package ppx_trace

  1. Overview
  2. Docs
A ppx-based preprocessor for trace

Install

Dune Dependency

Authors

Maintainers

Sources

trace-0.7.tbz
sha256=ebd0be29b30b49536c9659882790b9f0c121ffb06c0bec2eaeba8cfed4909339
sha512=d14b72db713315093c931351b9b04d2fd5ce793a8595970fa31cbf71477516ef25de129adddf4075514581fe9ea3e27d998530efacb17c0d00bb5616b8d18b91

README.md.html

Trace

This small library provides basic types that can be used to instrument a library or application, either by hand or via a ppx.

Features

  • [x] spans

  • [x] messages

  • [x] counters

  • [ ] other metrics?

  • [x] ppx to help instrumentation

Usage

To instrument your code, you can simply add trace to your dune/opam files, and then write code like such:

let f x =
  Trace.with_span ~__FILE__ ~__LINE__ "inside-f" @@ fun _sp ->
  (* … code for f *)

let g x =
  Trace.with_span ~__FILE__ ~__LINE__ "inside-g" @@ fun _sp ->
  let y = f x in
  (* … code for f *)

let () =
  Some_trace_backend.setup () @@ fun () ->
  let result = g 42 in
  print_result result

The file test/t1.ml follows this pattern, using trace-tef as a simple backend that emits one JSON object per span/message:

let run () =
  Trace.set_process_name "main";
  Trace.set_thread_name "t1";

  let n = ref 0 in

  for _i = 1 to 50 do
    Trace.with_span ~__FILE__ ~__LINE__ "outer.loop" @@ fun _sp ->
    for _j = 2 to 5 do
      incr n;
      Trace.with_span ~__FILE__ ~__LINE__ "inner.loop" @@ fun _sp ->
      Trace.messagef (fun k -> k "hello %d %d" _i _j);
      Trace.message "world";
      Trace.counter_int "n" !n
    done
  done

let () =
  Trace_tef.with_setup ~out:(`File "trace.json") () @@ fun () ->
  run ()

After running this, the file "trace.json" will contain something like:

[{"pid":2,"name":"process_name","ph":"M","args": {"name":"main"}},
{"pid":2,"tid": 3,"name":"thread_name","ph":"M","args": {"name":"t1"}},
{"pid":2,"cat":"","tid": 3,"ts": 2.00,"name":"hello 1 2","ph":"I"},
{"pid":2,"cat":"","tid": 3,"ts": 3.00,"name":"world","ph":"I"},
{"pid":2,"tid":3,"ts":4.00,"name":"c","ph":"C","args": {"n":1}},
…

Opening it in https://ui.perfetto.dev we get something like this:

ppx_trace

On OCaml >= 4.12, and with ppxlib installed, you can install ppx_trace. This is a preprocessor that will rewrite like so:

let%trace f x y z =
  do_sth x;
  do_sth y;
  begin
    let%trace () = "sub-span" in
    do_sth z
  end

This more or less corresponds to:

let f x y z =
  let _trace_span = Trace_core.enter_span ~__FILE__ ~__LINE__ "Foo.f" in
  match
    do_sth x;
    do_sth y;
    begin
      let _trace_span = Trace_core.enter_span ~__FILE__ ~__LINE__ "sub-span" in
      match do_sth z with
      | res ->
        Trace_core.exit_span _trace_span;
        res
      | exception e ->
        Trace_core.exit_span _trace_span
        raise e
    end;
  with
  | res ->
    Trace_core.exit_span _trace_span
    res
  | exception e ->
    Trace_core.exit_span _trace_span
    raise e

Alternatively, a name can be provided for the span, which is useful if you want to access it and use functions like Trace.add_data_to_span:

let%trace f x y z =
  do_sth x;
  do_sth y;
  begin
    let%trace _sp = "sub-span" in
    do_sth z;
    Trace.add_data_to_span _sp ["x", `Int 42]
  end

Dune configuration

In your library or executable stanza, add: (preprocess (pps ppx_trace)). The dependency on trace.core is automatically added. You still need to configure a backend to actually do collection.

Backends

Concrete tracing or observability formats such as:

  • [x] Fuchsia (see the spec and tracing. Can be opened in https://ui.perfetto.dev)

  • Catapult

    • [x] light bindings here with trace-tef. (Can be opened in https://ui.perfetto.dev)

    • [ ] richer bindings with ocaml-catapult, with multi-process backends, etc.

  • [x] Tracy (see ocaml-tracy, more specifically tracy-client.trace)

  • [x] Opentelemetry (see ocaml-opentelemetry, in opentelemetry.trace)

  • [ ] landmarks?

  • [ ] Logs (only for messages, obviously)

OCaml

Innovation. Community. Security.