Le 11/03/2014 10:10, Roy Xia a écrit :
Hi everyone,
Hello,
I suppose my first question to you all is whether there are any alternative suggestions or additional goals out there for overhauling the converter scheme.
To be frank I am not sure yet. We have some bugs in the current code, like: a sweave document[*] (that outputs Tnw, and should follow a Rnw->latex step) cannot use the platex (japanese) converter. platex (or rather japanese) may need to be a modifier.
[*] see the OutputFormat layout tag, that allows to preprocess latex files.
I also suspect that the path-finding algorithm will be further complicated by the presence of multiple flags; for example, I have noticed that pdf7 (PDF cropped) is not labeled as a vector graphics format whereas other pdf variants are, so if the pdf extension is to be consolidated I suspect that vector graphics format will also have to become a qualifier of some sort.
I think that many of our flags will become modifiers.
In that sense, there may need to be some judgement on which converter chain to use if, for example, the search only results in two valid paths, one with qualifier A but not B, and another with qualifier B but not A. All said and done though I'm not an expert on the converters, so perhaps someone could shed some light on other considerations or critique my analysis of the task at hand.
I suspect indeed that there will be several possible paths. We need some criterion to find the better path.
As for dealing with the latex flavors, within the context of the converter it seems they aren't used for much aside from generating log and aux files--in fact, unless I missed something,
It is used to set the OutputParams::flavor vluae and they change the content of the LaTeX file.
My gut instinct says to just have a general "latex" flag set for each of these formats and automatically determine the proper program to call from there without knowing what specific type of latex it is, but since I'm not that experienced with latex or any of its flavors I'm not sure if that's the right approach.
I would say that at first you decide to output "latex". If your final format is pdf:pdflatex, then all the converter chain (and all the formats in the list) inherits the "pdflatex" flag and therefore you know that your first format has to be "latex:pdflatex".
Finally after all of those changes are in place there would probably need to be some UI changes to reflect the modified converter scheme, but I think that's more of an afterthought.
Well our current converter setting UI is a mess so if there is enough time, they woll deserve more than an afterthought.
Finally, we may need to add some pref2pref python code to update the users preferences.
Thanks for reading all of that and I appreciate any advice you have to spare.
I think your thoughts are on the right track :) JMarc
