Numeric arguments behave unintuitively

[oantolin]

As far as I can tell giving iy-go-to-char a numeric argument n behaves as follows:

1. It always searches forward for the abs(n)-th occurence of the character. One might reasonably expect positive n to search forward and negative n to search backward.
2. Upon repetition, that either by retyping the character or by typing ;, it moves by one occurence. The move is forward if n>0 and backwards if n<0.

Possibly the most consistent behavior possible would be:

• Both the first jumps and repetitions go forward by n occurences. Here "forward n occurences" means "backwards abs(n) occurences" when n is negative.

But I see the value of point 2 above: you might include a count for the first search but if that didn't get you where you wanted to go, you probably want to adjust by one occurence forwards or backwards. But point 1 definitely seems counter-intuitive to me: why always forward even if n is negative?

I suggest changing the behavior so that negative n means even the initial search is backwards. This has the advantage that then M-- M-x iy-go-to-char would be equivalent to M-x iy-go-to-char-backward (which would then become unnecessary, if you ask me).

While I suggest changing the behavior, there might be a good reason it was designed this way, which I'd be glad to know. Even with the current behavior I would say there is a documentation bug: the documentation currently correctly states that repetitions move by one occurence instead of moving by n occurences, but it does not mention the stuff about how negative n affects the direction of search in the repetitions.