Re: [Rosegarden-devel] Re: [Rosegarden-bugs] gui main.cpp notationproperties.cpp notationproperties.h notationstaff.cpp notationvlayout.cpp notecharname.cpp notecharname.h notepixmapfactory.cpp notepixmapfactory.h notestyle.cpp notestyle.h

[Chris C. <ca...@al...>]

On Monday 27 Dec 2004 11:27, William wrote:
> I was relying on what I thought was a guarantee of NOTE_TYPE always being
> set, namely that [something complicated]. Have I missed some logic?

I don't know for sure without going to look at the code, and that's why you 
shouldn't do it.  Remember you're writing to us as well as the compiler.

Even if this test is in fact safe, the fact that it doesn't _look_ safe means 
that it could waste time by distracting anyone trying to locate mysterious 
crashes elsewhere later.  Especially since that previous conditional does 
include the extra existence test, which makes it appear necessary.

Defensive coding (which is all we want to do before 1.0) indicates that you 
should always do the extra test in a situation like that.  If we were in a 
more aggressive development phase and it turned out that these tests were 
unnecessary, it might make more sense to rip out all of them and add a 
comment saying why you think the property's existence is guaranteed at this 
point.  (Ideally this sort of thing would be done through some more rigorous 
pre and post condition tests of course.)  But we're not.

Moving on: The WHOLE REST ON STAFF (etc) Unicode/non-Unicode naming would be 
fine by me.

> > That does make for a potential incompatibility in the font .xml files,
>
> How so?

Because WHOLE REST used to refer to a whole rest without leger lines, and now 
it refers (or will refer) to a whole rest with leger lines.  So a new RG 
using old font .xml files will get the wrong character for that, and will 
also potentially have trouble with the fact that it asks for WHOLE REST ON 
STAFF that didn't exist in old .xml files.  (That's why we should have a 
fallback to the WHOLE REST version for when that character is not found.)

At the moment the font .xml files are all distributed with RG so this isn't 
likely to be a practical problem, but in principle people could have their 
own.  For all I know, maybe some people already do.

> Is there any graceful fall-back at the moment for any
> glyphs (not just MUSICAL SYMBOL WHOLE REST and MUSICAL SYMBOL HALF REST)
> that are unavailable?

Only in certain particular cases where it's thought likely that some fonts 
might not provide both characters -- as is also potentially the case here.  
For example, there are loads of fallback cases in 
NotePixmapFactory::makeTimeSigPixmap to cope with fonts that may or may not 
include special time signature digits, and in drawFlags to cope with fonts 
that have combining flags instead of separate glyphs for multiple flag sets.

I would suggest that any logic for handling this sort of fallback goes at the 
highest possible level (i.e. NotePixmapFactory), to avoid special cases in 
NoteCharName/NoteFont where possible.  Use the getCharacter method that 
returns a bool, and if it returns false, try the character with the opposite 
value of in/out of staff and use that instead.  The general principle is that 
it's better to render _something_ than to fail completely.


Chris