co-log-core
co-log
co-log-polysemy

co-log is a composable and configurable logging framework. It combines all the benefits of Haskell idioms to provide a reasonable and convenient interface. Though it uses some advanced concepts in its core, we are striving to provide beginner-friendly API. The library also contains complete documentation with a lot of beginner-friendly examples, explanations and tutorials to guide users. The combination of a pragmatic approach to logging and fundamental Haskell abstractions allows us to create a highly composable and configurable logging framework.

If you’re interested in how different Haskell typeclasses are used to implement core functions of co-log, you can read the following blog post which goes into detail about internal implementation specifics:

• co-log: Composable Contravariant Combinatorial Comonadic Configurable Convenient Logging

co-log is also modular on the level of packages. We care a lot about a low dependency footprint so you can build your logging only on top of the minimal required interface for your use-case. This repository contains the following packages:

• co-log-core: lightweight package with basic data types and general idea which depends only on base.
• co-log: taggless final implementation of logging library based on co-log-core.
• co-log-polysemy: implementation of logging library based on co-log-core and the polysemy extensible effects library.
• co-log-benchmark: benchmarks of the co-log library.

To provide a more user-friendly introduction to the library, we’ve created the tutorial series which introduces the main concepts behind co-log smoothly:

• Intro: Using LogAction
• Using custom monad that stores LogAction inside its environment

co-log also cares about concurrent logging. For this purpose we have the concurrent-playground executable where we experiment with different multithreading scenarios to test the library’s behavior. You can find it here:

• tutorials/Concurrent.hs

Benchmarks🔗

co-log is compared with basic functions like putStrLn. Since IO overhead is big enough, every benchmark dumps 10K messages to output. If a benchmark’s name doesn’t contain Message then this benchmark simply dumps the string "message" to output, otherwise it works with the Message data type from the co-log library.

To run benchmarks, use the following command:

cabal v2-run co-log-bench

Benchmarks	Time for 10K messages
Prelude.putStrLn	5.117ms
Text.putStrLn	9.220ms
ByteString.putStrLn	2.971ms
mempty	1.181ms
logStringStdout	5.107ms
logPrint	5.248ms
logTextStdout	5.351ms
logByteStringStdout	2.933ms
logByteStringStderr	17.482ms
ByteString > (stdout <> stderr)	17.715ms
Message > format > stdout	9.188ms
Message > format > ByteString > stdout	3.524ms
Message{callstack} > format > stdout	9.139ms
Message{callstack:5} > format > stdout	9.464ms
Message{callstack:50} > format > stdout	9.439ms
Message{Time,ThreadId} > format > stdout	54.160ms
Message{Time,ThreadId} > format > ByteString > stdout	54.137ms