World: r3wp

Thanks for those references.

Parse question on StackOverflow -- not yet answered:

http://stackoverflow.com/questions/1415340/rebol-parsing-rule-how-to-correct-the-rule-to-separate-paragraphs

Three answers now.

Steeve - move parse discussion here.

So... does such a function take an argument, such as the current 
index for the series being parsed?

That's how REPLACE works...

REPLACE?

i would say, no functions with parameters allowed

but if you can perform a do/next then parameters are allowed

Yeah. If you pass a function as the replacement value, that function 
will be called with the series at the replacement position as a parameter, 
and its return value is used. ARRAY does something similar too. My 
changes.

Say that the series position parameter is passed with APPLY rules. 
If the function takes a parameter it sees it; if not it doesn't.

Or we can provide 2 index by default, maintained by the parse engine.
& = the head of last rule matched
&&= the tail of the last matched rule.

Not a bad idea, but... this is how it starts. This is what led to 
the rule! type suggestion :(

See, your method would allow REPLACE/part to be called directly.

Sorry, REMOVE/part.

or even INSERT, REMOVE, CHANGE, without the need to develop a specific 
inlinned method for those functions

wer just need 2 pointers auto-handled by parse

Except your replacement code for those functions was wrong. And would 
be wrong in this case. Those inline operations were added to reduce 
common errors, not to provide missing functionality.

I am also concerned about the security implications of having PARSE 
call functions outside of parens. In parens you know what you're 
getting. This is why QUOTE and IF require parens for the REBOL code 
they execute.

All of the added operations could have been done before with code 
in parens and/or explicit position setting. It's easier this way.

I put a note up because of my silly misunderstanding of the intent 
of adding AND to PARSE.  But I get odd results with the likes of 
   parse "abeabd" [and [thru "e"] [thru "d'"]]  which behaves like 
ANY

Not a silly misunderstanding, a bug, bug#1238 in particular.

One of 4 parse bugs in a83.

OF course in STSC APL  "laod" was as good as "load" and in Smalltalk 
I still long for "slef" and "sefl" but I draw the line at "elfs" 
which is clearly unfit in the age of "octopuses"

And that doesn't even count the stuff not implemented yet.

I thought ONE (but no move) on the model of SOME and ANY when I was 
misunderstanding AND as "all"  as [ ONE [rule1 rule2 rule3 ] ]

ONE would be an interesting name for the OF proposal.

Shorter than UNIQUE

Sorry, that would be a different operation.

Shorter than FROM or GATHER though.

Wait, ONE wouldn't work, since you actually get all of the rules, 
just not in order.

I posted examples to show why a84 is more rebolish. http://www.rebol.net/cgi-bin/r3blog.r?view=0255#comments

looks good ...

Steeve - will better parse help with editor? IIRC it was you and 
Shadwolf, who did it?

A little, but the main drawback i have with parsing in editor is 
that parse doesn't handle incremental parsing.

Because i do parsing line by line to be able to parse only modified 
lines.

So that, i have to rewrite all the rules (describing the document) 
in an obfuscated way to deal with incremental parsing.

BrianH has some plans towards "streamed" parsing. Hopefully it can 
be done over ports in the future ...

Actually, we can simulate streamed/incremental parsing. But we need 
to transform all the input rules (it can be automatized).
I would prefer an inlined behavior of parse for such purpose.

It's why i asked to Carl if we could return the rule stack during 
parsing (i.e. with a special command).

What was the answer? I missed your request during yesterday's chat 
probably ...

Request delayed for R3.1

3.1 - that will be in Spring, so I think that we can wait ...

it seems

parse is in mode parse/all by default now, but we can remove the 
option, a bug ?

*we can't

/all still affects simple parse (splitting instead of rules).

With the PARSE changes, this puts us within the range of a parsing 
model with a reasonably solid theoretical foundation and a lot of 
experience with parser generators. We could compile PARSE rules into 
equivalent incremental rules.

yep we could start by building a compiler of rules into  incremental 
ones (with rebol i mean)

i made some tests in the past

About STAY, i don't see the interest to continue even if the following 
rule is not matched . Can someone give an use case ?

because when i  do this:

>> parse [1] [stay skip ?? to end]
to: [1]
== true

or

>> parse [ ] [stay skip ?? to end]
to: [ ]
== true

it's like doing:
>> parse [ ...] [?? to end]

STAY have no purpose to my mind...

It's just a little annoying, see:

>> parse a: "12" [remove copy val skip] print [a val]
==2 none


So, the [remove] treats well the [skip], but discard the content 
of [copy val]

Now, see:

>> parse a: "12" [remove [copy val skip]] print [a val]
== 2 1


The content of [val] is preserved in that case, don't know why here, 
but not above...

I would expect 'val being 1 in both cases .....