Andreas L Delmelle wrote: > On Jul 20, 2007, at 14:08, Andrejus Chaliapinas wrote: > > Hi > Hi > Thought I'd chime in on this one as well... > > First off: I agree with the general sentiment that this is a very > interesting initiative! > > That said, as a note of caution, I do wonder whether the motivation > (with or without the bounty) is enough by itself to make it happen. > I've recently also been looking in the direction of the patch in > bugzilla 40271, but I'm under the impression that certain other things > need to happen first before a decent auto-layout implementation > becomes a feasible project... Not Patrick's fault, --he was > motivated-- but maybe such an enterprise was too complex to take on as > a one-man GSoC project. Even if it didn't immediately succeed, it did > give us a better impression of other changes that could reduce the > complexity drastically. I am still motivated in participating in this feature. As you said there are things that need to happen, or at least in my case things that need to be understood before being able to do a decent auto-layout implementation.
I remember the main difficulty I had was to get information (i.e. the width) up the tree in order to compute the proper width of each column. Moreover a mechanism would have to be devised to avoid performing some operations twice (the first pass to determine the optimal width and the second one to render the table). >> (...) >> Some time ago I've tried to dig a little into those Layouts >> techniques, but >> quickly found that it's not the thing I could enhance myself easily >> (though >> it was some initial information provided by Jeremias and Simon and maybe >> others, but still not easy to understand without UML graphs of pieces >> interaction), >> so stayed mostly with work on various XSL/XML parts for my >> projects needs and left those FO things for real experts :). > > What helped me a lot in understanding certain parts of the overall > processing were not necessarily words or graphs, but mainly random > debug sessions. Stroll through the code, place a few breakpoints in > the parts of the code that interest you, debug the process, have a > look at the call stacks, step further to see which other methods are > called afterwards to get a hang of how the objects and the logic fit > together, and which methods are called at what point in the process... > Not perfect, but very educational. :-) I completely agree. I remember spending hours and hours just figuring out which methods are called at lower levels of the layout engine and how the information works its way back up the tree. > > Cheers > > Andreas Regards, Patrick
