tracer_api_plan1.rb

# https://github.com/st0012/ruby_tracer 

Tracer.new(:exception) {}
Tracer.new(:exception).trace {}
Tracer.new(:exception).config(a: true, b: 1, ...).trace {}
Tracer.new(:exception, config_a: true, config_b: 1, ...).trace {}

# The method name "trace" is nice, but maybe a better name?
# "Tracer#trace" seems redundant.

Tracer.new(:call) {}
Tracer.new(:line) {}

tracer = Tracer.new(:call)
tracer.enable or start
...
tracer.disable or finish or suspend

They are similar to TracePoint.
https://docs.ruby-lang.org/ja/latest/class/TracePoint.html#S_NEW

IMO, the Tracer API should NOT look like TracePoint's, because they're used in different ways:

• TracePoint: usually to implement something that'll stay (e.g. a tracer).
• Tracer: usually to be removed/added frequently, like puts. So it should be short and easy to remember.

And given its frequent-editing need, I'm favoring methods Tracer.trace_call over instances Tracer.new(:call) because the former can be autocompleted in editor or IRB.