retain-zero-exponent = true behaviour change

[ArchangeGabriel]

In v3, this now adds a ×10⁰ to things like \num{7}, while previously it would only display for \num{7e0}. Is there a way to fix that?

[ArchangeGabriel]

(Also the string seems gone from the documentation)

[ArchangeGabriel]

Also, in a similar fashion now \num{e4} produces 1×10⁴ whereas it used to produce 10⁴ (the former being produced by \num{1e4}). Certainly not directly related to retain-zero-exponent, but likely a similar change in the code?

[josephwright]

This is 'by design'. I was not happy in v2 that the same values could end up printing differently, as I wasn't careful enough about normalisation of input. In v3, I've tightened up the option names to show where the choice is made: this one is now print-zero-exponent, i.e. it's after parsing and is purely about visual formatting.

[ArchangeGabriel]

Hum, so what you say is that there is no way to have a different output for \num{7} and \num{7e0} as well as \num{e4} and \num{1e4} in 3.x without specifying on specific occurrences print-zero-exponent in the former and print-unity-mantissa in the later? The e0 cases is not too much a big deal (only two instances in all my .tex files), but the other one much more, I have hundred of both, on purpose…

[schtandard]

Would a third option print-zero-exponent=input (and accordingly print-unity-mantissa=input) be possible that prints the respective part of the number if and only if it is present in the input?

I'm not sure if there is any precedent for something like this and if this coupling of input and output is undesired, but I also want to mix both settings in the same document (depending on the context of the number) and being able to control the omission of these possibly redundant number parts by just inputting it accordingly used to be rather handy.

At least for print-unity-mantissa I find the new behavior a bit confusing, personally. After all, 1.00 is (correctly) treated differently from 1.0. If 1.0e9 means "two significant digits" and 1e9 means "one significant digit", I would expect e9 to mean "no significant digits" (i.e. just stating the order of magnitude).

[ArchangeGabriel]

@schtandard Thanks for your input on this, very good points and idea!

[schtandard]

Another option could be to handle this like drop-zero-decimal, i.e. instead provide two options drop-zero-exponent and drop-unity-mantissa that can drop but never add the redundant part. As I can imagine someone wanting to also add it, I would prefer the option outlined above, though. (Also, this would require changing option names, which is a more severe change.)

[josephwright]

I gave this a bit of thought and decided that @schtandard's pint about significant figures is reasonable. I've adjusted the code so the parser retains that difference: I may look at an option to normalise e10 to 1e10 for v3.1 (easy to do, just a question of if it's required).

[josephwright]

@ArchangeGabriel Will that work for you? I'm more reluctant to change the parsing of exponents, as I think all numbers do have an order of magnitude

[ArchangeGabriel]

Yeah, as I said  e0 is a very special case that I use only twice (in my lecture on writing numbers, to say that theoretically we should indicate ×10⁰ but that this is always omitted). So I’m fine manually handling those two. ;)

So thanks for fixing the no-mantissa case, that was actually my biggest grief (but I’ve discovered the change a bit after since the first case appears later in my lectures). :)