Skip to content

Latest commit

 

History

History
281 lines (223 loc) · 15 KB

README.md

File metadata and controls

281 lines (223 loc) · 15 KB
Raw

Publish Nuget

Description

This library provides a cross-platform service easily creating and running tasks on a schedule. It makes extensive use of NodaTime to allow for robust time zone handling, as well as Cronos for Cron expression support. Importantly, it is well architected to support testing, dependency injection and modern design methodologies.

Installation

The library is available via NuGet and is delivered via NuGet Package Manager:

Install-Package DevDecoder.Scheduling

If you are targeting .NET Core, use the following command:

dotnet add package 
Install-Package DevDecoder.Scheduling

Usage

The package exposes several key interfaces, and implementations that allow for easily creating and running tasks on a schedule.

For example:

// Only create one scheduler per application, and dispose when finished with it.
using var scheduler = new Scheduler();
...

// Run a job 3 times with a gap of 5 seconds between each execution.
var job = scheduler.Add(state => Console.WriteLine($"Execution {++counter}, due: {state.Due:ss.fffffff}"),
    new LimitSchedule(3, new GapSchedule(Duration.FromSeconds(5))));

// We can use the returned job, to execute manually or disable the job temporarily, etc..
job.IsEnabled = false;
Assert.False(job.IsExecuting);

IScheduler

Initialisation

The DevDecoder.Scheduling.IScheduler interface is the main scheduler service, it is implemented by DevDecoder.Scheduling.Scheduler concrete type. Note it also implements IDisposable, so should be disposed when your application terminates.

You can create a new scheduler in the main entry point of your code:

using var scheduler = new Scheduler();

However, the library is designed to be used with a dependency injection framework. You can register the service as a singleton, and have it injected automatically into your other services, or retrieve it manually:

// Add as a singleton, accessible by its interface.
services.AddSingleon<IScheduler, Scheduler>();
...
// Retrieve manually from the service provider
var scheduler = serviceProvider.GetService<IScheduler>();

Modern DI frameworks should correctly handle instantiation and disposal automatically, as well as suppplying a logger if registered.

Specifying a Clock

The scheduler retrieves the current time from an IPreciseClock. There are 4 clocks provided, which should cover every eventuality but you can easily create your own if desired.

  • StandardClock - This is equivalent to using the built in DateTime.UTCNow function, which is usually accurate to ~100ns. It is the default choice, an suitable for most applications.
  • FastClock - This uses the Query Performance Counters to get the most accurate timestamp, however, it is not synchronized to any external source, though it is often accurate to <100 clock cycles. On some systems, the clock is not available (see FastClock.IsAvailable, and so defaults to the StandardClock). It is rare that this clock is necessary.
  • SynchronizedClock - This clock uses GetSystemTimePreciseAsFileTime to get an accurate, synchronized time, where available. This is recommended for scenarios where it is important for multiple machines to stay in synch, e.g. during networking scenarios.
  • TestClock - This clock can be used during testing to allow you to control what times are returned when querying. It accepts a function that, when given an Instant returns the next Instant. Two static functions Fixed, which supplies a clock that returns the same Instant every time, and From which provides a clock that returns an Instant from a specific time, and increments by a set Duration each time it is queried, are supplied for convenience. There is also TestClock.Never which always returns Instant.MaxValue (the 'end of time'!).

All clocks have a static Instance property that can get their singleton implementation, and this can be supplied directly to the Scheduler on creation:

// Use the Synchronized clock.
using var scheduler = new Scheduler(SynchronizedClock.Instance);

However, you can also specify the ClockPrecision enumeration, e.g.

// Use the Synchronized clock.
using var scheduler = new Scheduler(ClockPrecision.Synchronized);

Using dependency injection:

// Add the clock singleton to the services collection.
services.AddSingleton<IPreciseClock>(SynchronizedClock.Instance);
// Add as a singleton, accessible by its interface.
services.AddSingleon<IScheduler, Scheduler>();

Timezone handling

The schedule uses NodaTime to ensure it handles timezones accurately. To that end it can be injected with an IDateTimeZoneProvider on creation.

Logging

The Scheduler constructor accepts an ILogger<Scheduler> for logging, this is normally injected via dependency injection, but here is an example of a simple console logger.

Maximum Execution duration

The MaximumExecutionDuration can also be specified during creation, or via the IScheduler interface. This defaults to Scheduler.DefaultMaximumDuration (which is 10 mins), but you can set this to any duration (including Duration.MaxValue) at any time (including during construction). It will set any job, that doesn't have the ScheduleOptions.LongRunning flag set, to cancel after the duration has elapsed. As such, if you do have a job that might take longer, you must set the ScheduleOptions.LongRunning flag on the schedule.

ISchedule

The scheduler runs jobs on a schedule. These can be as complex as your imagination allows, so long as they implement ISchedule, in particular:

ZonedDateTime? Next(IScheduler scheduler, ZonedDateTime last);

That is, given the last time a job started, or completed (based on whether the ScheduleOptions.FromDue flag is set for the schedule), it needs to return the next time the job should execute. On first execution, or first running after being disabled, this will be the current date and time.

The following schedules are built in for convenience:

OneOffSchedule

The one off schedule allows a task to run once, at a fixed date and time, e.g.

// Execute at midday (UTC) on 1st January 2023
new OneOffSchedule(new ZonedDateTime(Instant.FromUtc(2023, 1, 1, 12, 0), DateTimeZone.Utc));

GapSchedule

The gap schedule allows a task to run repeatedly, with a fixed interval. The interval can be measured from the start of the proceeding execution, or from it's conclusion, using the ScheduleOptions.FromDue flag, e.g.

// Execute with a 5 second gap between the start of each invocation.
new GapSchedule(Duration.FromSeconds(5), ScheduleOptions.FromDue);

FunctionalSchedule

The functional schedule is a convenience class that accepts a lambda to calculate the next date/time, e.g.

// Execute every 10 seconds (rounded up to nearest second).
new FunctionalSchedule(t => t.PlusSeconds(10), ScheduleOptions.AlignSeconds | ScheduleOptions.FromDue);

LimitSchedule

The limit schedule wraps any schedule, limiting how many times it will execute, e.g.:

// Execute 3 times, with a 5ms gap between each execution.
new LimitSchedule(3, new GapSchedule(Duration.FromMilliseconds(5)))

AggregateSchedule

The aggregate schedule is extremely powerful as it allows you to combine multiple schedules together, e.g.:

// Execute every 5th second (aligned) and every 3rd second (for the first 3 times).
new AggregateSchedule(
    new GapSchedule(Duration.FromSeconds(5), ScheduleOptions.AlignSeconds),
    new LimitSchedule(3, 
        new FunctionalSchedule(t => t.PlusSeconds(3), ScheduleOptions.AlignSeconds)));

CronSchedule

Finally, we also expose a schedule that can accept any chron expression, e.g.:

// Execute every 2 minutes from 1:00 AM to 01:15 AM and from 1:45 AM to 1:59 AM and at 1:30 AM
new CronSchedule("30,45-15/2 1 * * *")
// Note we also support the including the optional seconds format:
new CronSchedule("0 30,45-15/2 1 * * *", CronFormat.IncludeSeconds)

ScheduleOptions

Every schedule also exposes the ScheduleOptions flags which have the following meanings:

Flag Effect when set
LongRunning The job will not be limited by the MaximumExecutionDuration.
IgnoreErrors The job will not be disabled when an exception is thrown.
FromDue The ISchedule.Next method will be called with the time the previous execution was due to start, rather than when it finished.
AlignSeconds The result of ISchedule.Next will be rounded up to the nearest second by the scheduler.
AlignMinutes The result of ISchedule.Next will be rounded up to the nearest minute by the scheduler.
AlignHours The result of ISchedule.Next will be rounded up to the nearest hour by the scheduler.
AlignDays The result of ISchedule.Next will be rounded up to the nearest midnight by the scheduler.

Jobs

Any object that implements the simple IJob interface can be passed to the scheduler for scheduling:

public interface IJob
{
    /// <summary>
    ///     An optional job name.
    /// </summary>
    public string Name { get; }

    /// <summary>
    ///     Executes the current job.
    /// </summary>
    /// <param name="state">Information regarding the current job, (see <see cref="IJobState" />).</param>
    /// <param name="cancellationToken">The cancellation token.</param>
    /// <returns>An awaitable task.</returns>
    Task ExecuteAsync(IJobState state, CancellationToken cancellationToken = default);
}

However, there is also a convenient SimpleJob class that allows the creation of any job from an action or function, using one of the SimpleJob.Create or SimpleJob.CreateAsync overloads. Most conveniently though, there are numerouse extension methods on IScheduler that overload the Add and AddAsync methods to create a SimpleJob automatically, e.g.:

// Run a simple function in 10s, logging it's result on completion.
scheduler.Add(state => $"Log this result {state.Due}",
    new OneOffSchedule(scheduler.GetCurrentZonedDateTime().PlusSeconds(10)));

Although the scheduling system is really designed to run actions, if you supply SimpleJob with a function, it will log any result returned, e.g.

[2022-11-07 18:50:41Z] info: DevDecoder.Scheduling.Scheduler[0]
      The 'state => $"Log this result {state.Due}"' job returned: Log this result 2022-11-07T18:50:41 GMT Standard Time (+00)

You will also note that jobs are automatically named based on the arguments passed into the function/action, for easier debugging.

IJobState

When a job is executed it is passed an IJobState and a CancellationToken. The latter of these should be respected to allow for easy termination of overdue jobs. The first allows the execution function access to lots of useful information about the current execution, and allows the job to disable itself, preventing further execution.

/// <summary>
///     Holds information for the currently executing <see cref="IJob">job</see>.
/// </summary>
public interface IJobState
{
    /// <summary>
    ///     A unique identified for the scheduled job.
    /// </summary>
    Guid Id { get; }

    /// <summary>
    ///     An optional job name.
    /// </summary>
    string Name { get; }

    /// <summary>
    ///     The <see cref="IScheduler">scheduler</see> executing this job.
    /// </summary>
    IScheduler Scheduler { get; }

    /// <summary>
    ///     The <see cref="ISchedule">schedule</see> that triggered this execution.
    /// </summary>
    /// <remarks>Will be <c>null</c> if the job was <see cref="IsManual">executed manually</see>.</remarks>
    ISchedule? Schedule { get; }

    /// <summary>
    ///     The <see cref="Instant">instant</see> the job was due to run.
    /// </summary>
    /// <remarks>This will be when the job was requested.</remarks>
    ZonedDateTime Due { get; }

    /// <summary>
    ///     The current logger, if any; otherwise <c>null</c>.
    /// </summary>
    ILogger? Logger { get; }

    /// <summary>
    ///     If <c>true</c> then the current execution was triggered manually; otherwise <c>false</c>.
    /// </summary>
    bool IsManual { get; }

    /// <summary>
    ///     If <c>true</c> then the job is currently executing; otherwise <c>false</c>.
    /// </summary>
    bool IsExecuting { get; }

    /// <summary>
    ///     If <c>true</c> then the job is allowed to execute; otherwise <c>false</c>, prevents further executions.
    /// </summary>
    bool IsEnabled { get; set; }
}

IScheduledJob

Similarly, when a job is added to the IScheduler it returns an IScheduledJob. This is almost identical to IJobState, also allowing control over whether the job is enabled, but also allowing for manual execution of the job.

Note: A job will never be executed concurrently with itself. If a job is executed manually, whilst it is also executing as part of a schedule, the manual execution will receive the same task, and vice-versa. It is effectively 'debounced', meaning that a job execution is inherently thread-safe.

Job Removal

As well as disabling a job using IScheduledJob.IsEnabled, you can permanently remove it from the scheduler using IScheduler.TryRemove. A removed job can still be executed manually, but it's Scheduler and Due properties will always be null and IsEnabled will always be false. You cannot re-add a removed job, you must re-create a job using the Add or AddAsync methods. As such, if you want to temporarily 'remove' a job, use the IsEnabled property instead.

TODO

  • Explicit schedule serialization support.

Testing status

  • Unit tests can be found in the DevDecoder.Scheduling.Test project.

Acknowledgements