include() best practices

[asa_carter]

Hi

I have a questions related to the best practice of using include files.

1. What is the maximum number of include files that should be used on a page before server perfomance is affected?

2. I have read that require_once should be used instead of include. Why? Doesnt require exit the script on failure? So would inlclude not be better?

3. Would it be more efficient to put a lot of commonly used functions such as database connection, retrieving user details etc in one include file and then calling the functions instead of having each in separate include and then calling the include?

[AsaSchool]

I am not sure there is a number to how many includes can be used. I have pages where I am calling 5 - 8+ scripts to be included on one page. I do this because they are scripts that are used on more than one page (ie; navigation, features, etc..) and if I need to update anything on one of them, I only have to change one file instead of numerous files located across my websites. I have not noticed any change in performance or load-time with doing this and my sites are easier to update and make changes doing it this way.

If someone has any more information on this, I would also be interested as well. My experience in with this though is that it will not effect performance if the scripts are written well.

[asa_carter]

Yeh this what I'm also doing at the moment but I was thinking that in theory, If you had a lot of concurrent users, the server is having to keep fetching files and it would be far more efficient to only fetch one or two include files and much quicker to call a function than load a file.

[NogDog]

On any but the most overworked and/or underpowered server, you probably won't notice much of a difference. Any decent server should have a good sized disk cache, so if you are truly re-accessing the same files over and over, hopefully most are being accessed from RAM and not from the disk itself. In any case, with modern disk drives the seek and read times are awfully fast now.

But certainly it would not be to difficult to run some tests using [man]microtime/man to track how long it takes to include multiple small files versus a single large file, then do multiple test runs to see if you can detect any significant trends worth worrying about. (Most of us are not normally working on applications where a script that takes a few more milliseconds to run is ever going to make a difference, after all.)

[asa_carter]

Well that's put my mind at rest. What about require_once over include?

[NogDog]

The difference between require and include is only what happens if the file cannot be included, so from that respect, there should normally not really be any difference in performance. The "once" part will likely add a little time (we're probably down in the microseconds range here) since it needs to check if the file has already been included. But assuming there is a reason to use the "once" (which there often is), once again it's something not to lose sleep over unless in reality you really need to save every possible nanosecond. And frankly, if that's the situation you're in, you probably should not be creating the application in PHP in the first place.

[jazz_snob]

asa_carter wrote:

I have read that require_once should be used instead of include. Why? Doesnt require exit the script on failure? So would inlclude not be better?

Typically, if I'm including something I'm really requiring to reuse some functions. And since my code is depending on those function definitions being loaded then I'd rather have a fatal error if they cannot be found right away.

asa_carter wrote:

Would it be more efficient to put a lot of commonly used functions such as database connection, retrieving user details etc in one include file and then calling the functions instead of having each in separate include and then calling the include?

The granularity of release is the granularity of reuse. So if you have lets say you have 5 database functions that you've honed over the years and really want to use in every project cause they make your life simple. The optimum approach is to keep them in their own file so they can evolve on their own. To do that effectively the file (library) would be under version control (eg: CVS). That way your DB functions and your templating functions and image manipulation functions, etc, can all evolve independently. Moreover, I don't think having extra files hurts performance...php is still gonna churn over the same number of lines of code either way.

So now comes the time to code your app and you don't want 20 require_once statements at the top of every page. Well, there are lots of ways around that, but what I typically do is have an init.php file that contains nothing but require_once() statements that will load all "common" files I will need on each request, so wherever I need them, I just require_once('init.php');

[NogDog]

Also, if using OOP in PHP5, the [man]autoload/man function can save you from a having to type a lot of require_once() commands for loading class files (assuming you have rationalized your directory structure sufficiently to make autoload() reasonably usable).

[Weedpacket]

jazz_snob wrote:

Typically, if I'm including something I'm really requiring to reuse some functions. And since my code is depending on those function definitions being loaded then I'd rather have a fatal error if they cannot be found right away.

And to add to that: since you aren't allowed to declare a function more than once, you want to make sure that the required file is only loaded once. It's safer and more reliable to have the runtime keep track of which files have already been loaded and which haven't, and simply make sure that each file that is loaded requires exactly those files it directly requires (and if those files have any further dependencies, let them say what they are).