World: r3wp

What was the audience reaction about that?

Thanks. I've been talking about such techniques for many years in 
Syllable and in general, and nobody will ever believe me. I'm very 
happy to now have a tool to show them

The audience... sigh. They were friends, but there was only one programmer, 
formerly C++ and now Python. I asked him beforehand how long he thought 
his equivalent Python program would be. He didn't seem to be much 
into GUI programming, but he maintained it would be only ten lines...

How big is the GTK dialect implemenation?

If that is so, it is in some mystery Python binding that is hidden 
very well. In the known PyGTK binding, it would be 2.5 to 3 times 
as long, with a lot of boilerplate code inbetween

The binding is currently 900 lines. The dialect implementation is 
300 of that, but almost half of it is not used in this example

It will become much bigger when you support significant parts of 
GTK

Do you think that adding a symbol! type would be a big help for building 
such dialects at this point?

Yes, but it occurred to me a day or two ago that it can be done simpler, 
if you support first class constants, which I would like anyway

The gtk-position-center is the only #define in the example, the other 
words are function calls

So it's an integer, while there are already two integers for the 
size. Because they can be required to be adjacent, you can detect 
one more integer if it's at the proper distance of the pair, but 
that's the limit

You also have to use a complicated word, because defines of simple 
words could mess up the program elsewhere

To use center instead of gtk-position-center, you'd have to do:

center: 1

for example and then check for that code. But you'd get a manual 
sort of symbol table that would take space

If you could do

center: constant 1

and have constants be real types that you can check for, you'd be 
able to stretch simple dialects a lot further

Supporting first class constants would require to duplicate all existing 
types... I was thinking recently about adding a PROTECT keyword that 
would allow to mark any variable as immutable, so they would act 
as constants. But I guess that such option wouldn't help much for 
dialects, unless I would also add a PROTECTED? keyword to detect 
them.

Also, some kind of namespace support should also help implement more 
cleanly Red/System-level dialects. I might need that too for Red's 
runtime as it will grow much bigger than initially planned, due to 
Red compiling to Red/System.

Very nice Kaj!

cool, Kaj

Thanks

Doc, for dialects I only need constant integers. The immutability 
doesn't matter much, so it's more like the symbol! you suggested, 
but with control over the internal value

Namespaces would be good, too

Enumerations would also help; they're sort of a namespace type for 
constants

Note that to make the dialects more powerful, I'm making use of GTK's 
object system, that allows you to check types. So I can distinguish 
between different Red structs

If you would want to support that in Red/System itself, you'd need 
some sort of user defined type

I think that extending the RTTI system to distinguish different type 
of structs (or at least aliased structs) should be doable without 
that. Aliases are already a form of user-defined type.

Very nice Kaj, congrats!

Looks like we'll all have to install Syllable!

How's the Red timetable going?

Tick-tock.

Yes, very nice Kaj.

Nenad keeps the status updated on the RoadMap http://www.red-lang.org/p/roadmap.html


Since taking the decision for Red to compile to Red/System code, 
Nenad has been spending additional time extending  Red/System to 
be able to do so. It will probably result in less work being needed 
on the first Red static compiler.

Having more people testing red/system would probably help speed things 
up a little.

Agreed with Peter, as Red/System becomes more important now, it needs 
to be strengthen a bit more. I will publish an updated roadmap duing 
the SFD event in two weeks.

Detecting aliases would be very nice. It would replace GTK's RTTI, 
which is complex and buggy, and obviously wouldn't work cross-toolkit

Kaj: I have added in the last commit a new virtual path for querying 
aliases unique ID value: system/alias/<name> 


The system/alias/... paths are replaced during compilation with a 
unique ID that should match the one passed for "typed" functions. 
So now you basically have RTTI on aliases too.

As it might be a bit verbose in some cases, macros would be welcome 
to shorten them. In such case, the convention of a "type-" prefix 
on the alias name would be a good one to follow (will document that 
in the spec later today).

I also plan to add a SWITCH function soon, maybe before the SFD. 
It should be able to resolve alias names without the system/alias/ 
path prefix.

Let me know what you think about this solution.

Great, I'll try it

What I'm a bit worried about, though, is binary compatibility. How 
are the values of type IDs determined?

I suppose they shift when the program uses different aliases. That's 
no problem for monolithic programs, but it would be when you try 
to define an interface to a library or other subsystem

They are source-specific, they represent the order of each alias 
definition encountered by the compiler.

Yes, so they're not suitable for standard interfaces

Anyway, whatever value is attributed it should be transparent to 
users.

As long you as you use system/alias/ path accessors.

That's source compatibility, not binary compatibility