Add non-finite floats and clarify signed zeros #506
Conversation
| Values such as "infinity" and "not a number" that cannot be expressed as a | ||
| series of digits are not allowed. | ||
| Leading zeros are not allowed. Integer values `-0` and `+0` are valid and | ||
| identical to an unprefixed zero. Hex, octal, and binary forms are not allowed. |
There was a problem hiding this comment.
Yep, one thing at a time!
README.md
Outdated
|
|
||
| 64 bit (signed long) range expected (−9,223,372,036,854,775,808 to | ||
| 9,223,372,036,854,775,807). | ||
|
|
||
| Float | ||
| ----- | ||
|
|
||
| Floats should be implemented as 64-bit IEEE 754 values. |
There was a problem hiding this comment.
It should say IEEE 754 binary64 values, since there's a 64-bit decimal64 as well.
There was a problem hiding this comment.
Ah, fair enough, will change.
README.md
Outdated
| nf1 = +inf # positive infinity | ||
| nf2 = inf # positive infinity | ||
| nf3 = -inf # negative infinity | ||
| nf4 = nan # not a number |
There was a problem hiding this comment.
IEEE 754 has 2 types of NaNs -- sNaN and qNaN.
It might be worthwhile to mention that which one it is would be implementation dependant?
There was a problem hiding this comment.
Yeah, and many of each! I'll add a note as you suggest.
toml.abnf
Outdated
| @@ -117,14 +117,19 @@ underscore = %x5F ; _ | |||
|
|
|||
| ;; Float | |||
|
|
|||
| float = integer ( frac / ( frac exp ) / exp ) | |||
| float = integer ( frac / ( frac exp ) / exp ) | |||
There was a problem hiding this comment.
maybe now would be a good time to change this to?
float = integer ( exp / frac [ exp ] )
toml.abnf
Outdated
| @@ -117,14 +117,19 @@ underscore = %x5F ; _ | |||
|
|
|||
| ;; Float | |||
|
|
|||
| float = integer ( frac / ( frac exp ) / exp ) | |||
| float = integer ( frac / ( frac exp ) / exp ) | |||
| float =/ non-finite | |||
There was a problem hiding this comment.
Maybe call them "named float values"?
There was a problem hiding this comment.
From a bit of googling, it seems like "special values" is the preferred nomenclature. How does that strike you?
toml.abnf
Outdated
| @@ -117,14 +117,19 @@ underscore = %x5F ; _ | |||
|
|
|||
| ;; Float | |||
|
|
|||
| float = integer ( frac / ( frac exp ) / exp ) | |||
| float = integer ( exp / frac [ exp ] ) | |||
| float =/ non-finite | |||
There was a problem hiding this comment.
I went with just special-float. values seems redundant.
|
@pradyunsg Haha, thanks. =) I'll wait for a few more thoughts from others before final merge. |
toml.abnf
Outdated
|
|
||
| frac = decimal-point zero-prefixable-int | ||
| decimal-point = %x2E ; . | ||
| zero-prefixable-int = DIGIT *( DIGIT / underscore DIGIT ) | ||
|
|
||
| exp = "e" integer | ||
|
|
||
| special-float = ( [ minus / plus ] inf ) / nan |
There was a problem hiding this comment.
I wonder if allowing ± in front of nan also would be simpler – keeping the universal rule that numbers can have a sign prefix. Of course +nan, nan and -nan would all mean the same thing, but it could be simpler. Not a big issue, just a thought.
There was a problem hiding this comment.
That's an interesting point. It would keep the consistency across all values, which would be a nice behavior. I suppose if the implementation wanted to, it could carry over the sign to the NaN value, even though it's insignificant, but who knows!
CHANGELOG.md
Outdated
| @@ -2,6 +2,7 @@ | |||
|
|
|||
| ## HEAD | |||
|
|
|||
| * Add non-finite float values (+inf, inf, -inf, nan) | |||
There was a problem hiding this comment.
Perhaps only list one of +inf and inf here – since this says "values" and those are the same value written with different syntaxes.
There was a problem hiding this comment.
Fair enough, will tighten in up.
Closes #389.
This adds support for non-finite float values (
-inf,inf,+inf,nan) and clarifies that signed zeros (integer and float) are valid and what they mean.