Andre Poenitz <[EMAIL PROTECTED]> writes:
| On Mon, Oct 13, 2003 at 10:00:10AM +0200, Lars Gullik Bjønnes wrote:
>> | with FuncResult being
>> >
>> | struct FuncResult {
>> | dispatch_result res;
>> | idx_type idx;
>> | pos_type pos;
>> | };
>>
>> Are you sure that no dispatch ever uses the idx and pos passed in?
>
| Sure they do. Tab in a math array sort of increments the index.
>
>> | The reason is that we'd most likely have to add a third item soonish to
>> | model 'Parlist/Paragraph/Position' of InsetTabular and I do not want to
>> | touch all headers again. And it is less clutter...
>>
>> I have no problem with that, so ask me to do it when the time arises.
>
| But I have.
>
| Perhaps bad memories regarding Fortran subroutines taking 37 arguments
| and compilers not complaining if only 36 have been passed...
>
>> I'l continue to do monotonic improvements anyway.
>> (The return of FuncResult f.ex., which actually fit rather well with
>> my plans for dispatch.)
>
| I'd pretty much prefer the modifiable reference as the resulting code is
| much shorter.
>
| Compare
>
| void dispatch(FuncRequest & cmd)
| {
| ++cmd.idx;
| cmd.result = DISPATCHED;
| }
>
>
| with
>
| FuncResult dispatch(FuncRequest const & cmd)
| {
| FuncResult res;
| res.pos = cmd.pos;
| res.idx = cmd.idx + 1;
| res.result = DISPATCHED;
| return res;
| }
I actually like the second one a lot better, the first one smells of
voodoo.
And it would not be a realy problem to havd a FuncResult able to take
a FuncRequest as an arg to extract information from it:
FuncResult dispatch(FuncRequest const & cmd)
{
FuncResult res(cmd);
++res.pos;
res.result = DISPATCHED;
return res;
}
--
Lgb
