Re: RFC: locale-sensitive Emacs functions

[Paul Eggert]

Thanks, this looks like a reasonable way to proceed. Some comments:

There is no need for any of those backslashes at line end. Also, please 
indent consistently.

The manual needs to be updated, and there needs to be some test cases.

> +The \\=' flag is ignored for all arguments except %d.

This places too many constraints on the implementation. Instead, please 
say only that the behavior is defined for %d, and leave things undefined 
if the ' flag is used for any other conversion. We may want to add 
support for %'x at some point, for example.

This should document what happens when the 0 and ' flags are both used. 
Are grouping characters inserted before zero-padding, or after? I think 
it's before (as in POSIX), but you should check this.

> +                 if (apos_flag && INTEGERP (args[n]))

I don't see why that "&& INTEGERP (args[n])" is needed. Please remove it.

The code does not appear to do the right thing when the space or + flags 
are used, or when the integer is negative. It treats the leading sign or 
space as if it were a digit.

The code assumes that format-digit-separator is ASCII; it might not be.

Add test cases for the above corner cases.

Adjust SPRINTF_BUFSIZE to account for the worst-case buffer size when 
%'d is used. I suppose it should be the maximum of its current value, 
and (INT_BUFSIZE_BOUND (printmax_t) + (INT_STRLEN_BOUND (printmax_t) - 
2) * MAX_MULTIBYTE_LENGTH), but you should check this.

Use an O(N) rather than an O(N**2) algorithm when inserting grouping 
characters.