[Grace-core] Meeting Reminder NZ Wed 20, 11am-12 noon; US Tue 19, 2-3pm.

[Andrew P. Black]

On 14 Nov 2013, at 01:42 , James Noble <kjx at ecs.vuw.ac.nz> wrote:

> - need to annotate papers with reviewers comments  (andrew has started)

and finished.

> - need to make small fixes
> - need better examples
> -    particularly typechecker 
> - separate "examples of dialects" from "writing new dialects" sections 

One messages that came through very strongly is that the paper does not make clear what has actually been done, vs what might be done.  An example of this is the section on hiding helper methods used in constructing a dialect.  The paper talks about an "is local" annotation , but does not make clear whether that has actually been implemented (as part of the dialect for writing dialects), or is completely hypothetical.  I believe that some reviewers felt that we were deliberately obfuscating the true situation.

The other clear message (that I remember) is the need for evaluation.   This could be addressed by pesenting dialects that, for example, 

(1) make Grace "first order", by restricting blocks (used in the dialect) to be arguments to methods provided outside of the dialect.   

(2) improve an error message that an instructor finds troubling, or replaces a general-purpoise error message by one that is more appropriate to a sub-language (for example, insists that the arguments to while()do() be literal blocks, and not arbitrary expressions, and issues an informative and helpful error message in that case).

(3) need context information, collected at one point of the AST, and used somewhere else.  I'm not sure if the FSA example does this — perhaps a BDD dialect like Graceful cucumber?

Part of this would involve describing the AST-waking interface more deeply, and giving a pointer to the documentation of that interface.

An alternative approach, if we can't actually show that the dialects mechanism is ready for prime time, would be to leave the speculative langue in and re-direct to Onward!

	Andre