Graceful Death

As you may know, catching fatal errors in PHP is extremely painful aka nearly impossible. This is a library that (partially) solves this issue

Usage

<?php

require __DIR__ . "/vendor/autoload.php";

// The output will be:
//
// Yes, I can ;-)

GracefulDeath::around(function() {
    try {
        // Avoid to print the error in order to have clean output, don't try this at home :-)
        error_reporting(E_ALL ^ E_ERROR);
        // Creating an instance of an unknown class will cause a fatal error
        new UnknownClass();
    } catch(Exception $e) {
        // A fatal error is uncatchable
        echo "You cannot catch this, AHAHAHAHA!!!\n";
    }
})
->afterViolentDeath("Yes, I can ;-)\n")
->run();

Scenario

You have a piece of code that potentially can trigger a fatal error and you want to be able to clean up after it or retry it with some policy. With GracefulDeath you can put this piece of code in a closure and pass it as a parameter to the GracefulDeath::around static method. This static method returns an instance of a builder that let you configure how GracefulDeath will behave. With afterViolentDeath you can configure an handler that will be called whenever a fatal error is triggered. The handler could be

• An integer: the process will terminate with this integer as status code
• A string: the string will be printed on standard output (like in this example)
• A closure: the closure will be executed and its return value will be used as return value of the run method. The closure signature is function($status, $stdout, $stderr) where $status is the exit status of the code passed to GracefulDeath::around

There are a few other method that can be used to configure GracefulDeath

• afterNaturalDeath: like afterViolentDeath but used to configure an handler that will be called only when no errors are triggered

• afterDeath: used to configure an handler (like afterViolentDeath and afterNaturalDeath) that will be called after the code passed to GracefulDeath::around is terminated

• reanimationPolicy: used to configure the reanimation policy aka something that will be used to decide if the code passed to GracefulDeath::around should be executed again after a fatal error. The reanimation policy could be

  • A boolean: if true it will retry forever, if false it will never retry (default)
  • An integer: the number of times the code will be executed. The code will not be execute again if either the code terminates without error or the number of executions exceeds the number passed as argument
  • A closure: if the closure returns true the code will be executed again. The closure signature is function($status, $attempts, $stdout, $stderr)
    • $status: the exit status of the code passed to GracefulDeath::around
    • $attempts: how many times the code has been executed (think how many previous lives), starts at 1
    • $stdout: what the code passed to GracefulDeath::around printed on stdout
    • $stderr: what the code passed to GracefulDeath::around printed on stderr

  There are a few reanimation policies ready to be used

  • doNotReanimate: this is the default behaviour so not so useful to use but could be good for documentation or to make the code more explicit
  • giveMeAnotherChance: it will reanimate the child process only one time
  • liveForever: it will always reanimate the child process. If you use this reanimation policy don't forget also to use avoidFutileMedicalCare to avoid to continually reanimate a child process that is broken

• doNotCaptureOutput: avoid to capture stdout and stderr. Note that if output is not captured then it could not be given to the reanimationPolicy closure

• doNotEchoOutput: discard the (if any) captured output

• avoidFutileMedicalCare($numberOfFailures, $inAmountOfTime): avoid to reanimate a process that died too many times ($numberOfFailures default to 6) in a small amount of time ($inAmountOfTime default 60 seconds)

For all the options and methods look at the examples or at the tests 😄

Use Cases

• You have a long running process that leaks memory (see examples/memory_leak.php)
• You can use graceful-death as a supervisor, for that use ->liveForever() to always reanimate the child process and don't forget to use ->avoidFutileMedicalCare() to avoid to continually reanimate a child process that is broken

How Does It Work?

When run is called the process forks, the child process will execute the code passed to GracefulDeath::around, the parent process will act as a supervisor of the child. The supervisor will wait until the child dies and will act accordingly to the exit status and the given configuration.

Gotcha

It only works where pcntl_* function are available.

Self-Promotion

If you like this project, then consider to:

• Star the repository on GitHub
• Follow me on
  • Twitter
  • GitHub