New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
menu system does not allow some menu strategies for touch devices #3527
Comments
While workarounds exist to implement this without changes to the CMS, I agree that having a standard way to handle this would be a nice thing. |
I don't like using the template for this. Right now you could actually emulate this behaviour by making a page with a redirect to it's first child on it, so clicking it actually does something which is even better! How about a |
Actually you can't use redirect as you'll want the click to just open the dropdown on touch devices. |
Firstly: I think the title of this issue is rather alarmist and very mis-leading. I've deployed many touch-capable sites with CMS. The BS3 method of dealing with touch devices isn't the "correct" way, but a way and CMS has no particular affiliation with the BS3 project (other than knowing it is awesome =) Secondly, I do agree that the current menu system is somewhat limited and is often frustrating because we cannot make intermediate organisations that aren't also pages in their own right. In these cases, the CMS is the frustrating part. Sometimes (most of the time?) though, having a proper-page that is a "parent" for its subpages is a good thing and it is here where BS3 fails the CMS developer. For example, here's a recent way I've had to "bastardize" BS3 to make a drop down that also works as a direct link the Products page. In the case, you can click the words "Products" or you can click the caret to revel the pages beneath. |
+1000 @yakky's 'empty_page' flag. |
"empty page" is easier to understand than "non-leaf node", but maybe "menu node" or something like that would be even clearer (bikeshedding, yay!) |
I'm so bad at naming, that it's granted that my proposals are the most horrible anyone can come up with. |
+1 "menu node" |
Adding another field to the node was an alternative I also considered. |
@jrief go for it, thanks! Also I'd like to have this in 3.0.x branch, if the patch will not be too intrusive |
@yakky why 3.0.x? This is a new feature, so should go to 3.1, right? |
Just to be clear, so we'll go with an extra boolean field |
@yakky IMHO once you start distinguishing between "minor" and "major" features the whole distinction between "x.Y.z" and "x.y.Z" releases becomes moot. I think the following system is quite logical and good:
EDIT: If the problem is that this means it'll take a long time for this feature to be available, maybe we should try (like every release ;-) to make 3.1 happen in a reasonable timeframe. |
We should do this in the 3.1 branch, mainly because this is related to the menus. The primary distinction (so far) on 3.1 is that we've (by "we" I mean "Patrick" =) ripped out mptt and replaced it with treebeard. As Jacob (or whomever) develops this feature, he'll likely be touching some of this and I'd rather we didn't have to support two underlying tree systems and can instead optimize for one. Also, as we go forward with this change, we're likely to be further testing Patrick's treebeard replacement even further, and possibly spot some otherwise difficult to find bugs. I realize that the treebeard is at a lower abstraction layer, but we're bound to touch on it in a material way during the implementation of this feature. And if we don't, then that also tests that our abstraction layer is properly abstracted =) Also, 3.0.x is now labeled "Support 3.0.x"... we don't add new features directly into a support branch. This may creep in there as a back-port though. @ojii. I agree 100% with your major.minor.bug-fix system! In fact, if we stuck to it, then the mptt->treebread change would have been on a x.x.X branch, and this extra menu feature would be a x.X.0 release, because features are user-facing. In my book, if it doesn't affect a user in a visible way, then it's not a feature. But c’est la vie. tl;dr; this should go in the 3.1 branch. |
@mkoistinen changing the fundamental data model of a project isn't quite a "bugfix" though. arguably it should be 4.0 if anything. |
Oh and thanks for changing the issue title, @jacob =) |
That was actually me @mkoistinen (should've written that in a comment, sorry) |
+1 |
It would be exceedingly helpful if I could add Pages in multiple locations in the Menu in such a way as to have translation sensitivity. So for example I may have Page called "Processes" for which I could have a menu item somewhere else, (for example it's first child) possibly labeled as "Process Overview" which simply takes the user to the "Processes" Page but which also works when the Page has a different slug and name in different languages. |
@Noah-A how is this related to this issue? It sounds like a Page with a redirect on it to me. |
@ojii @mkoistinen +1 |
@ojii I asked @Noah-A to put this issue here. In retrospect it should probably be its own issue. The reason I asked him to note it here is that I think that there are larger issues with menus that should be addressed with a larger plan of attack and not a piece-meal approach. I will create a new issue from Noah's points. The issue is larger than what it seems. Created new issue: #3535 |
The empty page concept seems like it could/would solve a related issue: the pages grouped under "Learn", "Develop", "Discuss" in the navigation might not need or in some cases shouldn't have '/learn/', '/develop/' as a component of the url. For example, perhaps the site was originally designed with an "About Acme" page at the root (e.g. /about-acme/) but for navigation purposes, this is now grouped under 'Learn'. Cool urls don't change yadda yadda so if this 'Learn' was an empty page and did not contribute any slug bits to the pages beneath them in the tree, then 'About Acme' could be moved in the tree under 'Learn' so it appears in the right spot in the menu but without any 301 redirects or other customizations required. My current project has the exact use case the OP mentions but I could not wrap my head around how to get the menu system to work smoothly without adding '/learn/' as a prefix to every page, especially when there really isn't a 'Learn' page, it's just an artifact of the menu organization/UI and shouldn't add any cruft to the urls. |
Ugh, I'm having a moron Monday (month more accurately). Just now realizing/remembering there is "Overwrite URL". Apologies for the noise and +1 to a built-in solution for non-clickable top-level menu items. |
I am not keen on adding a new page flag for What I think we need is a scheme for arbitrary page attributes. Then we could have whatever flags we wanted:
Templates would have access to these attributes, and could act on them as appropriate. For example: For menu templates to act on these attributes we'd need to do make some revisions to the menu code, so that menu nodes have a generic FK to the objects they represent (which I think they should have anyway). https://github.com/evildmp/django-cms/compare/pageflags-patch does an excellent job of adding arbitrary page flags. Whatever we do, I would prefer to be cautious and not just solve the immediate problem with a quick fix that simply bolts some new functionality on. |
Isn't this what page extensions [1] do? [1] http://docs.django-cms.org/en/latest/extending_cms/extending_page_title.html |
During my first attempt to implement this, I came across another pagemodel field: |
@ojii yes of course, my brain is still on 2.3.5. |
@jrief I think this is to do with pages that require authentication: should they be displayed in menus? |
@evildmp welcome to 3.0 :D However this raises the question, does this actually require a change in core if it could be solved with a 3rd party cms page/title extension? I've never used this new feature so I'm not 100% sure on the implications, but this got added for a reason, so why not use it here? |
I think it would still require changes to the menu system - a way of getting from a menu node to a Page in the template. |
This'll be a bit more complicated than that, since not all menu nodes have a Page associated. But maybe a |
A generic FK to the object. |
Page extension requires an additional join which is not always good performance wise. |
I think we all need to realise that this and other issues stem from the fact that the Page model is being overloaded with two distinct concepts: "Page" and "Menu Node". They're not the same thing. We should be looking for a way to de-couple them. |
Menu nodes aren't database objects, so no (G)FKs here.
I'd say either do it in core proper, or a separate 3rd party package.
That's not 100% true. The default cms menu [1] turns most pages into menu nodes. I think this makes a lot of sense, since you usually want your cms pages to be navigateable. [1] https://github.com/divio/django-cms/blob/develop/cms/menu.py#L220 |
@mkoistinen actually pages are the only user-changeable materialization of a menu node in django CMS, so it's pretty natural to handle the relevant menu options at the Page model level |
When I looked at the code I had the same feeling as @mkoistinen , the menu/node system shall be decoupled from the pages. I remember a few posting from people using djangoCMS saying, that they only use the Page-model without menu system (or vice versa, I don't remember). Anyway, it means that there is a "market" for separating these the page from the node. I have to ask again: |
@ojii, here's how I see it for most projects and this doesn't even include the menus I end up hard-coding into the markup because I see no other way to properly add them ("Login" and "Logout" are regular examples). I only have one thing on the right: "App menus", but in reality, this is a pretty large collection of things in most projects. @yakky, I completely agree, but herein lies the issue. The concept and implementation of a Page is tantalizingly close to what we need for a menu item, so, we run with it. And, its not just the model, its also the slick GUI (page tree), which would be a large undertaking to make something new just for menus. But, the fact is, they are not the same thing. I've used Drupal and Wordpress before Django CMS. Now, I'll state right here and now my preference is for Django CMS by a wide margin, but in both of those systems, the concept of a menu is entirely decoupled from their version of a page. In fact, in both of those systems, they have multiple "things" that can be added to menus, not just pages. We have app-hooks, which serve the purpose of those other "things", and we can attach them to menus via app-menus in a limited manner (must be a child of a page to which they are attached, and, we can only attach one set of menus to that page), but in the other systems, you can also:
I'm not necessarily advocating that we just rip out what we have and replace it all. I think it would be very possible to fix some of these shortcomings. For No. 4, we can probably add I18N to redirects by leveraging the Title model as it was intended. For No. 2, I can imagine a non-core app that provides an operator-friendly GUI for attaching arbitrary objects (including pages and even arbitrary urls properly) as menu items into arbitrary places in the existing menu via Menu Modifiers. This may require some relatively simple changes (like allowing more than one menu attachable to pages, in an ordered manner), but shouldn't require a complete overhaul at all. |
Assuming categories are a model, can't you just use the menu extension system to add these categories? Is the main issue being able to re-arrange these nodes?
having login/logout hardcoded seems fine with me. I don't consider them part of the page/menu hierarchy, but rather "global" functionality. |
But they are! Our operators should be able to move those around from position to position or menu to menu without any hassle. |
While implementing this feature, I came to the conclusion, that it is a bad idea to add another, somehow magic flag, to the Page model. I now strongly believe that we should separate the menu tree from the page models. In the future, web-sites will be built with partial views, but each of them will have a separate URL. This means that we somehow need a way to drop the 1:1 relationship between menu-nodes and pages and replace it by a 1:n-relationship. I'll write an extra proposal about this. |
Bootstrap 3 introduced a navigation bar, where on mouse over, the sub/pulldown-menu does not automatically pop up. This brought to some controversial discussions on the Bootstrap mailing list, but considering the importance of touch devices (no mouse over possible), I agree with dropping support for this.
The menu system in django-cms is built in a way, which in practice requires the mouse over effect. The reason is, that one can not create a CMS "page", whose purpose is to only act as a node in the menu tree, without offering any page content.
For instance, consider the menu system of a typical site build with Bootstrap 3: https://www.angularjs.org/ . Here the main menu contains items: 'Learn', 'Develop', 'Discuss', which actually do not offer any page content.
Now, assume one of the above items shall additionally offer some page content. This can be achieved by adding a copy of an item from the main menu, into its own pulldown menu. Here, the first click onto the main menu's item opens the pulldown menu. Then, a second click on the copy of that item, opens the requested page.
In django-CMS we could use the flag "in menu", but when this is turned off, then all menu items below that item are unselectable, and that's definitely not want we want.
My proposal is to add another magic template. Currently django-cms has a built-in magic template "Inherit from Parent". We could add another magic template, say "Empty Page", which then tells the menu system to act as a clickable menu item, but not offering any page content. Then, while building the menu, the template tag
menu
can distinguish a page acting as pure navigation node from a real cms page.If this proposal is accepted, I can implement it.
EDIT (by ojii): Note, the title of this issue was changed by me.
The text was updated successfully, but these errors were encountered: