How to use multi-instance in PW 3.x

ProcessWire 3.0.30 and 2.8.30

This week work continued on preparing 3.x (and 2.8.x) for release. One of the features of 3.x that we'd not yet covered in much detail was the multi-instance support. So the primary focus this week was in making sure we clarified and simplified some things in that respect. This post covers all the details. In addition we've also got some $session updates that we think you'll like!

Using multi-instance in PW3

One of the things new to ProcessWire 3.x is multi-instance support. What this means is that you can create multiple instances of ProcessWire from the same code. You can pull or manipulate data from one instance into another and vice versa. This is particularly useful when you've got multiple ProcessWire installations in the same environment and you want them to be able to share data. It's a simpler, more efficient and powerful alternative to using feeds and web services (though obviously less portable).

Usage environment

When using multi-instance in a web environment, typically you'll have a master instance that is serving a page, and the template file [used by that page] that creates another instance of ProcessWire (pulling from another local installation). For instance, on our server we have the main site (you are here), and we've got 4 other ProcessWire installations for the modules site, directory site, cheatsheet site, and demo site.

Prior to multi-instance support, the only way these sites could share data was by pulling from feeds (like RSS feeds), and indeed we've done quite a bit of that. But with multi-instance support, any one of these installations can boot up another and have access to the entire ProcessWire API of that installation (where file permissions permit).

Another environment for multi-instance is from a PHP script (or shell script) running on its own. It can create and connect to as many ProcessWire installations as are visible to it. The same goes for instantiating ProcessWire from another software (like another CMS). In this post, we'll primarily focus on the environment described by the previous paragraph. However, note that everything mentioned here applies to either environment. The primary difference is that if your environment isn't already within ProcessWire, then you would need to include/require ProcessWire's core file first, like this:

require('/path/to/wire/core/ProcessWire.php'); 

You also have the option of letting the Composer autoloader handle it instead. PW3 isn't yet up on Packagist, but will be after it's in master release.

How to create a new instance

To connect to and create a new instance of ProcessWire, you need to know where it is located on the file system. The directory you provide can either be to the root of the installation (where PW's index.php file resides) or to the /site/ directory of the installation (where the config.php file resides). That's all you need to instantiate that site:

$site = new ProcessWire('/path/to/www/');

While optional, we also recommend providing the URL to the installation as the second argument (preferably including the scheme and hostname). That way if you make any $page->url() or $page->httpUrl() calls, it will know where to point them to:

$site = new ProcessWire('/path/to/www/', 'http://domain.com/'); 

Once you've got your ProcessWire instance ($site), you can access the entire API of that ProcessWire installation. For example:

$newestPages = $site->pages->find("sort=-created, limit=3"); 

That's a simple example above (finding the 3 newest pages), but that's just to communicate that "you have access to the API". Meaning, you have access to the entire API of that installation. But rather than accessing an API variable such as $pages via $pages or wire('pages'), you access it through $site->pages, where $site can be whatever variable name you've assigned the instance to.

A couple of things to note if you are booting a ProcessWire installation from a non-ProcessWire installation. First would be that you'll need to include/require the core ProcessWire.php file (as described in the previous section). Next would be that, assuming you aren't already in the ProcessWire namespace, you'd need to identify that namespace when creating the new instance, i.e.

$site = new \ProcessWire\ProcessWire('/path/to/site/');

Live example

The following data is being pulled directly from modules.processwire.com (a separate ProcessWire installation) using multi-instance, every time this page is rendered. This examples lists the 5 newest modules added to the directory. Following this example, you'll see the actual code used to deliver it.

newest-modules

Code used in the live example

Below is the exact code used to create the new ProcessWire instance (to modules.processwire.com) and produce the output shown above. This code appears in the blog-post template that is rendering this page you are reading now.

// Server path to the PW installation
$path = '/home/modules/www/';

// The root URL for the PW installation
$url = 'http://modules.processwire.com/';

// Create a new ProcessWire instance
$site = new ProcessWire($path, $url);

// Find the 5 newest items in modules directory
$items = $site->pages->find("parent=/modules/, limit=5, sort=-created");

// output list
foreach($items as $item) {
  echo "
    <p>
      <a href='$item->httpUrl'>$item->title</a><br />
      $item->version <em>by $item->author</em><br />
      $item->summary
    </p>
    ";
}

Pitfalls and considerations

Multi-instance support has not seen much broad use yet, so we'll be keeping this particular feature in beta even after 3.x is officially released. If you opt to use multi-instance support, please keep the following in mind:

  • We strongly suggest you test things out in a development environment before putting anything to use in a live environment. Be especially careful with anything that manipulates data in a multi-instance environment. Check and double check that it's all working how you expect.

  • While the ProcessWire core is aware of multi-instance, most 3rd party modules are not. Meaning, there is potential for problems there, though it probably doesn't matter to most modules. While we've not yet observed any issues, it's something to keep an eye out for.

  • Ideally all instances should be running the same exact version of ProcessWire if possible. At minimum, the versions must be close.

  • Keep an eye out for issues when multiple installations have very different sets of modules installed.

  • Be careful not to load objects from one instance and save them to another. Loaded objects are going to be connected with the instance they were loaded from. They might have a database ID unique to that installation, and so on. If you need to copy/move object data from one instance to save to another, convert to basic types and re-create the objects. For instance, if copying an image from one instance to another, do a $page->images->add($filename); rather than trying to add the Pagefile object from another installation.

  • Multi-instance is intended only for 3.x versions of ProcessWire. You can't create multiple instances of ProcessWire 2.7. While it may be possible to boot a 2.8.x instance from another 2.8.x instance, we aren't going to be officially supporting multi-instance on the 2.8.x version due to the lack of namespaces.

As you can see, there are a lot of potential pitfalls, so use care with multi-instance. However, we suspect many will be using multi-instance simply as a way to pull data in a read-only manner, and in that context you shouldn't have to worry about much.

Future multi-instance goals

Currently multi-instance requires that the PW installation are local and accessible on the file system. In the future, we'd like to extend this so that multi-instance can work over http in some respect. Meaning, you could make two completely independent PW installations on different servers talk with each other via the ProcessWire API. For instance, imagine if I could do this from here on processwire.com:

$tripsite = new ProcessWire("https://www.tripsite.com");
$tours = $tripsite->find("template=tour, country=Austria"); 

Currently to do stuff like this, I usually setup JSON web services and feeds. While it's not hard to do, it can be time consuming. If I could just make two PW installations talk to each other via the API instead, it would be a game changer. We're a long way off from this still, but it's fun to dream about.

More session control

Something small (or large, depending on your point of view) that was added this week is the ability to disable sessions. ProcessWire has always maintained sessions when delivering pages, which makes life simple for us. If we want to get or set some session data, we just use the $session API var and everything is ready to go.

But there are cases where maybe you don't want sessions to be active. For instance, sessions require a cookie on the client side, and maybe you don't want to set a cookie because of some law that says you have to have user consent. Or maybe your site is simple and you have absolutely no use for sessions. Or maybe some other reason. Whatever the reason, it's now supported in ProcessWire 3.0.30 via the $config->sessionAllow configuration property.

This configuration property may be set to a callable function that returns boolean (true or false) as to whether or not to allow sessions for the current request. Initializing the session is one of the first things that ProcessWire does, before determining the current page or user. So the callback function has to rely on what can be determined without the ProcessWire API, in order to determine whether a session is allowed or not.

Usage and example

Below is an example of using the $config->sessionAllow property. This example would go in your /site/config.php file. In this example, we allow sessions if a session cookie is already present, or if the requested URL looks like an admin URL. Otherwise we disallow sessions. This would have the effect of disabling sessions for an entire site, except for admin URLs, or for users that are already logged in. This callback function is provided a single argument, which is the $session API variable.

$config->sessionAllow = function($session) {

  // if there is a session cookie, chances are user is logged in
  if($session->hasCookie()) {
    return true;
  }

  // if requested URL is an admin URL, allow session
  if(strpos($_SERVER['REQUEST_URI'], '/processwire/') === 0) {
    return true;
  }

  // otherwise disallow session
  return false;
};

Hope that you all have a great weekend. See you at ProcessWire Weekly tomorrow!

Comments

  • Teppo

    Teppo

    • 8 years ago
    • 10

    "While it may be possible to boot a 2.8.x instance from another 2.8.x instance, we aren't going to be officially supporting multi-instance on the 2.8.x version due to the lack of namespaces."

    Ryan, would you mind clarifying this part a bit? To be honest this is news to me, and (most likely due to my lack of understanding of the benefits of namespaces) I'm not sure why namespaces would even matter in this case. Thanks! :)

    • Can

      Can

      • 8 years ago
      • 00

      more a guess as I don't really know 2.8 but namespace introduction brought more than just namespace ProcessWire at the top. It ensures every instance knows which instance it actually is so there not rising any conflicts. Haven't personally checked but Ryan mentioned something about changing/replacing/dropping all static methods to ensure this
      Might be totally wrong^^

    • ryan

      ryan

      • 8 years ago
      • 10

      Teppo, the conceptual difference between 3.x and 2.x is integrated environments vs isolated environments. Due to namespaces 3.x plays well outside of it's environment. For instance, if you try to include a 2.x PW instance in a WordPress install, it'll fail because of function and class name conflicts. Whereas that's no problem in 3.x. Multi-instance falls right in line with the conceptual reasons for 3.x, integration. While it's true that technically you should be able to make 2.8.x instances talk to each other, you won't be able to reliably make a 2.8.x instance talk to a 3.x instance using multi-instance (they are in different namespaces). Also due to lack of namespaces, multi-instance would be mostly limited to isolated environments in 2.8.x. I don't see a good reason to support it in 2.8.x due to built-in limitations. For these reasons, if people are thinking of using multi-instance, they should stick to 3.x and that's where we should support it imo.

      • Teppo

        Teppo

        • 8 years ago
        • 00

        Thanks for the clarification, Ryan!

        As a clarification from my side, I was thinking purely in terms of using two or more 2.8 sites together, *not* using ProcessWire site inside another application or combining 2.8 with 3.x. This is the main benefit of multi-instance for me: making multiple ProcessWire sites talk to each other in a simple way.

        Assuming that 2.8 + 2.8 means that all classes are in the root namespace (or no namespace at all, whatever is the correct terminology in PHP) and 3.x + 3.x just means that all classes are in the ProcessWire namespace, the end result should be more or less the same. As long as you don't do this inside another application or mix 2.8 with 3.x, that is :)

  • NikNak

    NikNak

    • 8 years ago
    • 00

    Thanks so much Horst - that does the trick :-)

  • Diogo

    Diogo

    • 8 years ago
    • 31

    This just saved me a load of work to migrate some data from one install to another!!! Thanks Ryan :)

  • Can

    Can

    • 8 years ago
    • 00

    Now I'm really looking forward to remote multi-instance support as mentioned :D
    Then I could finally do https://www.happygaia.com/blog/processwire-local-to-live-sync/ way more easily :D

  • Diogo

    Diogo

    • 8 years ago
    • 30

    This just saved me a load of work to migrate some data from one install to another!!! Thanks Ryan :)

  • Can

    Can

    • 8 years ago
    • 50

    Amazing, thank you Ryan! And it just worked! (after updating to 3.0.30 at least, but could be because 2. instance was a little older than 1.)

    • horst

      horst

      • 8 years ago
      • 30

      @niknak: doesn't $image->httpUrl work for you?
      http://cheatsheet.processwire.com/files/file-properties/file-httpurl/

      • NikNak

        NikNak

        • 8 years ago
        • 00

        Hi there

        I'm having fun experimenting with this, but have hit a hitch when trying to show images from one site on another. I need to retrieve the root url of the site I'm pulling in, but can't use $config->urls->root obviously and $newsite->config->urls->root doesn't work either (where $newsite is the new pw instance). Any pointers would be appreciated.

        thanks again
        Nik

        • Yannick Albert

          Yannick Albert

          • 7 years ago
          • 10

          Wow! After, finally, playing a bit around, sharing some data between sites and merging/backing up stuff - I'm totally keen to see what comes in the future. It feels a bit "rigid" right now, but some port forwarding (e.g. `ssh -L`) and you're ready to work with external/remote servers.

        • Andrew Arnold

          Andrew Arnold

          • 6 years ago
          • 00

          Making the session cookie optional is very cool! Would it be possible to get the name of the login page (ie. via id?) so the page could be renamed without remembering to change the config.php file?

         

        PrevProcessWire 3.0.29 and 2.8.29

        This week we covered close to a dozen issues and around ten pull requests, so you'll find lots of updates in today's versions (3.0.29 and 2.8.29). Plus some info for those that have inquired about ProcessWire site development help. More 

        NextTesla Model S + ProcessWire

        8

        While we work towards a 3.x master version, we'll shift gears for a moment this week to review the Tesla Model S that we borrowed for the weekend, and were truly inspired by it. We've also got some updates to several 1st party modules this week as well. More 

        Latest news

        • ProcessWire Weekly #515
          In the 515th issue of ProcessWire Weekly we’ll check out the latest core updates, new modules, and more. Read on!
          Weekly.pw / 23 March 2024
        • Invoices Site Profile
          The new invoices site profile is a free invoicing application developed in ProcessWire. It enables you to create invoices, record payments to them, email invoices to clients, print invoices, and more. This post covers all the details.
          Blog / 15 March 2024
        • Subscribe to weekly ProcessWire news

        “Indeed, if ProcessWire can be considered as a CMS in its own right, it also offers all the advantages of a CMF (Content Management Framework). Unlike other solutions, the programmer is not forced to follow the proposed model and can integrate his/her ways of doing things.” —Guy Verville, Spiria Digital Inc.