Module: Process Changelog

[teppo]

This module tracks changes, additions, removals etc. of public (as in "not under admin") pages of your site. Like it's name says, it doesn't attempt to be a version control system or anything like that - just a log of what's happened.

At the moment it's still a work in progress and will most likely be a victim of many ruthless this-won't-work-let's-try-that-instead cycles, but I believe I've nailed basic functionality well enough to post it here.. so, once again, I'll be happy to hear any comments you folks can provide

https://modules.processwire.com/modules/process-changelog/

https://github.com/teppokoivula/ProcessChangelog

How does it work?

Exactly like it's (sort of) predecessor, Process Changelog actually consists of two modules: Process Changelog and Process Changelog Hooks. Hooks module exists only to serve main module by hooking into various functions within Pages class, collecting data of performed operations, refining it and keeping up a log of events in it's own custom database table (process_changelog.) Visible part is managed by Process Changelog, which provides users a (relatively) pretty view of the contents of said log table.

How do you use it?

When installed this module adds new page called Changelog under Admin > Setup which provides you with a table view of collected data and basic filtering tools See attached screenshots to get a general idea about what that page should look like after a while.

For detailed installation instructions etc. see README.md.
 

[ryan]

Wow this looks great Teppo! I can't wait to try this one out tomorrow. 

[apeisa]

Amazing stuff Teppo aka Mr. Log (you can keep that name if you go to build career in adult entertainment business too ).

After testing this will definitely be on our default installation for all sites!

[Martijn Geerts]

Very nice one Teppo,

There are several sites where I could use such a functionality. 

[Pete]

Hi Teppo

Awesome module and very useful for a few of my sites with many editors

I know it's not really for version control, but I wonder if you'd considered somehow saving a text copy of the body field if that changes? Not sure how you would go about that (save to text file with page ID and timestamp maybe?), but it occurs to me that it's the one field where if someone changes you may need to quickly look at what was there before and change it back, so even a basic text dump would be of some use there I think. But yep, this ventures into the realms of version control then when I start thinking about things like this.

[Nico Knoll]

Looks really lovely!

[Adam Kiss]

Teppo,

you've just saved me some time, and my new favorite client some money. Great job!

One thing though: have you have any idea how this impacts performance of the website? (not much I guess, but I'd better ask)

`Thanks!

[ceberlin]

Hi, there were registered changes here today from a user "guest"... how can that be?

[teppo]

Hi, there were registered changes here today from a user "guest"... how can that be?

Most likely either API usage, or perhaps some internal PW feature.. for these user seems to be "guest", which is a bit confusing to say the least

If you're sure that neither of these were the cause, then I'd love to dig in a bit deeper. These are the cases that have triggered changes by "guest" for me in the past.

[teppo]

Teppo,

you've just saved me some time, and my new favorite client some money. Great job!

One thing though: have you have any idea how this impacts performance of the website? (not much I guess, but I'd better ask)

Sorry Adam, I've completely managed to miss this post! And thanks, by the way -- it's good to hear that someone has found this module useful

This is a bit late for a reply, but anyway: I wouldn't be too worried about performance impact. There's no heavy lifting happening code-wise, just some hook methods with a bit of relatively simple logic. One extra query (most of the time) for each page save shouldn't do much harm either and since this only happens when saving pages it's definitely not a problem on most pages.

On the other hand, if you do find that there's some kind of performance penalty, I'd be happy to check it out and see if things can be further optimized. (I'm sure they can, just haven't found a reason to do that so far.)

Only case where I would worry a bit about things getting out of hand would be a site where pages are constantly changed (ie. some public feature triggers page save all the time) and that data gets logged. In a case like that it would make sense to a) not log every action and/or b) limit the time log entries are stored for. I could also add a template setting at some point, so that admin can select which templates require logging in the first place..

[ceberlin]

Today we had an incident were lots of pages got unpublished. 

There were no normal editor activities around that time and there was no clearly seeable pattern.

The logs never should say "guest" if the api does something, better would be "system" or "API" and even better, which script.

Because of "guest" I immediately thought of a hacking attempt to the batcher, for example. But we found no sign of hacking.

700 "unpublishing"-changes in little time - this was a script and no person, for sure. We investigate at the moment.

[teppo]

@ceberlin: did couple of quick tests and it definitely looks like a) actions ran through Batcher get correctly logged as actions of the user that ran them and b) API actions via bootstrap method always point to Guest user.

In your case I'd keep looking for an external script that bootstraps PW and does some changes OR template code that alters content when a visitor loads a page with that template. Front-end edit forms would be potential causes for this too.

I have to agree that actions of API pointing to Guest is confusing, but as it's internal PW feature I prefer to keep it like that here too. Reasoning behind this seems to be that when doing something via API, whether triggered by unauthenticated visitor or some external script, you don't actually log in as any real user .. so the real issue here is probably just that "Guest" sounds very odd in this context.

[horst]

 The logs never should say "guest" if the api does something, better would be "system" or "API" and even better, which script.

 I have to agree that actions of API pointing to Guest is confusing, but as it's internal PW feature I prefer to keep it like that here too. Reasoning behind this seems to be that when doing something via API, whether triggered by unauthenticated visitor or some external script, you don't actually log in as any real user .. so the real issue here is probably just that "Guest" sounds very odd in this context.

I would really like something like a (maybe hidden) user named e.g. system or API or CLI that gets invoked for that (at least for bootstrapped scripts).

It has confused me too in the past.

Also if you create new pages that way, they get owned by "guest" in the system. Therefor I always try to log in first, when starting a bootstrapped instance. (But forget sometimes)

[teppo]

I would really like something like a (maybe hidden) user named e.g. system or API or CLI that gets invoked for that (at least for bootstrapped scripts).

Most of default content points (created_users_id DB field in pages table) to page ID 2, which is the /processwire/ page. This seems like a sensible solution to me.. even though /processwire/ obviously isn't real "user" either 

Edit: taking another look at that quote, I'd like to add that having user specific to bootstrapped scripts would be great. At least for those of us who use these quite a lot, it would help keeping track of things.

Edited July 22, 2013 by teppo

[ceberlin]

Finding that out of the sudden a "guest" batch-changed 700 file settings scared the hell out of me, I must say.

It scared me even more that I cannot see a pattern, why some files were changed and others not.

Glad to hear that "guest" was probably not a person and no hack, but only a script gone wild.

I really would find very useful if it was recorded somehow, what script was involved. But that is maybe to much to ask.

[teppo]

I really would find very useful if it was recorded somehow, what script was involved. But that is maybe to much to ask.

This won't help much with your current issue, but I've just pushed new version of this module to GitHub. This version adds an option to log which script triggered (ie. initiated) each action; a path or an URL, depending whether context was HTTP or non-HTTP.

This feature can be enabled via settings of Process Changelog Hooks module to either external scripts only or both external and local scripts (ie. current PW installation called in "normal way" -- bootstrap method is considered "external.")

Haven't been able to test it thoroughly enough yet, but so far it seems to work. If anyone has time to test this, I'd love to get some feedback.

[ryan]

I'd like to add that having user specific to bootstrapped scripts would be great. At least for those of us who use these quite a lot, it would help keeping track of things.

Guest means that no user is specifically logged in. But for things like your own bootstrapped scripts or other API usage, you can create a user specific to that need. Then in instances where you want them to be the current user (rather than guest) you can do a call like this:

$u = wire('user'); // save current user, if you want to
$users->setCurrentUser($users->get('api')); // set new user: api
// 
// .. your code that does stuff
// 
$users->setCurrentUser($u); // set back to the previous user, if you want to

[dragan]

I just installed this wonderful module.

Q: A client wanted to have such a feature. But his role / permissions are not "superuser", but "manager". Hence, he doesn't even see the "setup" items in the admin menu.

How can I add / change user roles / permissions, so that a user that has not "superuser" status, can access this module?

I tried cloning the admin tpl, switched the module to this "admin light" tpl, and tried accessing the Changelog page /processwire/setup/changelog/ (logged in as client) - but alas, I get a redirect to the main page/ menu...

Any tips? Is it at all possible? Really, this module is all the client should be able to see in the "setup" category...

[Wanze]

@dragan

Add 'permission' => 'your-permission'

to the array returned by the method getModuleInfo().

After this line: https://github.com/teppokoivula/ProcessChangelog/blob/master/ProcessChangelog.module#L35

Example: If all users who can edit pages should be able to access the module

    public static function getModuleInfo() {
        return array(
            'title' => __('Changelog'),
            'summary' => __('Keep track of changes (edits, removals, additions etc.)'),
            'href' => 'http://modules.processwire.com/modules/process-changelog/',
            'author' => 'Teppo Koivula',
            'version' => 126,
            'singular' => true,
            'autoload' => false,
            'installs' => 'ProcessChangelogHooks',
            'permission' => 'page-edit',
        ); 
    }

[dragan]

Thanks! That worked nicely.

I know I'm drifting away from this main thread topic now, but is there an easy way to add a custom admin link in the backend for a certain user / role?

i.e. adding "changelog" after "pages" in the top main admin nav? (without core-hacking)

I googled a fair bit, but didn't find any forum topics / modules / tutorials about this.

Problem: since access to this module is inside "setup", the "setup" menu doesn't appear for that backend-user. He has to remember the link page /processwire/setup/changelog/. I remember in MODX, there was an option to define a custom landing page after login. Or changing the admin dashboard (adding custom direct links). I was wondering if PW has something similar. (without creating a custom admin theme, or hacking the core)

Something like 

if ($user->role == this | $user->id == foo) {
  // echo | include ...
} 

inside a core file would probably be easy - but a hack...

[Wanze]

Nothing simpler than that. You know that every page in Admin is a page under "Admin"?

So you can drag the Changelog page from /Admin/Setup/ and drop it below the /Admin/Pages page.. et voila, the page is no longer displayed under "Setup"

[dragan]

Easy indeed! I tried moving it outside of the admin folder, but got errors - moving it just above "setup" worked

[kongondo]

@dragan...a clarification - if you have a chance, have a look at Wanze's code in his module Batcher. He creates a permission "batcher" which you can give a certain role. Although the batcher page appears under setup, it doesn't have to be moved manually from under setup in order for the non-supersuser (manager in your case) to see it. They will only see "batcher" or in your case "changelog" under the setup menu; nothing else. Try it and see!

Speaking of batcher, I prefer the route batcher takes of creating a permission to give access to the module rather than applying it site-wide with a page-edit permission (Wanze was just giving an example above). Of course, there will be times you don't mind giving permissions site-wide....OK, enough rambling...Hope this makes sense!

Edited for clarity

Edited February 1, 2014 by kongondo

[teppo]

Should've done this earlier, but I've just pushed to GitHub updated version of the module. This version is set to require "changelog" permissions and create it when the module is installed.

Dragan and any others who've already installed the module: I suggest that you get the latest version, add aforementioned permission manually and then apply it to those roles that should be able to access changelog page.

[dragan]

Thanks! Will try it out later.

Quick Q: 

I tried to use some of the vars to send an email each time something is changed. Something really simple like

$message = "operation: $operation 
page id: {$page->id} 
page template id: {$page->template->id}
details: $detailsStr";
		
@mail($to, $subject, $message);

$detailsStr is just a var_dump of $details. But it remains blank in my test emails.

I inserted it after line 283: $this->insert($operation, $page->id, $page->template->id, $details);