On Mon, 2007-01-29 at 08:36 -0800, Rob McDonald wrote: 
> On 1/29/07, Lars Clausen <[EMAIL PROTECTED]> wrote:
> > As I see it, there's two possible situations:  Either the text is plain
> > text or it's a LaTeX formula (of some kind).  The formulas we have no
> > chance to fit correctly until a separate LaTeXObject is introduced (which
> > could start out as a clone of TextObject) which the MetaPost renderer can
> > handle specially.  Plain text we actually have a chance to not just
> > position right, but also have a reasonable width of -- which is used in
> > the diagram, for sizing boxes etc.  If we always leave the size to
> > LaTeX/MP, the output is 99% guaranteed to be different from what you see
> > in Dia.  If we give LaTeX/MP the size (width!), any plain text will be
> > just as in Dia, but formulas will need fixing.
> 
> For the plain text case:
> 
> Your plan works if you are perfectly exact in calculating the width of
> the text.  However, if you get a slightly different result from the
> text rendered in LaTeX, you will wreak havoc.  Lets say you can
> predict the text width within 1%.  If you use my MetaPost macro to
> force the text to fit your estimated width, you will end up with
> inconsistent results.  Some text, you underpredict its width, so the
> macro shrinks it slightly.  Some text, you overpredict its width, so
> the macro grows it slightly.  What you end up with is subtle (random)
> variations in the font size throughout the document.

Rather than changing font size (which we have had bad experiences with),
can't we just have each text line be fitted in a LaTeX box of the
appropriate size?  Then we can make use of LaTeX's fine kerning
abilities to get the width right, and font size wouldn't suffer.

> On the other hand, if you leave things the way they are...  You can
> use your 1% error estimate of text width within DIA to calculate
> position, box size, etc.  You can output these diagrams and let
> MP/LaTeX handle the text sizing.  All text will have exactly the same
> (deterministic) font size.  However, things like the margins in a box
> drawn around text will be subtlety wrong.  When I do a flowchart, I
> usually pick a box size big enough to surround all of my text.  I then
> use a consistent box size, even if sometimes it means the margins are
> larger than others.  So, for me, slight variations in the margins are
> unimportant.

Unfortunately, 1% is when we're lucky.  I've seen many cases of 10%
difference or more.  I'll show you samples as soon as I get the metapost
thing to work.

> Of course, for truly complex LaTeX text, we don't have any hope of
> calculating the size inside of DIA until we integrate a LaTeX renderer
> into DIA in some way.

Indeed.

-Lars

_______________________________________________
Dia-list mailing list
[email protected]
http://mail.gnome.org/mailman/listinfo/dia-list
FAQ at http://www.gnome.org/projects/dia/faq.html
Main page at http://www.gnome.org/projects/dia

Reply via email to