Re: patch for thingatpt.el

[Stefan Monnier]

>>>>> "Drew" == Drew Adams <address@hidden> writes:

>> > Below is a patch for thingatpt.el that provides additional
>> > functionality and new thing and form functions. In particular:
>> > (1) it lets you supply a syntax-table argument to most functions,
>> > to affect the definition of what a word etc. is, and (2) it
>> > provides functions to return things near point, not
>> > just precisely at point.
>> 
>> I believe that point 1 is misguided.  We have `with-syntax-table'
>> for those cases and it is quite sufficient AFAIK.

> Right, and it's a much better approach too. Sorry about that. My enhancement
> predated `with-syntax-table', and I wasn't aware of it.

> Here is an updated change log and patch, which also incorporates the `defun'
> bug fix I sent after the previous patch.

> ! (defun symbol-at-point-with-bounds (&optional non-nil)
> !   "Return (SYMBOL START . END) with START and END of SYMBOL.
> ! Return nil if no such Emacs Lisp symbol is found.
> ! SYMBOL is the `symbol-at-point' (which see).
> ! If optional arg NON-NIL is non-nil, then the nearest symbol other
> !   than `nil' is sought."
> !   (with-syntax-table emacs-lisp-mode-syntax-table
> !     (form-at-point-with-bounds
> !      'symbol (if non-nil (lambda (sym) (and sym (symbolp sym)))
> 'symbolp))))

This is wrong.  The name is clearly generic, whereas the docstring then goes
on to say it's only specific to Emacs Lisp symbols.  I think the name is
right and the docstring (and code) are wrong.

Check "recent" changes to symbol-at-point where I've fixed a similar problem
(basically you shouldn't call `read').


        Stefan