Juergen Spitzmueller wrote:

> Conversion fails always. LaTeX is complaining about a missing *.eps file
> (pdflatex about missing pdf files). In 1.3, there's the additional error
> that \usepackage{graphics} is not added to the preamble.

I'll have a look next year. Maybe it is me who is to blame here.

> OTOH I do not wonder that nobody noticed this. Who uses the external inset
> to insert files which can also be inserted via the graphics dialog?

I did use it some time ago, because I liked the "edit" button. Then I wanted
both preview and edit, and implemented the edit button in the graphics
inset.

>> Why did you change insetexternal? I am not sure if things like
>>
>> > bool ResizeData::usingScale() const
>> > {
>> > - return !float_equal(scale, 0.0, 0.05);
>> > + return (!scale.empty() && scale != "0");
>> > }
>>
>> work.
> 
> Look at the archives for the threads "float2string" and "validators for
> the rest". tostr(float) -- i.e. boost::lexical_cast -- is broken and Lars
> strongly argued for never storing float values internally, but strings
> (and LyXLengths). We had a longer discussion (mostly JMarc and Lars) and
> decided to go this way. So I am inbetween the process of changing all
> occurences of floats to strings. Some work has already been applied.

I did follow this thread, but I understood it that this should only be done
in the frontends. But maybe I am wrong.
In this particular case I wonder wether scale is always "0" when zero is
meant. Maybe it is "0.0"? or "0.00" etc. I believe that this value is read
from the .lyx file. Even if lyx always writes "0", we don't know what
happens in lyx2lyx, tex2lyx and maybe other programs that write .lyx files.
BTW, I appreciate what you are doing, I only want to avoid that it breaks
something.

>> > + float rotAngle = strToDbl(angle);
>>
>> Why not double?
> 
> Why double? Float is sufficient for this value.

Because strToDbl returns a double. And Lars prefers it. (thread [patch]
float -> string #1, 2004-12-20). I never use float personally, but I may be
a bit biased (doing mainly that sort of numerics where memory is not an
issue, but runtime).


Georg

Reply via email to