Mathematical const e is not defined

[siddMahen]

The following should be added to float.j:

const e = 2.71828182845904523536

On another note, very cool idea.

[StefanKarpinski]

I think that taking a single-letter variable name for a global constant may be unacceptable. You can also do exp(1), although having a constant would be nice. The question is what's a good name for it. We went with im for the imaginary unit (rather than i which would steal a really common variable name). Maybe en?

[JeffBezanson]

I noticed, to my horror, that exp(1) differs from the above value by 1 ulp.

[StefanKarpinski]

Eeek!

[JeffBezanson]

Actually, exp(1) is giving the correct round-to-even answer. But in this case rounding down gets you a teensy bit closer. Maybe we should call the constant e1 or exp1?

[StefanKarpinski]

So that you can write twice Euler's constant as 2e1? Oh, wait, or does that mean 20.0? Oops. We definitely can't call it that. However, since it's Euler's constant, how about eu? Also possible would be E although that does also conflict with floating-point literal notation at the moment. I've never much understood allowing both 1e2 and 1E2, however.

[StefanKarpinski]

You know. Maybe using e isn't so bad after all. It doesn't prevent using e as the name of a parameter of a local variable, actually. And that's not really a very common variable name even. The only thing it prevents is using e as a variable at the global level. Which is acceptable. (In the case of i, on the other hand, I think that even that would be unacceptable.)

If we do this, we should make sure that expressions like 2e parse as 2*e. It's unambiguous since that's not valid floating-point literal anyway, but does require some look-ahead, rather than unconditionally committing to 2e always being the beginning of a float literal.

[JeffBezanson]

Ok, the 2e fix is in. We can now add e if you want.

[StefanKarpinski]

Closed by b092738.