We have PHP code in production that sometimes fails with "Call to member function on null", although the same code path executes fine several times before that in one invocation. We have a test that reproduces the error consistently at the same run of the loop.
I already proved that the object gets created correctly in the factory even if it gets returned as null. The factory method must not return null in any case, as indicated in the DocBlock. This question is not related to nullable return types or something like that.
The process does not exceed memory or runtime limitations and I already tried turning off the garbage collector, but no luck. The error happens both in PHP 7.0 and 7.3 on Debian, did not try on other versions or operating systems.
I am not allowed to paste the real code here, but I wrote a simple mockup to explain in more detail. Please keep in mind that this demo code will not result in the error, it is just meant to show the general structure of the program that runs into this fault.
// Three placeholder classes with common methods
class Bender
{
public function common()
{
echo "Bend, bend!" . PHP_EOL;
}
}
class Clamper
{
public function common()
{
echo "Clamp, clamp!" . PHP_EOL;
}
}
class Worker
{
public function common()
{
echo "Work, work!" . PHP_EOL;
}
}
// abstract class with static factory to produce objects
abstract class MomCorp
{
/**
* Factory to create one of several objects
*
* #param string $name
* #return Bender|Clamper|Worker
*/
public static function factory($name)
{
$type = self::managementDecision($name);
switch ($type)
{
case "bender":
$robot = new Bender();
break;
case "clamper":
$robot = new Clamper();
break;
default:
$robot = new Worker();
}
// optional QA works flawlessly here, object is fine all the time!
// $robot->common();
return $robot;
}
public static function managementDecision($name)
{
// irrelevant magic happens on $name here
return "bender";
}
}
foreach (['Rodriguez', 'Angle-ine', 'Flexo', 'Billie'] as $newName)
{
echo "$newName: ";
// The next two lines break after some loops - why?
// The perfectly functional object in factory gets returned as null
$bot = MomCorp::factory($newName);
$bot->common();
}
// SAMPLE OUTPUT
// Rodriguez: Bend, bend!
// Angle-ine: Bend, bend!
// Flexo: Bend, bend!
// Billie: Call to a member function common() on null
Has anyone experienced the same and has any hints on what might cause such an error and how to fix it?
I'm attempting to create a wrapper class around the mikehaertl\php-pdftk\pdf object for the purposes of populating PDF form fields. When trying to chain commands via the documentation the pdf fails to correctly execute the second command (or any after the first). It looks as though this is an issue with the underlying temp file handling and the tmep file not being written out as I watch my temp folder. As I debug, the temp file is there, but of 0 size.
Sample code demonstrating the issue
use mikehaertl\pdftk\Pdf;
class PDFTKTest extends TestCase
{
public function testPdfTkOperations()
{
$cmdPath = 'D:\PDFtk\bin\pdftk.exe';
$formPath = 'D:\test\sample_files\test.pdf';
$options = ['command' => $cmdPath];
$pdf = new Pdf($formPath, $options);
$this->assertNotNull($pdf);
//Get fields from PDF
$fields = $pdf->getDataFields();
$this->assertNotNull($fields);
//Set some field Values
$values = ['full_name' => 'John Q. Programmer'];
$pdf2 = new Pdf($pdf, $options); //chaining broken
//$pdf2 = new Pdf($formPath, $options); //works fine creating a new Pdf object
$this->assertNotNull($pdf2);
$res = $pdf2->fillForm($values)->execute();
//Next assertion fails using chaining
$this->assertTrue($res, "Execute failed: \n". $pdf2->getError());
//Get fields with the updates
$fields = $pdf2->getDataFields();
$this->assertNotNull($fields);
//Next assertion fails, getDataFields fails on a chained command
$this->assertGreaterThan(0, count($fields));
}
}
I've got a work around where I use separate \Pdf objects for each action and manage my own temp file, I was just hoping to take advantage of the classes functionality a bit more and not have to do so much of the mundane. Is this functionality broken, or am I using it incorrectly?
After looking deeper in to the PDFTK library which mikehaertl\php-pdftk\pdf wraps and reading the documentation on the dump_data_fields option I came up with the folowing observations:
PDFTK doesn't produce an output file for the dump_data_fields command
The php-pdftk class does create the underlying temp file when calling getDataFields, but it is empty and remains that way.
When chaining another Pdf object, it references the empty temp file from the previous command. Here lies the rub.
Solution
When I call getFieldData I create a new Pdf object and chain it to the previous, however I don't save a reference to that. I only save the newly chained object if it is form a command that creates actual output.
Here's an exmaple to demonstate:
<?php
use mikehaertl\pdftk\Pdf;
class PDFTKFormService
{
protected $pdf = null;
/**
* #return array|bool|\mikehaertl\pdftk\DataFields
*/
public function getDataFields()
{
//get data fields doesn't output a new file
//so we need to use the existing instance or create a new one and
$pdf = $this->getNextPdf();
$fields = $pdf->getDataFields();
if ($fields === false)
return [];
return $fields;
}
/**
* #param array $data
*
* #return resource The stream resource
*/
public function setDataFieldValues($data = [])
{
$this->pdf = $this->getNextPdf();
$this->pdf->fillForm($data)->execute();
}
protected function getNextPdf()
{
$options = ['command' => 'Path\To\PDFTK\binary'];
if ($this->pdf === null) {
return new Pdf($this->getTemplatePath(), $options);
} else {
return new Pdf($this->pdf, $options);
}
}
}
Hopefully this can help someone else.
I've been stuck on this for a while and I'm not sure why PHPunit can't see that the function is being called.
This is the code I'm trying to test:
public function handle()
{
$path = $this->request->getPath();
$requestMethod = $this->request->getMethod();
if (!$path) {
$this->redirect('home');
} else if (!$this->isMethodPathFound($path, $requestMethod)) {
$this->redirect('404');
} else {
$handler = $this->getControllerFullName($this->routes[$path]['handler']);
if (is_callable($handler)) {
call_user_func($handler);
} else {
$this->redirect('404');
}
}
}
/**
* #param string $path
* #param int $statusCode
*/
public function redirect($path, $statusCode = 303)
{
if (defined('TESTING_ENVIRONMENT') && TESTING_ENVIRONMENT) {
return;
}
header(
'Location: ' . $this->request->getProtocol() .
$this->request->getHost() . '/' . $path,
true,
$statusCode
);
die();
}
The TESTING_ENVIRONMENT variable is set for the header function so it does not trigger on running PHPunit (I don't want to create another class to have that redirect function just to be able to mock it for one test) and this is the testing code:
public function testHandlePathIsEmpty()
{
$requestMock = $this->getMockBuilder('\services\Request')->getMock();
$requestMock->expects($this->once())->method('getPath')->willReturn('');
$requestMock->expects($this->once())->method('getMethod')->willReturn('GET');
$routerMock = $this->getMockBuilder('\services\Router')
->setConstructorArgs([$this->routes, $requestMock])
->enableProxyingToOriginalMethods()
->getMock();
$routerMock->expects($this->once())->method('redirect')
->with('asdasd')->willReturn(true);
$routerMock->handle();
}
The $routerMock object should definitely invoke the "redirect" function, and it says that it does not get invoked..even though when I var_dump/die inside the function, it does go inside of it.
Thanks for the help!
Though you hesitated to show the complete output of phpunit's error, your problem is very likely not that your method is not called, but that it is not called with all the expectations you defined.
Your code
$routerMock->expects($this->once())->method('redirect')
->with('asdasd')->willReturn(true);
translates to the following expectations: The method redirect must be called exactly once with an argument 'asdasd' and will return true.
From your testcode I do not see that there is asdasd passed to the redirect method. Your test will most likely succeed when you remove the with expectation.
Just to make this clear. If you have to mock the class u want to test, your code is way to complex and you should think about implementing your logic in another way.
How about not mocking the class you are actually testing, create the new instance by passing the Request and a Router Mock (Router mock might not have any logic since you are not going to use it) and then do the following in your code:
public function handle()
{
$request = $this->request;
$path = $request->getPath();
if (!$path) {
$this->redirect('home');
} else if (!$this->isMethodPathFound($path, $request->getMethod())) {
$this->redirect('404');
} else {
$handler = $this->getControllerFullName($this->routes[$path]['handler']);
if (is_callable($handler)) {
call_user_func($handler);
} else {
$this->redirect('404');
}
}
}
In your Unit-Test, you now can just test for
$requestMock
->expects($this->never())
->method('getMethod');
I see that this would only cover the second case to not being executed but the third one could happen aswell. Thats always a point why your code is not clean enough.
You should read something about KISS and SOLID to make your code more testable. This method is just too complex as you could test it correctly.
I've been using Codeigniter for the past two years and really have become a big fan, but over the past year I've found myself writing more and more javascript than PHP.
In the begining, I would write everything with PHP, but now I find myself using $.ajax all the time. And I sort of feel like Im repeating myself between javascript and php.
I know that CI does give you some good control over ajax, but Im still having two write a ton of javascript and I'd like to consolidate if at all possible.
I guess what I am looking for is a php framework that integrates tightly with jQuery's $.ajax.
I use this piece of code in Javascript. Backend wise things are organized in a MVC type of organisation, so things affecting one module are usually grouped together. In general I also create a sperate module for a seperate model, but in some cases you may deviate from this principle.
My setup is with symfony at the back and plain jquery at the front. There are some approaches that automatize this part, like http://javascriptmvc.com/, I find it too restricting in many parts. Here is my workflow for integrating php and jquery.
PHP
Execute a piece of code and wrap it inside a try/catch block. This way error messages may be propagated to the frontend. This method helps in that regard to convert exceptions to a readable error. (to debug from json).
try {
//... execute code .. go about your buisness..
$this->result = "Moved " . count($files) . " files ";
// result can be anything that can be serialized by json_encode()
} catch (Exception $e) {
$this->error = $e->getMessage() . ' l: ' . $e->getLine() . ' f:' . $e->getFile();
// return an error message if there is an exception. Also throw exceptions yourself to make your life easier.
}
// json response basically does something like echo json_encode(array("error" => $this->error, "result" => $this->result))
return $this->jsonResponse();
For error handling I often use this to parse errors.
public function parseException($e) {
$result = 'Exception: "';
$result .= $e->getMessage();
$trace = $e->getTrace();
foreach (range(0, 10) as $i) {
$result .= '" # ';
if (!isset($trace[$i])) {
break;
}
if (isset($trace[$i]['class'])) {
$result .= $trace[$i]['class'];
$result .= '->';
}
$result .= $trace[$i]['function'];
$result .= '(); ';
$result .= $e->getFile() . ':' . $e->getLine() . "\n\n";
}
return $result;
}
Javascript side
/**
* doRequest in an ajax development tool to quickly execute data posts.
* #requires jQuery.log
* #param action (string): url for the action to be called. in config.action the prefix for the url can be set
* #param data (object): data to be send. eg. {'id':5, 'attr':'value'}
* #param successCallback (function): callback function to be executed when response is success
* #param errorCallback (function): callback function to be executed when response is success
*/
jQuery.doRequest = function (action, data, successCallback, errorCallback) {
if (typeof(successCallback) == "undefined") {
successCallback = function(){};
}
if (typeof(errorCallback) == "undefined") {
errorCallback = function(data ){
alert(data.error);
};
}
jQuery.log(action);
jQuery.post(action, data, function (data, status)
{
jQuery.log(data);
jQuery.log(status);
if (data.error !== null || status != 'success') {
// error handler
errorCallback(data);
} else {
successCallback(data);
}
},'json');
};
Note: the error callbacks are very nice if you combine them with something like pNotify
Look into Agile Toolkit, which is a PHP UI Framework. UI means it takes care of HTML, JavaScript, CSS and AJAX while allowing you to develop in plain, object-oriented PHP language.
http://agiletoolkit.org/intro/javascript
There is also a blog post comparing it with CodeIgniter: http://agiletoolkit.org/blog/agile-toolkit-for-codeigniter-developer/
p.s. I'm co-author for Agile Toolkit.
I am starting a new web application in PHP and this time around I want to create something that people can extend by using a plugin interface.
How does one go about writing 'hooks' into their code so that plugins can attach to specific events?
You could use an Observer pattern. A simple functional way to accomplish this:
<?php
/** Plugin system **/
$listeners = array();
/* Create an entry point for plugins */
function hook() {
global $listeners;
$num_args = func_num_args();
$args = func_get_args();
if($num_args < 2)
trigger_error("Insufficient arguments", E_USER_ERROR);
// Hook name should always be first argument
$hook_name = array_shift($args);
if(!isset($listeners[$hook_name]))
return; // No plugins have registered this hook
foreach($listeners[$hook_name] as $func) {
$args = $func($args);
}
return $args;
}
/* Attach a function to a hook */
function add_listener($hook, $function_name) {
global $listeners;
$listeners[$hook][] = $function_name;
}
/////////////////////////
/** Sample Plugin **/
add_listener('a_b', 'my_plugin_func1');
add_listener('str', 'my_plugin_func2');
function my_plugin_func1($args) {
return array(4, 5);
}
function my_plugin_func2($args) {
return str_replace('sample', 'CRAZY', $args[0]);
}
/////////////////////////
/** Sample Application **/
$a = 1;
$b = 2;
list($a, $b) = hook('a_b', $a, $b);
$str = "This is my sample application\n";
$str .= "$a + $b = ".($a+$b)."\n";
$str .= "$a * $b = ".($a*$b)."\n";
$str = hook('str', $str);
echo $str;
?>
Output:
This is my CRAZY application
4 + 5 = 9
4 * 5 = 20
Notes:
For this example source code, you must declare all your plugins before the actual source code that you want to be extendable. I've included an example of how to handle single or multiple values being passed to the plugin. The hardest part of this is writing the actual documentation which lists what arguments get passed to each hook.
This is just one method of accomplishing a plugin system in PHP. There are better alternatives, I suggest you check out the WordPress Documentation for more information.
So let's say you don't want the Observer pattern because it requires that you change your class methods to handle the task of listening, and want something generic. And let's say you don't want to use extends inheritance because you may already be inheriting in your class from some other class. Wouldn't it be great to have a generic way to make any class pluggable without much effort? Here's how:
<?php
////////////////////
// PART 1
////////////////////
class Plugin {
private $_RefObject;
private $_Class = '';
public function __construct(&$RefObject) {
$this->_Class = get_class(&$RefObject);
$this->_RefObject = $RefObject;
}
public function __set($sProperty,$mixed) {
$sPlugin = $this->_Class . '_' . $sProperty . '_setEvent';
if (is_callable($sPlugin)) {
$mixed = call_user_func_array($sPlugin, $mixed);
}
$this->_RefObject->$sProperty = $mixed;
}
public function __get($sProperty) {
$asItems = (array) $this->_RefObject;
$mixed = $asItems[$sProperty];
$sPlugin = $this->_Class . '_' . $sProperty . '_getEvent';
if (is_callable($sPlugin)) {
$mixed = call_user_func_array($sPlugin, $mixed);
}
return $mixed;
}
public function __call($sMethod,$mixed) {
$sPlugin = $this->_Class . '_' . $sMethod . '_beforeEvent';
if (is_callable($sPlugin)) {
$mixed = call_user_func_array($sPlugin, $mixed);
}
if ($mixed != 'BLOCK_EVENT') {
call_user_func_array(array(&$this->_RefObject, $sMethod), $mixed);
$sPlugin = $this->_Class . '_' . $sMethod . '_afterEvent';
if (is_callable($sPlugin)) {
call_user_func_array($sPlugin, $mixed);
}
}
}
} //end class Plugin
class Pluggable extends Plugin {
} //end class Pluggable
////////////////////
// PART 2
////////////////////
class Dog {
public $Name = '';
public function bark(&$sHow) {
echo "$sHow<br />\n";
}
public function sayName() {
echo "<br />\nMy Name is: " . $this->Name . "<br />\n";
}
} //end class Dog
$Dog = new Dog();
////////////////////
// PART 3
////////////////////
$PDog = new Pluggable($Dog);
function Dog_bark_beforeEvent(&$mixed) {
$mixed = 'Woof'; // Override saying 'meow' with 'Woof'
//$mixed = 'BLOCK_EVENT'; // if you want to block the event
return $mixed;
}
function Dog_bark_afterEvent(&$mixed) {
echo $mixed; // show the override
}
function Dog_Name_setEvent(&$mixed) {
$mixed = 'Coco'; // override 'Fido' with 'Coco'
return $mixed;
}
function Dog_Name_getEvent(&$mixed) {
$mixed = 'Different'; // override 'Coco' with 'Different'
return $mixed;
}
////////////////////
// PART 4
////////////////////
$PDog->Name = 'Fido';
$PDog->Bark('meow');
$PDog->SayName();
echo 'My New Name is: ' . $PDog->Name;
In Part 1, that's what you might include with a require_once() call at the top of your PHP script. It loads the classes to make something pluggable.
In Part 2, that's where we load a class. Note I didn't have to do anything special to the class, which is significantly different than the Observer pattern.
In Part 3, that's where we switch our class around into being "pluggable" (that is, supports plugins that let us override class methods and properties). So, for instance, if you have a web app, you might have a plugin registry, and you could activate plugins here. Notice also the Dog_bark_beforeEvent() function. If I set $mixed = 'BLOCK_EVENT' before the return statement, it will block the dog from barking and would also block the Dog_bark_afterEvent because there wouldn't be any event.
In Part 4, that's the normal operation code, but notice that what you might think would run does not run like that at all. For instance, the dog does not announce it's name as 'Fido', but 'Coco'. The dog does not say 'meow', but 'Woof'. And when you want to look at the dog's name afterwards, you find it is 'Different' instead of 'Coco'. All those overrides were provided in Part 3.
So how does this work? Well, let's rule out eval() (which everyone says is "evil") and rule out that it's not an Observer pattern. So, the way it works is the sneaky empty class called Pluggable, which does not contain the methods and properties used by the Dog class. Thus, since that occurs, the magic methods will engage for us. That's why in parts 3 and 4 we mess with the object derived from the Pluggable class, not the Dog class itself. Instead, we let the Plugin class do the "touching" on the Dog object for us. (If that's some kind of design pattern I don't know about -- please let me know.)
The hook and listener method is the most commonly used, but there are other things you can do. Depending on the size of your app, and who your going to allow see the code (is this going to be a FOSS script, or something in house) will influence greatly how you want to allow plugins.
kdeloach has a nice example, but his implementation and hook function is a little unsafe. I would ask for you to give more information of the nature of php app your writing, And how you see plugins fitting in.
+1 to kdeloach from me.
Here is an approach I've used, it's an attempt to copy from Qt signals/slots mechanism, a kind of Observer pattern.
Objects can emit signals.
Every signal has an ID in the system - it's composed by sender's id + object name
Every signal can be binded to the receivers, which simply is a "callable"
You use a bus class to pass the signals to anybody interested in receiving them
When something happens, you "send" a signal.
Below is and example implementation
<?php
class SignalsHandler {
/**
* hash of senders/signals to slots
*
* #var array
*/
private static $connections = array();
/**
* current sender
*
* #var class|object
*/
private static $sender;
/**
* connects an object/signal with a slot
*
* #param class|object $sender
* #param string $signal
* #param callable $slot
*/
public static function connect($sender, $signal, $slot) {
if (is_object($sender)) {
self::$connections[spl_object_hash($sender)][$signal][] = $slot;
}
else {
self::$connections[md5($sender)][$signal][] = $slot;
}
}
/**
* sends a signal, so all connected slots are called
*
* #param class|object $sender
* #param string $signal
* #param array $params
*/
public static function signal($sender, $signal, $params = array()) {
self::$sender = $sender;
if (is_object($sender)) {
if ( ! isset(self::$connections[spl_object_hash($sender)][$signal])) {
return;
}
foreach (self::$connections[spl_object_hash($sender)][$signal] as $slot) {
call_user_func_array($slot, (array)$params);
}
}
else {
if ( ! isset(self::$connections[md5($sender)][$signal])) {
return;
}
foreach (self::$connections[md5($sender)][$signal] as $slot) {
call_user_func_array($slot, (array)$params);
}
}
self::$sender = null;
}
/**
* returns a current signal sender
*
* #return class|object
*/
public static function sender() {
return self::$sender;
}
}
class User {
public function login() {
/**
* try to login
*/
if ( ! $logged ) {
SignalsHandler::signal(this, 'loginFailed', 'login failed - username not valid' );
}
}
}
class App {
public static function onFailedLogin($message) {
print $message;
}
}
$user = new User();
SignalsHandler::connect($user, 'loginFailed', array($Log, 'writeLog'));
SignalsHandler::connect($user, 'loginFailed', array('App', 'onFailedLogin'));
$user->login();
?>
I believe the easiest way would be to follow Jeff's own advice and have a look around the existing code. Try looking at WordPress, Drupal, Joomla, and other well-known PHP-based CMS to see how their API hooks look and feel. This way you can even get ideas you may have not thought of previously to make things a little more robust.
A more direct answer would be to write general files that they would "include_once" into their file that would provide the usability they would need. This would be broken up into categories and NOT provided in one MASSIVE "hooks.php" file. Be careful though, because what ends up happening is that files that they include end up having more and more dependencies and functionality improves. Try to keep API dependencies low. I.E fewer files for them to include.
There's a neat project called Stickleback by Matt Zandstra at Yahoo that handles much of the work for handling plugins in PHP.
It enforces the interface of a plugin class, supports a command line interface and isn't too hard to get up and running - especially if you read the cover story about it in the PHP architect magazine.
Good advice is to look how other projects have done it. Many call for having plugins installed and their "name" registered for services (like wordpress does) so you have "points" in your code where you call a function that identifies registered listeners and executes them. A standard OO design patter is the Observer Pattern, which would be a good option to implement in a truly object oriented PHP system.
The Zend Framework makes use of many hooking methods, and is very nicely architected. That would be a good system to look at.
I am surprised that most of the answers here seem to be geared about plugins that are local to the web application, ie, plugins that run on the local web server.
What about if you wanted the plugins to run on a different - remote - server? The best way to do this would be to provide a form that allows you to define different URLs that would be called when particular events occur in your application.
Different events would send different information based on the event that just occurred.
This way, you would just perform a cURL call to the URL that has been provided to your application (eg over https) where remote servers can perform tasks based on information that has been sent by your application.
This provides two benefits:
You don't have to host any code on your local server (security)
The code can be on remote servers (extensibility) in different languages other then PHP (portability)