Support underscores for mpz/mpq assignments from strings

[Hans Åberg]

> On 14 Jun 2021, at 15:51, Vincent Lefevre <vincent at vinc17.net> wrote:
> 
> On 2021-06-14 14:50:43 +0200, Hans Åberg wrote:
>> 
>>> On 14 Jun 2021, at 10:59, Vincent Lefevre <vincent at vinc17.net> wrote:
>>> 
>>> On 2021-06-11 18:20:01 +0200, Hans Åberg wrote:
>>>> But you can probably cannot change mpz_set_str,
>>>> because it is modelled on C behavior.
>>> 
>>> It isn't. In C, whitespace isn't a separator.
>> 
>> The GMP manual, sec. 5.2, is not clear on this, as it says that whitespace is discarded, whereas in C/C++ it is only the initial whitespace.
> 
> I think that the GMP manual is clear: mpz_set_str ignores whitespace
> anywhere in the string (contrary to C's strtod, etc.).

Then perhaps it is OK to add it to ignore '_' as well: In C/C++ I think the idea is that one should be able to pick up a sequence of numbers separated by spaces. The idea of mpz_set_str may be to pass the whole number, and only that, to the argument.

Using the low line (underbar) '_' may be a good choice, better than space ' ' as it does not cause a confusion with a sequence of tokens. But also better than '.' and ',' (the WP also mentions thin space, or half-space " ", and apostrophe "'"), because traditionally numbers are monospace even in variable space fonts so that they line up in tables. So if '_' has the same width as the numbers, that tradition can be kept. Not all variable with fonts do that, though.

But check with the GMP developers what they think.