtime classpath and leave it null so the engine is discovered automatically.

The agent must have APM enabled (DD_APM_ENABLED=true); it listens on port 8126 by default.

Mapping notes:

• log4k's OpenTelemetry-style hex ids are converted to Datadog's numeric ids: the low 64 bits are used, and for 128-bit trace ids the high 64 bits are preserved in the _dd.p.tid tag. Non-hexadecimal ids (e.g. from a remote/propagated context) are hashed (FNV-1a).

• Finite numeric span tags become Datadog metrics; everything else — including non-finite numbers, which metrics cannot represent — goes to meta (stringified).

• Span events are encoded as a JSON array in meta.events, mirroring the OpenTelemetry ingestion convention. An error status sets error=1 plus error.type/message/stack.

Delivery is best-effort: agent failures are reported to the console, and the batch is dropped — the appender itself keeps running.

Parameters