Deprecate ma.INFINITY and add ma.MAX

[dariosanfilippo]

Hello, people.

After some discussion on the topic, I was thinking that it could be appropriate to deprecate ma.INFINITY and add ma.MAX with the same definition. This is because ma.INFINITY and the actual INF special value behave differently and it can be inconsistent. For example: ma.INFINITY - ma.INFINITY = 0 while inf - inf = NaN.

Ciao,
Dario

[sletz]

• do you mean ma.INFINITY was incorrectly named and should be renamed as ma.MAX ? (see https://docs.microsoft.com/en-us/cpp/c-language/limits-on-floating-point-constants?view=msvc-160)
• and what do you mean by the actual INF special ?

[dariosanfilippo]

I don't think that we can remove it as I can think of some code using it, for example, Bart's limiter, if I'm not wrong, but we could add ma.MAX as the corresponding FP MAX constant.

I'm distinguishing between the INF generated by, say, 1.0/0.0 and the constant ma.INFINITY which is actually a number.

[magnetophon]

I don't think that we can remove it as I can think of some code using it, for example, Bart's limiter, if I'm not wrong, but we could add ma.MAX as the corresponding FP MAX constant.

I am indeed using ma.INFINITY in https://github.com/grame-cncm/faustlibraries/blob/master/basics.lib

Should those be search-replaced with ma.MAX?

[sletz]

Yes better, although ma.INFINITY is maintained for compatibility reasons.

[sletz]

OK, done 358a804.