Chronograph.Serilog 2.2.1

Chronograph

Chronograph is an instrumentation object which encapsulates Stopwatch instance and serilog parametrized messages for operation start and end reporting.

It implements IDisposable for convenient operation wrapping. When Dispose method is called on Chronograph instance the internal Stopwatch gets stopped and operation duration along with action description and optional end action message gets written to a logger which Chronograph was associated with upon creation.

The general pattern of the Chronograph output is :

[{Time} {Log event level}] Started {action description}
[{Time} {Log event level}] Finished {action description} {optional end action report mesage} [{action duration}]

Following is the example output during the simple (without end action message) Chronograph lifetime.

[16:28:38 INF] Started getting products.
[16:34:03 INF] Finished getting products. [0:05:25.3627059]

Where:

• The action description is getting products
• The operation duration is [0:05:25.3627059]

Following is the example of the extended Chronograph output.

[16:28:38 INF] Started getting products with batch size of 100.
[16:34:03 INF] Finished getting products with batch size of 100. Got 65536 products. [0:05:25.3627059]

Where:

• The action description is getting products with batch size of {batchSize}.
• The action description template parameter is 100.
• The end action report message is Got {productsCount} products from.
• The end action report message parameter is 65536. These parameters got calculated after operation completes.
• The operation duration is [0:05:25.3627059]

There can be multiple action description and end action report message template parameters.

Event level

The Chronograph uses its own event level ChronographLoggerEventLevel to abstract away the concrete logger event levels.

Te supported event levels and their respective translations to Microsoft.Extensions.Logging.Abstractions.LogLevel and Serilog.Events.LogEventLevel loggers are

• Verbose ⇒ LogLevel.Trace, LogEventLevel.Verbose
• Debug ⇒ LogLevel.Debug, LogEventLevel.Debug
• Information ⇒ LogLevel.Information, LogEventLevel.Information
• Warning ⇒ LogLevel.Warning, LogEventLevel.Warning
• Error ⇒ LogLevel.Error, LogEventLevel.Error
• Fatal ⇒ LogLevel.Critical, LogEventLevel.Fatal
• None ⇒ LogLevel.None applies only to Microsoft.Extensions.Logging logger

The default event level for each concrete logger messages can be configured using static specific logger helper class DefaultChronographEventLevel property.

For Microsoft.Extensions.Logging this helper class is called MicrosoftExtensionsLoggerHelper. For Serilog this helper class is called SerilogLoggerHelper.

The default event level for all the concrete loggers is ChronographLoggerEventLevel.Information.

The same class for each concrete logger contains static methods for converting abstract ChronographLoggerEventLevel event level to concrete one ToTargetEventLevel and backwards ToAbstractEventLevel.

Message enrichment

Chronograph enriches all end ('Finished xxx') messages with the following field:

• OperationDurationMilliseconds - the measured operation duration in milliseconds - for later aggregation convenience;

In addition to default field, Chronograph can enrich end message with custom fields. To provide fields to enrich message with use WithParameter or WithParameters methods.

Simple Chronograph creation

The main functionality of the chronograph is implemented in Dispose method, so it is recommended to use Chronograph in using statements. Although it is not always convenient to do so. In this case, call Dispose method manually (or use C#8 using declaration syntax when full support of this feature lands).

Chronograph class provides Create factory method. This method is used to create empty (without any message) non-started chronograph for subsequent fluent configuration methods calls. To start chronograph, call one of the Start method overloads.

using(Chronograph.Create(Logger).WithEventLevel(LogEventLevel.Information).Start("action description"))
{
    // operation to time
}

To shorten the call, use provided Chrono extension method on ILogger instance.

using(Logger.Chrono().Start("action description"))
{
    // operation to time
}

Or even shorter.

using(Logger.Chrono("action description"))
{
    // operation to time
}

The result of the all the above calls will be as follows (operation duration may vary depending on operation to time).

[16:28:38 INF] Started action description.
[16:34:03 INF] Finished action description. [0:05:25.3627059]

Chrono() method automatically sets the LogEventLevel to Information (you can edit this settings by modifying MicrosoftExtensionsLoggerHelper.DefaultChronographEventLevel or SerilogLoggerHelper.DefaultChronographEventLevel property depending on which structured logging library you use).

Action description lowercasing and punctuation is automatic, so there is no need to lower/upper-case action description message or write full stops in message templates.

If you want to prevent automatic lowercasing of the action description message - prefix it with a space. That space will be removed and message will be outputted in the original casing.

using(Logger.Chrono(" Action description"))
{
    // operation to time
}

Will output action description in the original casing

[16:28:38 INF] Started Action description.
[16:34:03 INF] Finished Action description. [0:05:25.3627059]

Specifying parameters

To specify action description parameters, use either of the following.

Logger.Chrono()
    .Start(
        "{parametrized} action description with {parameterCount} parameters", 
        "first_parameter_value", 
        2
    )

Logger.Chrono()
    .For(
        "{parametrized} action description with {parameterCount} parameters", 
        "first_parameter_value", 
        2
    )
    .Start()

The result of the both calls will be as follows (operation duration may vary depending on operation to time).

[16:28:38 INF] Started first_parameter_value action description with 2 parameters.
[16:34:03 INF] Finished first_parameter_value action description with 2 parameters. [0:03:44.3622359]

Parameter values are params and optional.

Please note that when you are using parameterized messages or other extensions you need to explicitly Start() the chronograph. If you won't the Chronograph instance will issue a warning upon disposal.

As a rule of thumb - call Start() in any case except for when using the most shorthand chronograph creation method : Logger.Chrono("action description").

Someday I will add a Roslyn analyzer to ensure that you do.

Specifying end action report message

To specify end action report message, use either of the following.

using(Logger.Chrono()
    .For("acton description")
    .Report(
        "Loaded {productsCount} products",
        () => loadedProducts.Count
    )
    .Start())
{
    loadedProducts.AddRange(await LoadProducts());
}

using(Logger.Chrono()
    .Report(
        "Loaded {productsCount} products",
        () => loadedProducts.Count
    )
    .Start("acton description"))
{
    loadedProducts.AddRange(await LoadProducts());
}

The result of the both calls will be as follows (operation duration may vary depending on operation to time).

[16:28:38 INF] Started acton description.
[16:34:03 INF] Finished acton description. Loaded 100 products. [0:01:44.3622359]

Note that end action report message also uses parameters, but to specify those you need to use lambda expressions. This is because these parameters got calculated after operation completes and the chronograph instance is created before it. Using lambda expressions allows us to use closures over some parameters that will be changed during the course of teh timed operation, so we can get actual final values of those changed parameters when the operation completes and chronograph message gets rendered.

Important: if you use ReSharper, the closure upon changed variable will trigger AccessToModifiedClosure diagnostic. If you suppress this diagnostic for a method, please do watch for other places in your method where you might (unintentionally) use closures over modified values to avoid subtle bugs.

If you do not need to output counts or other dynamically changed (during timed operation) data, simply return a constant literal (this does not trigger ReSharper diagnostic).

You can combine both parametrized versions of end action report and action description.

End action report message parameters are params and optional.

Specifying end action report message on manual chronograph object Dispose()

You can override the end action report message with another one. To do so - manually call Dispose(string endMessageTemplate, params Func<object>[] countProviders) on the desired Chronograph instance.

 var chrono = logger.Chrono()
    .For("Some operation")
    .Report(
        "Teport count={CounterValue}, secondCount={CounterValue2}",
        () => 42,
        () => 1567
    )
    .Start();

// further in the code

chrono.Dispose(
    "Overridden end message with parameter={Parameter}", 
    () => 43
);

Long-running operation report

You can set up the chronograph to issue standardised or custom message when the operation is running longer than some set amount of time.

To do so use either of the WithLongRunningOperationReport builder method overloads.

Chronograph WithLongRunningOperationReport(
    TimeSpan longRunningOperationThreshold, 
    string longRunningOperationReportMessage = null,
    params object[] longRunningOperationReportMessageParameters);

Chronograph WithLongRunningOperationReport(
    TimeSpan longRunningOperationThreshold,
    string longRunningOperationReportMessage = null,
    params Func<object>[] longRunningOperationReportMessageParameterProviders);

If you omit the longRunningOperationReportMessage, the standardised message will be written.

$"{specified action description here} took a long time to finish >({specified long running operation threshold here}) : [{Elapsed:g}]"

The second overload WithLongRunningOperationReport with Func<object>[] formal parameter works the same as the similar Report overload does. It evaluates the functions upon message rendering allowing for closures to report some values that will be changed during the timed operation run.

On start and on end operation actions

Sometimes it is useful to call some code on start and/or on end of the operation chronograph.

To register such action for start operation use WithOnStartAction(Action<IReadOnlyList<object>> onStartAction) builder method. This action has a single argument IReadOnlyList<object> that contains all the start action parameters if any. This action gets called before chronograph writes the start action message.

To register such action for end operation use WithOnEndAction(Action<Stopwatch, IReadOnlyList<object>> onEndAction) builder method. This action has two arguments : an internal Stopwatch instance (stopped by the time the action is called) and a IReadOnlyList<object> with all the end action parameters that contain specified start action parameters, end action parameters and a string, containing elapsed operation time. This action gets called before chronograph writes the end action message.

var chrono = chronograph
    .For("test operation {IntParameter}", 42)
    .Report("Operation result {IntParameter}", () => 1567)
    .WithOnStartAction((parameters) => { /*some code here*/ })
    .WithOnEndAction((sw, parameters) => { /*some code here*/ })
    .Start();

Sampling

For cases when the chronograph wraps a frequently called code, it's beneficial to output only some of the start/end action messages and reports to the logger. This is where sampling can be handy. To enable message sampling call the WithSampling(uint samplingFactor, bool shouldAlwaysReportLongRunningOperations = true) builder method.

var chrono = chronograph
    .For("test operation {IntParameter}", 42)
    .WithLongRunningOperationReport(
        TimeSpan.FromMilliseconds(1),
        "Long running operation {LongParameter} detected",
        () => 1567)
    .WithSampling(50, shouldAlwaysReportLongRunningOperations: false)
    .Start();

This method have two parameters.

• samplingFactor, which is a value between 1 and 100 that indicates the rough percentage of the messages to output.
• shouldAlwaysReportLongRunningOperations (optional), which indicates whether the samplingFactor should apply to the long-running operation report if it is enabled. By default, long-running operations got reported regardless of the samplingFactor.

The sampling is implemented by getting a random value between 1 and 100 at the chronograph creation and checking samplingFactor against this value.

If you know how to improve this documentation, don't hesitate to suggest any changes by pull request.