On Wed, 24 Aug 2005, Michael Sperber wrote:

> As someone pointed out, how do you point to the location of
>
> (
>
>   3
>       4
> )

In principle, I think attaching location only to identifiers is enough to do
this, /if/ we give the programmer access to the full source (or optionally
some parsed representation of it).  Given the location of the keyword at the
head of the macro invocation, the user can then walk from there as she sees
fit.  (The head location could also be a dynamic property of the current macro
invocation rather than being part of the identifier.)

In fact, if programmatic access to source information is a good thing, I would
say this is the ultimate in source tracking, and the current
practice of attaching as hoc information piecemeal to syntactic
data is but a poor approximation to it ;-)

The problem with taking syntax objects as the parsed representation of the
source is that it that it has limits.  There are aspects of the source that do not fit
naturally into syntax objects as they are currently understood.

On the other hand, given the full source, we can do all kinds of nice things:
displaying error messages with code fragments as the user typed it, with
comments and whitespace, including some of the surrounding code for
easier identification of the context; extracting comments, bracket type,
whitespace, presence or absence of dots, etc.

Cheers
Andre