co-log / co-log-core Goto Github PK

I wonder if a constructor Success would be useful for the Severity data type to distinguish critical operations that completed successfully in the log stream. Of course all of this can go into Info, but I feel an extra layer would help.

Green might be a fitting colour, if you're willing to consider this addition.

Our main goal in co-log-core is to create a minimal package with only base in dependencies and enabling more people to use co-log capabilities in a more flexible way.

As proposed by @dcoutts, it might be helpful for other people to:

maximise the ability to use the pattern in different applications by minimising the assumptions

and by moving some parts of co-log-core into co-log.

So let's start a discussion about possible changes in the co-log-core interface.

• Fix Haddock with CPP for TypeErrors instance
• Better docs for cmapMaybe
• Fix hoistLogAction example of type performPureLogsInIO
• Better top-level docs for Colog.Core.Action module
• Typo in cfilterM docs

Would be nice to support this version such that Stackage could upgrade

• commercialhaskell/stackage/issues/6875

I find myself wanting this type:

data WithSeverity a = WithSeverity a Severity

i.e. a message tagged with a severity. Then I can do something like this

data Log1 = ...
data Log2 = ALog1 Log1 | ...

adapt :: LogAction m (WithSeverity Log2) -> LogAction m (WithSeverity Log1)
adapt = cmap (fmap ALog1)

Given that co-log-core provides Severity, perhaps it would be nice to provide WithSeverity also?

Current README duplicates previous co-log README. Ideally, we want to see here something like in the following README of prolens with a short description and simple usage.

• https://github.com/kowainik/prolens#readme

It's not uncommon for logging systems to have a level below Debug, often called Trace. I do find this useful: often there's a distinction between "debugging information that's not useful during normal operation" (Debug) and "just vomit out everything" (Trace).

It would be really nice to have tests for the logging functionality, especially property-based testing. However, it's not really clear how to test it. It would be really nice though to have property-based tests for laws for every data type. However, in order to do this, we need to implement the following two pieces:

• Question 1: Some way to generate arbitrary LogAction
• Question 2: Some way to compare LogAction

Question 1 is still open. For Question 2: instead of comparing LogActions we can compare results of PureLogger.

We have doctest for unit testing, so this should be enough in terms of unit tests (though, more functions always can be covered with doctest). But having property-based tests for laws would be really good.

Any thoughts are appreciated!