>>>>> "Juergen" == Juergen Spitzmueller <[EMAIL PROTECTED]> writes:
>> Ha! I knew the question would eventually arise... The problem is
>> that ideally, we do not want to detect the characters around the
>> inset, but the latex output around the inset. It is especially
>> difficult to detect what is going to be output later.
Juergen> OK, I'm beginning to understand the problem. Thanks for the
Juergen> explanation.
Here is a clean way to make this work, and more:
We need to define our own variant of an ostream, that has knowledge of
how TeX works. It would be a normal ostream with the following
additional opeartions, implemented either as member functions or
iomanip thingies (I don't care). The names I propose are bogus.
- macrowoargs(): this is to be used to indicated that you have output
a macro without args with a alphabetical name, like "\foo".
Action: just set a flag
When the next character is given, the flag requires to take the
following action:
* c == ' ': output "{}" before c
* c alphabetical: output ' ' before c
* otherwise, just output c
The flag is then reset
- unwantedligature(string liglist): this is to be used to indicate that
the latest character may produce an unwanted ligature with any of
the characters in liglist.
Action: remember liglist
When the next character is given, if it is in liglist, output
\textcompwordmark{} before it and reset liglist.
- breakline(): This allows to break lines without fearing to add a
paragraph break. This can simplify a lot of code and improve output
quality.
Action: output a \n if and only if we are not at the beginning
of the line.
- breakpar(): this allows to obtain a paragraph break without adding
to many empty lines.
Action: output \n\n if not at the beginning of line; \n if at
beginning of line but not beginning of paragraph; nothing otherwise.
- other features may be automatic wrapping of lines and automaic
texrow accounting (or whatever scheme Lars wants to implement).
Those two mean a big simplification of code.
Note that implementing this `latexostream' can begin with any one of
the features enumerated above, since they do not depend on each other.
Basically, the big work is to write the class (probably easy, but I do
not know ostream interface well enough to know which method(s) should
be reimplemented) and change all of our latex() methods to take a
latexostream as argument instead of std::ostream.
After this is done, the changes can be done gradually.
Do this look like a plan? Now, we only have to find someone to
implement it :)
JMarc
