Can you recommend a script or software to quickly detect these crossed BCPs that are often created in an extrapolated weight? FL's FontAudit does not do it.
Personally I use extrapolation quite a lot and, if controlled properly, with predictable proportional relations between node handles it actually works quite well.
In my image I moved a bit the handlers so the problem is more visible. Otherwise it looks like a segment when in reality the handlers are crossed.
It happens when you extrapolate (maybe also interpolate) a segment and a segment looking curve. Yes, sometimes I do use extrapolations when I want to go lighter or darker of the masters, but of course then I have to correct well the extrapolations because any small inconsistency is going to be increased.
I found out Glyph Nany extension for Robofont is a great contour audit tool. much more powerful than FL 5's audit. It's helping me to spot these problems.
FontAudit covers that, at least is should give you "Unnecessary Extremum" error in (at least) two places
Nope. FontAudit will detect the problem if the handlers have been moved like in my image, the real problem is when everything is resting on the same line. It's no visible and FontAudit does not detect it.
FontAudit covers that, at least is should give you "Unnecessary Extremum" error in (at least) two places
Nope. FontAudit will detect the problem if the handlers have been moved like in my image, the real problem is when everything is resting on the same line. It's no visible and FontAudit does not detect it.
Now, that is a bit different story - all resting on same line. Yes I have encountered the problem, but rarely and by chance - otherwise it is not easily spotted.
Now I can confirm your results, that FontAudit will not detect the problem, but only under certain circumstances (or at least in all my tests). Audit will not report the problem if the BCP handles of the node are no longer than 14 units (strange), otherwise it will return "Unnecessary Extremum".
Glad to know that "Glyph Nanny" does such a good job. I took a brief look at its code and I liked, what I saw - made me think about porting it to FontLab.
Comments
But yes, FL flags it just fine.
In my image I moved a bit the handlers so the problem is more visible. Otherwise it looks like a segment when in reality the handlers are crossed.
It happens when you extrapolate (maybe also interpolate) a segment and a segment looking curve. Yes, sometimes I do use extrapolations when I want to go lighter or darker of the masters, but of course then I have to correct well the extrapolations because any small inconsistency is going to be increased.
I found out Glyph Nany extension for Robofont is a great contour audit tool. much more powerful than FL 5's audit. It's helping me to spot these problems.
Nope. FontAudit will detect the problem if the handlers have been moved like in my image, the real problem is when everything is resting on the same line. It's no visible and FontAudit does not detect it.
Now I can confirm your results, that FontAudit will not detect the problem, but only under certain circumstances (or at least in all my tests). Audit will not report the problem if the BCP handles of the node are no longer than 14 units (strange), otherwise it will return "Unnecessary Extremum".
Glad to know that "Glyph Nanny" does such a good job. I took a brief look at its code and I liked, what I saw - made me think about porting it to FontLab.