World: r3wp

There is new blog about small change to modules - http://www.rebol.net/r3blogs/0179.html

Eh Ammon ! how do you manage selection over several text gobs ?

I'm trying to do it currently, but i'm not sure to have the best 
design

I haven't gotten that far yet.  I recently had to rewrite my parser 
from the ground up, mostly due to the fact that NOT isn't implemented 
yet and I want to be able to handle unicode without too much trouble. 
 I think I'm going to have to handle a lot events like text selection 
by passing them through the text faces to the main scrollable area 
and let it handle them.  I expect I'll be rewriting the highlighting 
from the ground up.

i'm rewritting completly the key handler with my own cursor (a gob). 
I don't like the default handler

I like the design of the default handler.  =D  I'll need to modify 
or extend it before I'm through though.

too much ovverrides have to be done, redoing it from the scratch 
is simpler

You know that you can get the source from DevBase, right?

my cursor is gob i'm moving from gobs to another gobs

i have the sources

Yup, I'll be doing that as well.

in the default design, the cursor is not a gob

well, the handling of text editing is very powerful.  I will definitely 
be using the same design just extending it a bit.

i don't like it so much ;-)

but perhaps i'm wrong

It's a simple state machine.  It's taken me a long time to grok what 
can be done with such a system but now they are starting to dominate 
my code.

A state machine ? What a big word. It's just a serie of switch commands 
:-)

Instead of having to add a lot of refinements and refinement handling 
to related functions you can just put them together into an object 
and as each of the functions are called they are setting values that 
are local to the collection of functions and each of those values 
modify the behavior as if you were using refinements and passing 
in more values.  It's kind of an inference engine.

Again, what a big word. I do that too to have a more readable, and 
compact code. 

I don't like to pass thousand of parameters. But nothing is related 
to a state machine to my mind

You may be right.  I may be misinterpreting the meaning of state 
machine...

I'm seeing some strange behavior with Parse...

ws: charset rejoin [space tab newline]

parse "   This is a test!" [s: some ws e: (probe copy/part s e) to 
end]
; == false

parse/all "  This is a test!" [s: some ws e: (probe copy/part s e) 
to end]
; "  "
; == true

I shouldn't need /all when I provide a rule block should I?

In truth it doesn't always fail but I'm not sure what causes it to 
change it's behavior.

Heh.  Or rather I don't know why SOME was ever working but I saw 
it work once.


parse "  ^- this is a test!" [s: any ws e: (probe copy/part s e) 
to end]

ANY appears to work all the time though.  Strange.

charset is bugous in R3 currently

So this is a problem with Charset not Parse?

I made my own function.
charset: func [str /local bitset][
	bitset: make bitset! 256
	foreach char str [append bitset char]
]

to be right, is used this to make complement of charsets

you made a mistake bout the understanding of /all refinement, see 
help

You're not making much sense.

Sorry, i mean, forget my charset function. your problem is the missunderstanfing 
og /all refinement

/all -- Parses all chars including spaces.

if not, spaces are skiped.
spaces are : space, newline, tab

Yeah, I read that part.  For some reason I thought parse didn't skip 
whitespace if you provide a block but this clearly shows otherwise...


>> parse " a b c ddd " [ "a" (print 'a) "b" (print 'b) "c" (print 
'c) some "d" (print 'd)]
a
b
c
d
== true

But why does ANY work without /all while SOME doesn't???

because any is an optionnal rule not some

basics

That doesn't make sense to me still.  If the parser is skipping whitespace 
unless you use /all then any should fail the same way that some does.

No, Optionnal means that if ANY fails , the parsing process continue 
with the next rules

Ah, but ANY didn't fail.  The paren was evaluated which means that 
ANY succeeded.

show me your code

But whatever.  I have what I need and as long as parse is behaving 
as Carl intended then, whatever...

I posted the code a few messages back.

Then, it's not related, the parent is not associated with the ANY 
rule. So that, the ANY rule can failed and the parent will be executed 
each time

cardinality of ANY is  0,N
cardinality of SOME is 1,N

I changed my mind, i uses the default caret and highlight behavior, 
much simple.

Ammon some doesn't work for SOME reason   ^^ .... probably  ^^

Most widgets are rectangular in shape ... any reason for this?  Are 
other shapes feasible?

So far, there is not much widgets, which could have different shape, 
no?

Lack of imagination?

edge / hgit testing is the main reason

easier computationally?