Distance field flips depending on TTF/OTF file.

[hak]

Generated SDF flips depending on TTF/OTF file.
For example, in 'Times New Roman.ttf' (coming with OSX), SDF is generated for outer direction
while in 'NotoSansCJKjp-Bold.otf' (https://www.google.com/get/noto/), it's generated to inner direction.

Need to check a winding direction for a robust SDF generation?

[daredevildave]

I've found some other fonts that appear to be inversed, either completely or partially.

This one seems to be completely inverse: http://www.dafont.com/sports-world.font

[Chlumsky]

So does the font file specify which direction it uses or am I supposed to guess it from the geometry somehow? Unless it does, I'm thinking of just adding a switch that will invert it. Anyway, for now, you should be able to fix the distance field if you just invert the colors.

[behdad]

TrueType-flavored OpenType fonts (aka TTF) conventionally have clockwise outer contours, while CFF-flavored ones (aka OTF) have the reverse. That said, these are conventions, and both work perfectly fine in most apps if they do not follow the convention. I have some code in GLyphy for fixing up contour directions if you care to check.

[boulabiar]

I had the same problem with svg downloaded from thenounproject.com
I've made some tests in the shader and I have
float sigDist = 0.5- median(sample.r, sample.g, sample.b);
or
float sigDist = median(sample.r, sample.g, sample.b) - 0.5;
depending on the case. (it's just a temporary solution..)

[Chlumsky]

I have added an option (-reverseorder) that will fix the output for shapes defined in reverse order.

[chaoticbob]

@Chlumsky - I borrowed the logic from invertColor for OTF fonts and the result looks in correct. I changed the 0.5 to 1.0 and it seems to produce correct results. Wondering if I'm missing something on this? Font is Chalet Comprime Cologne 1980. It also lives in various places on the internet.

Also wondering if anyone has seen this on TTF font as well?

[daredevildave]

The -reverseorder flag doesn't seem to be working correctly as noted above.

Compare the two results below. First is generated using no -reverseorder flag and then the image is inverted in graphics magic. Second is generated using -reverseorder flag.

I believe the error is as noted above. If you change the value in invertColor from 0.5 to 1.0. The results come out the same.

[Chlumsky]

You are absolutely right, sorry, I have no idea how I overlooked this.

EDIT: Fixed.

[Chlumsky]

I'm not getting the same behavior. Actually, what you're claiming is the output with reverse order looks a lot like what would happen before the fix.

[vkalpias]

Looks like you're right. I had a weird local merge on my fork of your repo which seemed to use the old version! Thanks and sorry I'll delete my posts.

[Chlumsky]

Since 0e68504, the standalone version will automatically guess the orientation of the input by sampling a distance outside the shape's bounds. This can be disabled by -keeporder or -reverseorder.