Re: [xwiki-devs] XAR source projects should allow source files

On 26 Aug 2016, at 13:50, Vincent Massol <vincent(a)massol.net> wrote: Hi Paul, I’ve finally taken the time to read this discussion (a bit quickly for the last mail though, I admit ;)). Let me give my POV: * What I understand is that you’re not proposing to change the XAR format but to introduce some tooling to generate a XAR from a set of source files, at build time. Correct? * IMO we should not touch the existing XAR plugin. This is current the standard way to generate XAR files and we need to keep, if only for backward compatibility with existing sources. * IMO you should develop your tool as a new maven plugin in xwiki-contrib and use it on some contrib projects to validate and show to others how it works. Of course you can depend on the existing XAR plugin if you need tor reuse some code (it’s even possible to introduce a new api module if there’s an important need to share common code). * FTR this is exactly what Jean Simard as done with the XFF tools (https://github.com/xwiki-contrib/api-xff). He’s done is separately to validate it and show it. We need to know where we stand on XFF since it’s currently the most advanced POC that we have on introducing a new format for storing wiki pages in the SCM. It also supports both directions: from source to XAR and from wiki pages to source. I’m planning to check what’s missing in that progress to be able to use it (not much IMO).

* If you were to develop your maven plugin, you would just compete with the XFF code. This is fine, although I’d have preferred that we spend our time finishing the XFF work. It’s your time so your choice though. In the future, we’ll be able to choose which solution we want to use between the 3 that will exist (current XAR plugin, XFF solution and your solution). * FTR Fabio had also worked on this topic in the past, see http://design.xwiki.org/xwiki/bin/view/Design/DirectoryStructureforXWikiApp…. Caleb has also worked on something related: https://github.com/xwiki-contrib/xwiki-tools-node So for me the important part is for you to develop your solution in a separate module in xwiki-contrib as a demonstator and possibly start using it on some contrib project to validate it. The alternative is to work on XFF and finish it. Personally I’m not a big fan at all of modifying the existing XAR plugin for various reasons: * We don’t know yet what solution we’ll choose in the future. At this point in time I even have a preference for XFF. So it wouldn’t be fair and nice to modify the existing XAR plugin with your solution and then have to remove the code if we don’ want it in the future. * It will be much simpler for you to develop your own project in xwiki-contrib since you’ll have commit access (which you won’t have in xwiki-commons). WDYT? Thanks -Vincent > On 22 Aug 2016, at 15:59, Paul Libbrecht <paul(a)hoplahup.net> wrote: > > > Hello fellow developers, > > many of you are developing XAR-packaged resources. The exchangeability > of these archives is one of the precious aspects of XWiki, I believe. > > However, if one see all the many GitHub-contributed xar packages, the > source is not there: it's in a Wiki with which one would "open it". A > pity since much of the content is code. > > I'd like to propose to allow XARs to be built form source code which is > true source that one edits. > Code files with their proper syntax coloring, editors, auto-completion, > search. Media-files that are edited with our tool of choice by a simple > double-click and not a complex pull, edit, and replace. > > Thus, my proposal is to allow inclusions within the XML files. > Inclusions should be of type: > - text to include page content or object properties (e.g. css of cssx > objects) > - attachments to include attached objects > > My current setup is documented here:http://jira.xwiki.org/browse/XWIKI-13643 > It does binaryInclude to generate base64 and textInclude to generate > XML-text. > Thus, I can advertise a github project which is a macro and whose > complex function is in a velocity file that is in a .vm file. However, I > do not need to explode everything, e.g. the text-content describing the > macro or (in this case too small) translation set. > > What do you think? > Should I rather generate complete attachment elements (name, length, > modif date, ...) instead of doing a binaryInclude that just generates > base64 ? > Is fetching files for inclusion beyond the boundaries of the project of > any use? (it is currently prohibited for safety) > > thanks in advance. > > Paul