[Lazarus] Windows coordinate limits

[Hans-Peter Diettrich]

Alexander Klenin schrieb:
> Apparently, Windows places some limitations on the sizes of lines drawn.
> See attached project -- it seems the maximum length is $7FFFFFF = 2^27 - 1

That's perfectly sufficient for currently available screen and paper 
formats.


> 3) Write some wrapper(s) to work around the problem?

There is no problem with the output, the problem is the data source. 
When you want a picture of the whole earth (Google Earth like), you need 
an appropriate coordinate system and storage format for the data, and 
means to map the data into an canvas (zoom factor, section).


> On the more generic note, I have a need (in TChart) for some
> floating-point geometry
> routines, like intersection of line with rectangle.
> Is there something like that in LCL? If not, should I add them to
> TChart or to some place in LCL?

GDI allows to map arbitrary world coordinates to an viewport (mapping 
modes: MM_ISOTROPIC, MM_ANISOTROPIC). When the mapping parameters are of 
insufficient accuracy, the world coordinates will have to be mapped into 
viewport (canvas) coordinates outside the drawing system, with 
appropriate clipping. This is what you can provide.

I'm undecided how such mapping should be made available. One solution 
were a "high precision" canvas, with all methods overloaded for world 
coordinates as Int64 and/or Double/Extended, and added properties for 
the viewport mapping (genercis?). Another solution, with less methods, 
would require to map all coordinates prior to passing them to the canvas 
methods; not so nice to use, but with far less added methods, or 
encapsulated in a dedicated mapping object.

DoDi