bug#14395: 24.3; [PATCH] new feature smie-highlight-matching-block

[Stefan Monnier]

> I want something for octave mode that looks like something in the
> attached screenshot. But since this is generic I would like to put it in
> smie.el. Do you have any objections or comments?

Looks like a good feature, thank you.

> It doesn't make sense for this feature and smie-blink-matching-open to
> be on at the same time. So in the patch nothing is enabled.

I don't think enabling it in octave-mode makes sense: this is more like
"blink-paren vs show-paren-mode", i.e. a personal preference.  So the
enabling/disabling should be done via code in smie.el.

> +  (when (and (bound-and-true-p smie-closer-alist)

It's defvarred to nil, so don't test if it's boundp.

> +    (let ((open-re (concat "\\_<"
> +                           (regexp-opt (mapcar 'car smie-closer-alist))
> +                           "\\_>"))
> +          (close-re (concat "\\_<"
> +                            (regexp-opt (mapcar 'cdr smie-closer-alist))
> +                            "\\_>"))

The string returned by smie-forward-token-function is usually the same
as the representation of the token in the buffer, but not always.
So the above is not strictly correct.

Instead you want to call smie-for/backward-token-function and then
compare the result via (r?assoc tok smie-closer-alist).

> +         ((funcall beg-of-tok open-re)
> +          (with-demoted-errors
> +            (forward-sexp 1)
> +            (when (looking-back close-re)
> +              (funcall highlight (match-beginning 0) (match-end 0)))))

I think this should not use with-demoted-errors but instead should
explicitly catch the scan-error and turn it into a message.
After all, the user doesn't want to be thrown in the debugger just
because his sexp is not properly closed yet.  And also this way you can
provide a much nicer error message.


        Stefan