FloatFormat should allow scientific notation with given precision

[dschrempf]

Hello!

I was wondering if we could get the possibility to format a Float or Double in scientific notation with a given precision (number of decimal places). That is:

scientificFixedPrecision :: Int -> FloatFormat

Further, it would make sense to format Float and Double in standard notation, but with a given number of significant digits.

[sjakobi]

scientificFixedPrecision :: Int -> FloatFormat

There was a brief discussion about this in #365 (comment). Quoting @la-wu:

I don't think it would be particularly hard to implement something for precision for FExponent (we should be able to largely follow what is in base https://github.com/ghc/packages-base/blob/master/GHC/Float.lhs#L602 as I did for the FFixed). But ideally we would move to using a raw buffer-formatter for both FFixed and FExponent with precision so they could be equally efficient there.

Are you interested in working on this? Maybe @la-wu could advise?!

---

Further, it would make sense to format Float and Double in standard notation, but with a given number of significant digits.

Can you clarify the difference of what you're proposing to the existing standard formatter?

> formatDouble (standard 4) 12.345678
"12.3457"

[dschrempf]

Can you clarify the difference of what you're proposing to the existing standard formatter?

What I meant is the following: Let x=0.0039. One could think of a format formatDouble (significant 1) x which prints 0.004. formatDouble (significant 2) x would print 0.0039, formatDouble (significant 3) x would print 0.00390. Does this make sense?

Are you interested in working on this? Maybe @la-wu could advise?!

I will have a look at the discussion and the linked file, thanks for the references!

[sjakobi]

What I meant is the following: Let x=0.0039. One could think of a format formatDouble (significant 1) x which prints 0.004. formatDouble (significant 2) x would print 0.0039, formatDouble (significant 3) x would print 0.00390. Does this make sense?

Yes, this makes sense to me. I think it would be best to have a separate issue to track this idea though.

Regarding the name, I think it would be nice to have one that marks this format as a variant of the existing standard format. standardWithSignificantDigits is very long though…