RE: How do I highlight word at point?

["Drew Adams" <drew.adams@oracle.com>]

> here's what i would want to have ideally:
> press a key, expand selection to the current word, press again, expand
> to the next semantic unit (with respect to the current lang/mode),
> press again, expand further.
> 
> This idea is borrowed from Mathematica's IDE (aka the Front End).
> I have wrote full description of this feature, see:
> http://xahlee.org/emacs/syntax_tree_walk.html
> 
> I have a draft version for over a year now:
> 
> (defun mark-semantic-unit ()
> "Select a semantic unit.
> Select a word under cursor. "word" here means alphanumeric sequence
> plus "_" or "-".
> 
> When this function is run for the first time and there is no active
> region, then it selects the current word (i.e. sequence of
> alphanumerics plus hyphen or underscore).
> When the function is run again, or if there is a region, it extends
> the selection to the next level of enclosing delimiters."
>  (interactive)
>  (require 'thingatpt)
> (let ((deactivate-mark nil))
> (if (or (and (eq last-command this-command))
>         (and transient-mark-mode mark-active))
>     (let ((e1 (region-beginning))  (e2 (region-end)) b1 b2)
>       (goto-char (- e1 0))
>       (skip-chars-backward "^<>("{[?<\"'")
>       (setq b1 (point))
>       (goto-char (+ e2 0))
>       (skip-chars-forward "^<>)"}]?>\"'")
>       (setq b2 (point))
>       (set-mark b1))
>   (cond
>    ( (looking-at "\\_<")
>     (setq pt (point))
>     (push-mark pt nil t)
>     (forward-symbol 1))
>    (t
>     (forward-symbol -1)
>     (setq pt (point))
>     (push-mark pt nil t)
>     (forward-symbol 1))))))
> 
> this version works when your cursor is inside a matching pair. For
> example, if your cursor is at A in: (something here A and that)
> then invoke twice it'll select the whole paren.
> 
> However, it doesn't work when the cursor is in a screwed nested
> position. For example: (something here A (and) that)
> if your cursor is at A, invoke twice won't get the whole outer paren,
> because the code doesn't track nested parens, it only looks for chars
> in a dumb way.
> 
> Ideally, this mark-semantic-unit should just extend a semantic unit,
> where what's considered a semantic unit depends on the language. But
> this i imagine would be rather a non-trivial problem. I am not sure
> emacs's syntax table system is rich enough to be used for this.
> 
> So, taking a step back, i tried to have the code just extending to
> outer matching pairs. (effectively a extend-semantic-unit for lisp) I
> think now i can make this code work by using the sexp navigation
> functions. I should code this soon.
> 
> If anyone can implement this, or better yet, write a proper extend-
> semantic-unit for langs with C like syntax (C++,java,javascript),
> that'd be a killer feature i think.

I'm not sure that it would help you toward what you want to do, but you might
look at `thing-cmds.el'. It has some commands for selecting successive things of
different types.
http://www.emacswiki.org/emacs/ThingAtPointCommands.

See also perhaps:
http://www.emacswiki.org/emacs/MarkCommands#mark-a-word-or-thing.