Sven Neumann wrote:
Hi,
Grzegorz Borowiak [EMAIL PROTECTED] writes:
There could be problem if brush size is big and line has to be
half-transparent. E.g. brush radius is 10 and intensity is 0.5. There
would be 1-pixel-close overlapping brush hits, whose intensities would
accumulate. But this problem can easily be compensated if we use more
sophiscated tool_t function, which would reasonably decrease its intensity
(simply dividing nominal intensity by radius size and then by sine/cosine
value of line angle) for all pixels except two ends, and these two ends
would be painted with full nominal intensity, but only half of brush would
be painted (this half which does not overlap with any mid-line brush hit).
I don't think this would work well. Instead I'd suggest to draw the
brush totally opaque to intermediate tiles, then blend these with the
choosen brush opacity onto the drawable. Might be more memory-intensive
but should give perfect results. On the other hand it would diverge a
lot from what happens if you draw a line by hand. Would it make sense
to consider a line-drawing tool? I think it would be especially useful
to stroke selections and paths. Our current approaches to stamp the
brush at equidistant positions along the outline does not give pleasant
results and all attempts to improve this will most probably only mess up
the (already weird) code.
I agree.
Such a line-drawing tool wouldn't work with brushes but would have
configurable line-width and cap settings. I might be wrong, but I think
the ink tool might be suited reasonably good for proper line drawing.
If it would have support for drawing straight lines using the Shift
key, we could test it and this should be easy to implement.
For single pixel lines I think we should really go with a bresenham
type solution.
The antialiasing on the ink tools looks visually a little off to me.
not how easy it would be to fix that.
I have a design in mind for an AA bresenham line drawing function that
supports wide lines with end caps, mitred corners and customizable tip
profile shapes but I doubt I'll ever have time to code it up.
Jay Cox
[EMAIL PROTECTED]
___
Gimp-developer mailing list
[EMAIL PROTECTED]
http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer