Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Overlapping Paths
#21
For the problem of the leading A's, this could be a Gimp bug. The A is obviously truncated in text layer for "Adobe Caslon" (while the text path is OK), and it seems the same applies to "American typewiter" even if it is less obvious because the serif is thin (but its serifs are very wide). Printing a sequence of 'A' and comparing the A's at both ends with the A's in the middle could be a good way to demonstrate the problem.

For Arnold, it appears that the holes strokes are listed before the outlines. Same for Anabelle, but this isn't consistent, this only happens on the "e"'s. I will see if I can improve the algorithm, checking if the bounding box of a hole is completely within the bounding box of the next, and if so swap the strokes.
Reply
#22
Doing some testing for me:

1) get latest ofn-path-edits from Sourceforge (for improved "summary" function)
2) Create using the "Arnold" font a path that is just one letter having a least one loop (or even two, like uppercase "B")
3) Using OPE's summary, in which order are the strokes listed? Does the outline (longest stroke, normally) comes before of after the holes? And are the holes in the opposite direction to the outline (CW vs CCW)?
Reply
#23
If it is any help.

Using the Arnold font: Longest path definitely first and clock-wise. The hole anti-clockwise.

A different font 'Simply Rounded' Two clock-wise holes first, then the longest anti-clockwise.

screenshot: https://i.imgur.com/tZqR4WY.jpg

kubuntu 18.04 / Gimp 2.10.14
Reply
#24
(05-11-2020, 07:46 AM)Ofnuts Wrote:
  • I wonder what is the deal with the leading "A" of the "Adobe Calson Pro" and "American Typewriter"

That might by my fault. When I was resizing some of the fonts to get them a bit closer in visual size and not just the same font size, some of the fonts got cut off at the beginning. (not sure why that happened. switched to fixed instead of dynamic possibly?) I fixed the ones I noticed by inserting a leading space. Those two fonts must not have caught my eye.

Anyway, hopefully that clears up that issue. It should not normally be a problem.
Reply
#25
(05-11-2020, 06:15 PM)rich2005 Wrote: If it is any help.

Using the Arnold font: Longest path definitely first and clock-wise. The hole anti-clockwise.

A different font 'Simply Rounded' Two clock-wise holes first, then the longest anti-clockwise.

screenshot: https://i.imgur.com/tZqR4WY.jpg

kubuntu 18.04 / Gimp 2.10.14

Sort of what I expected alas. So I'm looking for a more robust algorithm. All those I envision all involve figuring out which stroke is completely enclosed by which other one.  I had such a (crude) algorithm in OPE's Break text path apart, but your Anabelle font demonstrates that it is not robust enough, so I'm using a better algorithm (known algorithm, I just need to write a decent python implementation).
Reply
#26
OK, so I have a working stroke-nesting algorithm(*). Interesting side-effect, I no longer need the orientation (the top-level strokes are mandatorily outlines, and the outline/hole type is determined by to the level (odd: outline, even: hole). This even solves the outline-overlaps-hole issue.

(*) Not very efficient, could be a killer with longer text Smile
Reply
#27
Sounds like it's much more bullet-proof which is more desirable than speed IMO.

Looking forward to trying it out!
*rubs hand together briskly*
Reply
#28
Not completely there yet. I have to investigate why some holes aren't cleared:

   

OK, got an explanation, my detection of enclosure isn't accurate enough, I work on polygon as defined by the path anchors. Some anchors are very far away from each other, and the straight line between them cuts the polygon of some inner stroke, so the stroke aren't considered nested. I'll have to add a few anchors to improve accuracy (Gimp also has a path interpolation function that could be even easier...).
Reply
#29
Latest results, after increasing the accuracy of the polygons I use for nesting determination. Besides the truncated serifs in the As, I could only spot one  significant difference with your initial rendering. Can you see it?

   
Reply
#30
Sorry for the late reply. I'm writing this on my media center NUC. The smoke genie escaped from from my desktop earlier today and I'm afraid it's pretty much a complete loss. I ordered a new computer but the lead time is expected to be at least 2 weeks. No gimp for me for a while.....

Nothing is jumping off the page as to a significant difference. I don't have my imagus extension installed on my browser yet so not as easy to zoom in and inspect. Looks great to me!
Reply


Forum Jump: