Open bryannaegele opened 4 years ago
How about adding an option to disable logging for specific events? Although that is mostly equivalent to filtering, but tailored to only that specific use-case.
If we add filtering, I think it needs to be implemented via keep
, drop
, to avoid adding new options.
If we added it here then it would just be adding a second arg to keep/drop with the measurement values.
The other alternative is I just silently drop events with nil
measurement values but warn on missing measurement keys. 🤷🏻
I want to ship v1.0 and this is an open issue which may require breaking changes. How do we feel about this 3 years later? Wouldn't you be able to discard idle=nil events using keep/drop?
We don't currently accept measurements as an arg to the predicates. If we wanted to add it later, it wouldn't be a breaking change, so not a blocker for 1.0
Ah, we could easily make it a two arity function with measurement, metadata
, so it sounds good to me. I can work on a PR for this. :)
I'm currently logging a debug "skipping event due to missing measurement" which is helpful for identifying an issue but is sometimes an unavoidable thing, such as
:idle
beingnil
in an Ecto query.I could add a "report once" check but that would require another ets lookup on something that could be very frequent. Not a huge deal. Or we could enable filtering on measurements here, which we discussed during the addition of filtering on metadata.
Thoughts?