Using the spl_autoload_register(), should I use a single autoload function that contains all the logic to determine where the include files are or should I break each include grouping into it's own function with it's own logic to include the files for the called function? As the places where include files may reside expands so too will the logic of a single function. If I break it into multiple functions I can add functions as new groupings are added, but the functions will be copy/pastes of each other with minor alterations.

Currently I have a tool with a single registered autoload function that picks apart the class name and tries to predict where it is and then includes it. Due to naming conventions for the project this has been pretty simple.

if has namespace 
  if in template namespace look in Root\Templates
  else look in Root\Modules\Namespace
  look in Root\System
if file exists include

But we are starting to include Interfaces and Traits into our codebase and it hurts me to include the type of a thing in it's name. So we are looking at instead of a single autoload function that digs through the class name and looks for the file and has increasingly complex logic to it, we are looking at having multiple autoload functions registered. But each one follows the same pattern and any time I see that I get paranoid about code copying.

function systemAutoloadFunc
  logic to create probable filename
  if filename exists in system include it and return true
  else return false

function moduleAutoloadFunc
  logic to create probable filename
  if filename exists in modules include it and return true
  else return false

Every autoload function will follow that pattern and the last of each function if filename exists, include return true else return false is going to be identical code. This makes me paranoid about having to update it later across the board if the file_exists include pattern we are using ever changes. Or is it just that, paranoia and the multiple functions with some identical code is the best option?

  • I have a feeling that you're trying to reinvent the bicycle. Did you look at composer's autoload generator? It is VERY efficient
    – JimiDini
    Aug 17, 2014 at 12:28
  • Gotcha, what I am asking is what are the cost/benefits of having multiple registered autoload functions, each specifically tailored to the place/type of object in my tool, or a single autoload function that I add more special cases as expand the structure of my tool. Aug 18, 2014 at 17:22
  • Hope that is a better intro to the question. Aug 18, 2014 at 22:43

1 Answer 1


Registering multiple autoloaders lets you separate different naming conventions for directory and file names and will be easier to maintain in the long run.

Imagine needing to add or remove functionality for the autoloader. Should you need to dive into a monolithic function or simply tweak or add a smaller function?

Similar calls to "file_exists() / include() / return" can be put into a helper function instead of copypasting that code.

function autoload_attempt($path){
    if (file_exists($path)) {
        include $path;
        return true;
    } else return false;

    $path = ... naming logic ...
    return autoload_attempt($path);

Registering different ones also gives you easier control over what order they are invoked. For example, if you want to have a master override directory, you would register that first.

From the manual: http://php.net/manual/en/function.spl-autoload-register.php

If there must be multiple autoload functions, spl_autoload_register() allows for this. It effectively creates a queue of autoload functions, and runs through each of them in the order they are defined. By contrast, __autoload() may only be defined once.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.