MEL.Flex (FSharp Logging EXtensions for Microsoft.Extensions.Logging) adds the ability to use string interpolation for strongly typed strings while getting the benefits of structured logging by converting the interpolated string into a message template transparently.
Structured logging with message templates are great, except they do one a few problems.
-
They are (usually) positional. If I have the following code:
let userName = "KirkJ1701" logger.LogWarning("Some user: {UserName} has logged in", userName)
if I wanted to add an IpAddress
let userName = "KirkJ1701" let ipAddress = "KirkJ1701" logger.LogWarning("Some user: {UserName} has logged in from {IpAddress}", ipAddress, userName)
I can easily mess up the arguments. This of course looks easy but gets more difficult as you add more logs to structure or rearrange your logs.
-
Normalizing your keys with semantic convetions makes it easier for you to search across your logs.
let userName = "SpockV" logger.LogWarning("Some user: {user_name} has logged in", userName)
Adding this log to our application logs the
user_name
but earlier we were logging aUserName
. If I needed to search across these logs, I would need to know each potential variation of that key name. (Yes some tools allow you to do that mapping on their side but not all).
One of the currently supported ways is to use tuples. Taking an example above:
// Some file that containts your normalized names
module LoggerKeyConsts =
let [<Literal>] UserName = "UserName"
// .. Some function
let userName = "SpockV"
logger.LogFWarning($"Some user: {(LoggerKeyConsts.UserName, userName)} has logged in")
The important changes are:
LogWarning
->LogFWarning
$
in front of the string{Username}
became{(LoggerKeyConsts.UserName, userName)}
- The latter part of this syntax is a tuple with two values.
Slightly better would be to create helper functions that generate the tuples:
module LogConsts =
let [<Literal>] ``user.name`` = "user.name"
let inline userName (s : string) = struct (``user.name``, s)
// .. Some function
let userName = "SpockV"
logger.LogFWarning($"Some user: {LogConsts.userName userName} has logged in")
This way, you could apply any type safety or normalization to your log data.
- Unfortunately F# does not yet support DefaultInterpolatedStringHandler which means you will still take the interpolated creation hit. However, this library does implement it's own log formatter which allows for lazy construction of the interpolated string -> message template until it is required or possibly not at all if the LogLevel configuration is set to a higher threadshold than the log statement.
GitHub Actions |
---|
Package | Stable | Prerelease |
---|---|---|
MEL.Flex |
Make sure the following requirements are installed on your system:
- dotnet SDK 3.0 or higher
- Mono if you're on Linux or macOS.
or
CONFIGURATION
will set the configuration of the dotnet commands. If not set, it will default to Release.CONFIGURATION=Debug ./build.sh
will result in-c
additions to commands such as indotnet build -c Debug
GITHUB_TOKEN
will be used to upload release notes and Nuget packages to GitHub.- Be sure to set this before releasing
DISABLE_COVERAGE
Will disable running code coverage metrics. AltCover can have severe performance degradation so it's worth disabling when looking to do a quicker feedback loop.DISABLE_COVERAGE=1 ./build.sh
> build.cmd <optional buildtarget> // on windows
$ ./build.sh <optional buildtarget>// on unix
The bin of your library should look similar to:
$ tree src/MEL.Flex/bin/
src/MEL.Flex/bin/
└── Debug
└── net50
├── MEL.Flex.deps.json
├── MEL.Flex.dll
├── MEL.Flex.pdb
└── MEL.Flex.xml
Clean
- Cleans artifact and temp directories.DotnetRestore
- Runs dotnet restore on the solution file.DotnetBuild
- Runs dotnet build on the solution file.DotnetTest
- Runs dotnet test on the solution file.GenerateCoverageReport
- Code coverage is run duringDotnetTest
and this generates a report via ReportGenerator.WatchTests
- Runs dotnet watch with the test projects. Useful for rapid feedback loops.GenerateAssemblyInfo
- Generates AssemblyInfo for libraries.DotnetPack
- Runs dotnet pack. This includes running Source Link.SourceLinkTest
- Runs a Source Link test tool to verify Source Links were properly generated.PublishToNuGet
- Publishes the NuGet packages generated inDotnetPack
to NuGet via paket push.GitRelease
- Creates a commit message with the Release Notes and a git tag via the version in theRelease Notes
.GitHubRelease
- Publishes a GitHub Release with the Release Notes and any NuGet packages.FormatCode
- Runs Fantomas on the solution file.BuildDocs
- Generates Documentation fromdocsSrc
and the XML Documentation Comments from your libraries insrc
.WatchDocs
- Generates documentation and starts a webserver locally. It will rebuild and hot reload if it detects any changes made todocsSrc
files, libraries insrc
, or thedocsTool
itself.ReleaseDocs
- Will stage, commit, and push docs generated in theBuildDocs
target.Release
- Task that runs all release type tasks such asPublishToNuGet
,GitRelease
,ReleaseDocs
, andGitHubRelease
. Make sure to read Releasing to setup your environment correctly for releases.
git add .
git commit -m "Scaffold"
git remote add origin https://github.com/user/MEL.Flex.git
git push -u origin master
-
paket config add-token "https://www.nuget.org" 4003d786-cc37-4004-bfdf-c4f3e8ef9b3a
- or set the environment variable
NUGET_TOKEN
to your key
- or set the environment variable
-
- You can then set the environment variable
GITHUB_TOKEN
to upload release notes and artifacts to github - Otherwise it will fallback to username/password
- You can then set the environment variable
-
Then update the
CHANGELOG.md
with an "Unreleased" section containing release notes for this version, in KeepAChangelog format.
NOTE: Its highly recommend to add a link to the Pull Request next to the release note that it affects. The reason for this is when the RELEASE
target is run, it will add these new notes into the body of git commit. GitHub will notice the links and will update the Pull Request with what commit referenced it saying "added a commit that referenced this pull request". Since the build script automates the commit message, it will say "Bump Version to x.y.z". The benefit of this is when users goto a Pull Request, it will be clear when and which version those code changes released. Also when reading the CHANGELOG
, if someone is curious about how or why those changes were made, they can easily discover the work and discussions.
Here's an example of adding an "Unreleased" section to a CHANGELOG.md
with a 0.1.0
section already released.
## [Unreleased]
### Added
- Does cool stuff!
### Fixed
- Fixes that silly oversight
## [0.1.0] - 2017-03-17
First release
### Added
- This release already has lots of features
[Unreleased]: https://github.com/user/MEL.Flex.git/compare/v0.1.0...HEAD
[0.1.0]: https://github.com/user/MEL.Flex.git/releases/tag/v0.1.0
- You can then use the
Release
target, specifying the version number either in theRELEASE_VERSION
environment variable, or else as a parameter after the target name. This will:- update
CHANGELOG.md
, moving changes from theUnreleased
section into a new0.2.0
section- if there were any prerelease versions of 0.2.0 in the changelog, it will also collect their changes into the final 0.2.0 entry
- make a commit bumping the version:
Bump version to 0.2.0
and adds the new changelog section to the commit's body - publish the package to NuGet
- push a git tag
- create a GitHub release for that git tag
- update
macOS/Linux Parameter:
./build.sh Release 0.2.0
macOS/Linux Environment Variable:
RELEASE_VERSION=0.2.0 ./build.sh Release