[PoC] feat: Implement a more feature complete telemetry system #2389
Conversation
|
@staabm @infection/core happy to get a first round of feedback about the functionality (i.e. not the code). The description is up to date, but there is more options available about what to display. A few notes to keep in mind/consider:
We parsed fileA, it resulted in X mutations/mutants. Heuristics = checks done on the mutation to see if we can kill it before executing it in a separate process. I use the term
What I want to achieve with this system, is to have a reliable way to assess the execution of infection, to evaluate if a change is an improvement or not. I think the base framework introduced here is a good base, but I'm unsure of what it is missing to be able to use it for this purpose comfortably. I already invested quite some time on it, hence I would appreciate a round of feedback before proceeding any further. I am totally fine with "this is not going anywhere, let's drop it", or defining what sort of result we want to see. |
This PR is not ready for review yet.
This PR is a draft for me to explore a more complete telemetry system. The goal is to have more insights about the flow & performance for a given run.
The current terminology employed is borrowed from well established telemetry systems (e.g. OpenTelemetry or Sentry).
Currently each span* has the following information:
*: To those not familiar with the concept of span, the general idea is that it is a unit of an operation. Here is an example of spans with a UI (which I find easier to interpret than the blob I printed bellow):
The snapshot includes:
Currently at the end of the run command the trace is serialized and dumped into a file. One can then view it using
$ infection telemetry:dump-trace build/trace/<trace-id>.The current output is something like this:
example of trace dump
Missing pieces:
--profile?)