On Tue, Aug 14, 2012 at 10:35:14PM +0200, Thien-Thi Nguyen wrote: > > between the source and target vertices. A better explanation is at: > > (info "(elisp) Circular Objects")
Ok. > That said, we want to be nice to programs that have only basic ‘read’ > capabilities, and should plan to support references "manually". (That's > what INDEX is for -- to support on-demand dereference, sidestepping the > requirement that everything be held in memory.) So the full answer to > your question is "yes, but doing so would be bad design (IMVHO, YMMV)". I think it really depends on what is to be represented. To represent the link between an @author command and the associated @quotation, I don't think an index should be used, but instead such kind of reference could be ued. But for index commands and printindex, floats and listoffloats, sure they belong to a special INDEX part. But I don't really mind in fact, just tell me what you want, I'll construct it. > As for the information that may be wanted in a renderer, looking at > what is available now, there is the index informations (index > entries, merged indices), informations on floats (for @listoffloats), > informations on labels (@node, @anchor, @float with label) especially > for internal @*ref, a possibility to access the tree of some > @-commands, like @copying, @documentdescription, @top, @settitle..., > and some misc informations, like the succession of @direntry and > @dircategory, the input file name. > > We mustn't forget the all-important ‘@paragraphindent’ [froth froth] !!! > I have been waiting a long time for that to be runtime-tweakable... That is to be found in the tree, as it affects all the Texinfo coming after the @-command. Some @-command affect the whole document, so would be available right at the beginning of the rendering. > This points to another subtlety of texinfo: IIUC the lifetime of a > variable or setting (or macro) is indefinite. That is, if you ‘@set’ a > variable in 1.3.5.7 (a subsubsection in the first chapter), that value > will hold for all subsequent nodes (even 2.1, i.e., higher level), until > EOF or another ‘@set’ changes it. This means we need to record where > such settings are done to know their "span of influence". It is worse than that. User defined macros and @value expansion may happen out of a balanced tree. So these 2 constructs are expanded when constructing a proper tree in texi2any. It should be possible to keep a 'mark' where they appeared (it is on the TODO list), but they cannot be kept in the tree like the other commands are, since the tree would not be a tree anymore. If a generated sexpr tree is generated through makeinfo/texi2any, my idea is that the tree would have user defined macros and @value already expanded (macros definitions and @set would be in the tree, though). -- Pat
