Re: RFC: String interpolation

[Paul Eggert]

On 12/12/2016 11:51 AM, Ted Zlatanov wrote:
> I like how this separates expressions from strings. It might be more
> readable if the % begins the next string:
>
>      (format "Altitude= " alt "%.2f, direction = " dir "%.2f.")

Unfortunately although this syntax is clean it would not be upward 
compatible, as 'format' currently silently discards excess trailing 
arguments. In contrast, Lars's suggestion:

> (format "Altitude= %.2f%" alt ", direction = %.2f%" dir ".")

is upward compatible. Also, I suspect that Lars's suggestion will be 
less error-prone than mine, as it would be more likely to diagnose stray 
occurrences of "%" at the end of a format.

> It has some downsides: it only works in one function; `(apply 'format ...)`
> can become surprising; and repeating a parameter requires duplicating
> the code.

True for repeating a nontrivial parameter. However, that is not common 
and 'let' should be good enough for that, e.g., (let ((dup (complicated 
expression))) (format "appearance1 = %f%" dup ", appearance2 = %g%." dup)).

I'm not sure what is meant mean by 'only works in one function'. 
format-message should behave like 'format', and any function that calls 
'format' or 'format-message' will get the extended behavior for free. 
For example, (message ...) would get the extended behavior.

(apply 'format FMT ...) would not get the extended behavior unless FMT 
itself uses the extension, and this should be good enough to avoid most 
surprises.

Ah, perhaps you're thinking about mixed forms like (format "x = %1d, y = 
%2d%" EX WYE ", z = %3d" ZEE)? To avoid confusion, the extension could 
work only if the format string has just one conversion specifier, at the 
end (just before the "%" at end of string). The idea would be to 
encourage people to write either the extended (format "x = %1d%" EX ", y 
= %2d%" WYE ", z = %3d%" ZEE) or the traditional (format "x = %1d, y = 
%2d, z = %3d" EX WYE ZEE) instead of trying to write a confusing mixture 
of the two.