[Help-smalltalk] Re: Question regarding the parser(s)

[Paolo Bonzini]

On 09/12/2010 07:21 PM, Holger Freyther wrote:
> Hi Paolo,
>
> I try to come up with a test case for my parser issue and I am a bit confused
> and maybe you can help me out.
>
> 1.) if I use FileStream fileIn it will end at a primitive and will always
> use the internal C Smalltalk parser?!

Unless you use "PackageLoader fileInPackage: 'Compiler'."

> 2.) Behavior evaluate: 'Object subclass: Foo'. In the default implementation
> this will be wrapped by a 'DoIt [ ^ [', code, ']]' and then executed?
> Internally this will be passed to compile which should end in
> VMpr_Behavior_primCompile?
>
> Now the weird part is, that part does not appear to know about GST Syntax.

Yes, and the fact that is wrapping your code with a method, is exactly 
why it doesn't know about the GST syntax.  The GST syntax is not valid 
in methods.

To parse GST syntax, you can pass a Smalltalk stream to #fileIn, like

    st> 'Object subclass: Foo []' readStream fileIn
    a ReadStream
    st> Foo
    Foo

> 3.)
>    PackageLoader fileInPackage: 'Compiler'.
>    Behavior evaluate: 'Object subclass: #Foo3', returns nil and not
>    Foo3 as without the Compiler.

That's a bug:

    st> PackageLoader fileInPackage: 'Compiler'.
    ...
    st> Behavior evaluate: '3'
    nil

I added to the issue tracker.

>         b) Should the Compiler return the (last?) class it created?

Only because it's the return value of #subclass:.

Paolo