Re: BIFF loader source

[Nick Lamb <njl98r ecs soton ac uk>]

On Thu, Jan 17, 2002 at 10:37:49AM +0000, Michael Meeks wrote:
>       Given that there are 1450 lines of relatively complex formula parsing /
> converting code that are then modified by 20 lines of changes -
> including large cut and paste tables etc. it doesn't seem like there is
> much sharing proposed.

Proposed? Lots. Actually implemented? None. So far.

If you think I butchered that code then you're right. The one thing you
shouldn't do when properly engineering code is to hack away at it until it
compiles, and then try to fix it later. But that's exactly what I did
to get the formula parser into that state.

My only defense is: It was fast, and it can and should be fixed after
both plugins are in CVS where they can easily be kept in sync.

>       Clearly - cut and paste of large chunks of (complex) code is barbaric,
> and worse when the code is complex (fragile?) than otherwise. I would
> strongly recommend then that these (great) changes are merged back into
> the mainline - so the work isn't wasted - during the Gnome 2.0 port,
> where (presumably) it is ok to touch the importer. But that it remains a
> standalone, extra for 1.0 - in which case cut & paste is perhaps
> justified.

That's pretty much what I was hoping for. There are two types of changes
made versus the original formula parser, as you can probably recognise.

#1 Handling older versions of structures (members got wider in later
versions of Excel, some positions were moved around). These can go into
the existing Excel plug-in but they'll never be triggered. They should
be considered where they simplify sharing code and otherwise ignored.

#2 Differences in the rest of the parser code, like the function
parameters and the global structures. These mostly need fixing in my
code, but the Excel plugin needs to bend a little too.

It's late. More tomorrow?

Nick.