wrote:

wrote:

Instead the workflow is attached to already existing pages.

At least that is how people around here handle it. Maybe this is simply the wrong approach? Clemens

Hi, Yes, this makes sense because in the previous "doc:PageB" version, the link was in relative form (no space specified) and it was resolved to the current space (which was "Public" when PageA was in the "Public" space). After NS, using "doc:Draft.PageB.WebHome" means using a very explicit and absolute reference that does no resolution so it will behave the same, regardless of the context where PageA(.WebHome) is located (hence, it will link to the "Draft" space). There are a couple of cases where you could get away with it without any refactoring and using advanced knowledge on xwiki's link syntax resolution algorithm. TL;DR; These are the ones that involve links coming from higher in the hierarchy towards pages lower in the hierarchy or when linking between sibling pages (i.e. same level in the hierarchy). You can find more details on http://www.xwiki.org/xwiki/bin/view/Documentation/UserGuide/Features/XWikiS… (see the "[[.PageA]]" and "[[.PageA.PageB]]" examples and then the details on the resolution of "Untyped" links). Now, I have mentioned this just for the sake of completeness on the matter and would not really advise to rely on these "relative" links, since they are currently more of a side effect of our current model than a feature that is well-thought and recommended (e.g. we don't have support for relative links upper in the hierarchy like "../../PageX"). Given the proper support for relative links in our syntax, indeed, the listener and link fixing solution would be the recommended one. Hope this helps, Eduard On Sat, Dec 16, 2017 at 2:06 PM, Anca Luca &lt;lucaa(a)xwiki.com&gt; wrote: > Hello Clemens, > > Indeed, this is _one_ of the problems that the workflow application has on > a "nested" environment. > > I think it's a good approach for the fix, the only thing that you should > try to take care of is to make this as generic as possible wrt the > configured draft space in the publication workflow config of the document > being published (fix links relative to this space). Don't hesitate, if > needed, to add a configuration of the "default publication space" in the > publication workflow config, I found the need for it in other situations as > well, so it won't hurt. > > Also, one issue I can think of (which is related to the other issues that > publication workflow has on nested) is that the page to which the link is > might not yet be published, and in this case, imo, the link should be > refactored and be a "broken link" until the target page gets published. > > Thanks for working on this, > Anca > > On Thu, Nov 30, 2017 at 12:10 AM, Clemens Robbenhaar < > c.robbenhaar(a)posteo.de >> wrote: > >> Hi Devs, >> >> some users reported a problem to me concerning the Publication Workflow >> Application and the nested pages feature. >> The problem is as follows: >> >> In their old instance (some 6.4.x) they had a draft space and a public >> space and workflows attaching each page in the draft space with a page in >> the public space in an 1:1 manner, like: >> >> Draft.WebHome --> Public.WebHome >> Draft.PageA --> Public.PageA >> Draft.PageB --> Public.PageB >> >> Now if a link in, say, PageA in the Draft space points to another page, >> e.g. PageB, this is done via a relative link like [[link > text>>doc:PageB]] >> After publishing PageA, the link of the published variant points to the >> published variant of PageB (and not the variant in the Draft space), > which >> is the desired effect from the users point of view. >> >> Now after migration to 9.10 the situation is as follows: >> >> Draft.WebHome --> Public.WebHome >> Draft.PageA.WebHome --> Public.PageA.WebHome >> Draft.PageB.WebHome --> Public.PageB.WebHome >> >> Now a link to PageB in PageA looks like: [[link >> text>>doc:Draft.PageB.WebHome]] because both pages are no longer in the >> same space, but each in their own space. >> Publishing the draft makes the link in the published variant point back > to >> the Draft space, of course. However this is different from the behavior >> before nested pages and the users experience this as a bug. >> >> A manual solution would be to edit all pages and make the links point to >> the published variant of the pages. However this is not only a cumbersome >> and error prone task, but also the users are used to the "automatic" link >> conversion that made links in the published variant point to published >> pages and are wont to continue setting links to pages in the draft space, >> expecting them to point to the "right place" after publishing. >> The fact that now titles are used throughout in the navigation tree, and >> the titles of the draft and published variant are usually the same and > thus >> indistinguishable complicates the problem even more. >> >> To fix this I could write a separate extension that listens to Publish >> events send from the Workflow extension, and which converts the links in >> the published variant before it gets saved. >> However I remember that I once experienced strange issues where the > events >> were sometimes not send to the event listener in question, probably due >> some class loader issues or the like. >> (I only experienced these problems in a production instance and was not >> able to hunt them down in a development instance.) >> >> Also I cannot think of a use case where users want to link from a >> published variant back into the draft space (except for pointing to the >> draft version of the page itself, which is provided by the panel, > however.) >> >> Thus I would like to implement the link transformation in the extension >> itself. If a link in a page points to a page in the draft space of that >> workflow, I would change that link in the published variant to point to > the >> corresponding published variant of the link target. >> I could add a checkbox to the workflow class so people can switch of that >> link transformation if they do not want it. >> >> >> What do you think? Are there any objections to implementing such a >> feature? Or do you have different ideas how to solve the problem? >> >> Best Regards >> Clemens >> >