Re: Patch for sql.el

[Kevin Rodgers]

Michael Mauger wrote:
> --- Stefan Monnier <address@hidden> wrote:
>>Is there any particular reason why you don't merge all those submodes
>>and simply handle a language that's a superset of all?  Basically
>>accept @ # $ as symbol components, and accept all the keywords of all
>>the known servers...  That should simplify your code.
>
> The problem with SQL is that there is no real standard.  The core is
> standard, but the functions and procedural extensions vary significantly
> from one product to another.
>
> One of the reasons I originally started using sql-mode was that I was
> using two database products at the same time (oracle and ms).  The
> product specific features helped keep me sane.  For example, Oracle calls
> its sub-string function `substr' whereas MS (and ANSI) call it
> `substring'.  I go back 20+ years with Oracle so my instinct is the
> abbreviated form.  When I used the abbreviated form in MS scripts I had
> immediate feedback that I had used the wrong form.
>
> I have one more patch coming for sql.el that removes, at compile time,
> keywords in product specific lists that are also in the ANSI list.  This
> slows the compile time, but seems to speed up the fontification.  It also
> allows me to keep the entire product specific keyword lists in the source
> without manually filtering out the keywords duplicated in ANSI.

Why not define the vendor-specific modes as derived modes?  Would that
simplify things?

--
Kevin Rodgers