[fpc-pascal] fpc-passrc expression parser (issue #16914)

[Michael Van Canneyt]

On Tue, 13 Jul 2010, dmitry boyarintsev wrote:

> Hello, Michael Van Canneyt,
>
> Sorry, I can't write to the resolved issue (without re-opening) it.
>
>> - better class tree (split out binary/unary/primitive operators in subclasses)
>> These may break backwards compatiblity; if you prefer to do them yourself, I can say what I have in mind.
>
> That's how it was originally written. However, I personally prefer to
> keep less subclasses.

I prefer more, with less fields, it saves memory on big expressions :-)

> I would like to hear your ideas.

Well, at least I would have expected 4 subclasses:
TPasExpr
  +-- TUnaryExpr (field operand)
  +-- TBinaryExpr (fields left, right)
  +-- TPrimitiveExpr (field 'value')
  +-- TFunctionCallExpr (fields value/params)

I must say that I am not familiar with the parsing style you used. 
I wrote 2 expression parsers meanwhile, and both used a different 
style. (see fcl-base/fpexprpars.pp; the other one is in fcl-js. 
A third will follow in the sql parser I'm about to commit)

> thanks,
> dmitry
>
> P.S.. the expression parser still doesn't resolve ranges (i.e. in
> sets), neither simple nor complex, like:
>
> Char(6) in [Char(sizeof(Integer))..Char(SizeOf(int64))]

That should be relatively easy to add, no ?

Michael.