World: r3wp

obviously we could alter the rules again to account for data-less 
tokens, but this would require a bit different structure.

great working ;-)

oops... in the above text...    " this rule"   should read..  "the 
next rule"

if you have time to show me how i am ok. but for now i must take 
my children to school and go to work. thanks a lot again

basically you have to create two complete (& and alternate) rule 
structures. and separate them with an "|" .


but you have to be sure that the first rule doesn't "pre-empt" the 
second one.... meaning that the first rule must not also match the 
second rule, or else, you will never reach the second rule.

for example..

some [["a" | "aa"]]


here we will never reach "aa" because "a" will be satisfied and the 
alternative will never be attempted ... so instead of matching "aa" 
you'd always match "a" twice .

where as specifying [some ["aa" | "a"]]  will always match "aa" IF 
there is still more than one "a" to parse... and will only ever reach 
"a" if the sequence is an odd number of "a" characters (or just one, 
obviously).


so "aaaa"  will match the "aa" rule twice, and "aaa"  will  match 
 "aa" then "a" .

IMHO, this is the basic premise of all of parse.   once you really 
understand how this applies to a rule which has sub rules... you 
really understand parse.... and then you can start doing more funky 
stuff.

Ladislav, your use-rule is effectively the same implementation that 
I had in mind when I made the USE operation proposal in the first 
place, but mezzanine instead of native. With the same overhead that 
made Carl initially reject the proposal when they were being implemented. 
I'm glad that you are making more headway towards getting him to 
accept it this time :)

Actually, both variants are implemented, even the one without the 
overhead (which I implemented first).

(or, to be more precise, maybe there is a possibility to make a variant 
not binding the rule at all, which would then deserve to be called 
"without the overhead" rather than any of my variants)

But, as you said, one of my motivations was to write it as a mezzanine 
to have some "inspiration"/experiences with it for Carl.

, since I guess, that this way, he will not have to just go into 
an "unknown territory"

I must say, that I was actually surprised, how people (including 
me) have struggled to circumvent this problem, while having such 
an elegant way available to solve it.

a regex question ...  ([0-9]{4})(-([0-9]{2})(-([0-9]{2})(T([0-9]{2}):([0-9]{2})(:([0-9]{2})(\.([0-9]+))?)?(Z|(([-+])([0-9]{2}):([0-9]{2}))))))

is apparently failing this string : 2010-10-18T07:06:25.00Z

What tool can I use to check this string against this regex ?

Regexlib has a different ISO-8601 date matching regex:
    http://regexlib.com/REDetails.aspx?regexp_id=2092

And the ability to enter any regex and target strings to test what 
happens:
    http://regexlib.com/RETester.aspx?

found this one too http://www.fileformat.info/tool/regex.htm

and it seems my string is passing ... hmm

The problem with regexes is they are impossible to debug.....Best 
just to rewrite continually until they work :)

I'm trying to validate some XML against an online validator and it's 
rejecting my dates :(

how do you specify an element to be of the type any-type! except 
none! ?

I am afraid, that you need to list all types excluding none

does R3 solve this? if not, maybe that would be a good problem to 
solve.

R3 can let you define that typeset and use it any time you like

ok, that is possibly good enough for generating specs.

I don't remember what all we did Henrik, but some of our test generation 
stuff on another world had some support for typesets IIRC.

Gregg, ok

Henrik, with a parse rule ?

Steeve, yes.

R3 does it

Graham, try http://gskinner.com/RegExrfor working out regexes. It 
has a really nice UI where you can hover over the components of the 
regex and see exactly what they do.

Thanks

Question on StackOverflow.....there must be a better answer than 
mine, and I'd suspect it involves PARSE (better answers usually do:)

    http://stackoverflow.com/questions/4093714/is-there-finer-granularity-than-load-next-for-reading-structured-data

Use fixed length records

Anyone got a parse rule that strips out everything between tags in 
an "xml" document

whitespace: charset [ "^/^- " ]
    swsp: [ any whitespace ]
    result: copy ""

    parse/all pqri-xml  [ some [ copy t thru ">" (append result t) swsp 
    to "<" ]]

Posted an answer mentioning the test framework, which does almost 
exactly what Fork asked

also, Carl's clean-script and script colorizer use parse + load/next 
to do the same thing. my Wetan uses the same method.

http://www.colellachiara.com/soft/MD3/emitters/wetan.html#section-4.2

basically, as long as you skip over [, (, ), and ] you can just use 
load/next. I'm also skipping over #[ because I want to preserve literal 
values while formatting (that is, preserve what the user typed)

How to use the new INTO parse keyword? Could it be used to avoid 
the temp parse like in this (very simplified example)?

  parse "<a>123</a>" [thru "<a>" copy tmp to "</a>" (probe tmp  probe 
  parse tmp ["123"]) to end]

Note that I know that in this example it's easy to use just one parse 
and avoid the temp.

INTO is neither new, not it is meant for string parsing

You can take advantage of using it when parsing a block and needing 
to parse a subblock (of any-block! type) or a substring

(of the said block)

can you give me a simple example, please?

>> parse [a b "123" c] [2 word! into [3 skip] word!]
== true

>> parse [a b c/d/e] [2 word! into [3 word!]]
== true

I understand now, thanks.

it's very useful, I woder why I've not found it earlier :)

The substring property is just a recent addition