[GiNaC-devel] bugs related to parser

[Alexei Sheplyakov]

Hello,

On Sun, Dec 04, 2011 at 12:59:53PM +0100, Cedric Studerus wrote:
> When working with the parser to read in user defined functions from strings
> I noticed 3 things:
> 
> 1)
> parsing a user defined function (declared to have 1 argument) from a string
> does not throw an exception when the string contains this function with
> zero arguments.

This looks like a bug.

> 
> 2)
> parsing products of user defined non-commutative functions returns a
> commutative product

Built-in GiNaC parser does not handle non-commutative products. You can
either 
1) Transform all non-commutative products into some dummy function(s), and
   explain the parser how to handle them (i.e. define and set the custom
   prototype table).
2) Use the archiving facility to (de)serialize expressions.

> 3)
> The parser accepts userdefined functions with zero arguments and parsing
> such a function does not throw when there is 1 argument.
> 
> 
> I have attached a small file "main.cpp" such that you can reproduce this
> behavior.
> 
> I'm using these user defined functions with zero arguments but it looks to
> me like a hack and I would like to know if you are going to allow them (
> e.g. with macro definition) or if you completely suppress them in future
> releases of GiNaC.

Zero-argument functions sound a bit strange (why don't you use symbols
instead?), and I doubt we actually support them (they might work by accident).
I don't think we will disable defining zero-arguments functions on purpose:
preventing users from doing silly things typically prevents them from doing
smart things, too. However I doubt zero-argument functions will be supported
properly (unless someone sends a patch).

Best regards,
	Alexei