World: r3wp

Whoops, an error. Change:

        ["*" a: skip b: to "*" c: skip d: | "~" a: skip b: to "~" c: skip 
        d: ] :a (
to:

        [a: "*" b: to "*" c: skip d: | a: "~" b: to "~" c: skip d: ] :a (

Silly me :(

w: complement charset "*"
rule: [	
	to "*" here: "*"
	opt[ 
		copy item some w "*" there:
		(change/part :here join "" [<strong> item </strong>] :there)
	]
]
parse/all str [some rule]

Tomc, that will crash older versions of REBOL, and not work on newer 
versions. You need to reset the parse position to before the change, 
before the paren where you make the change. Otherwise parse will 
be referencing a point off the end of the string at the end of the 
paren, before you can reset it. This used to crash REBOL so bad the 
interpreter disappeared.

brianh   please supply a str that fails on current versions, so I 
can see what you mean

To fix your example, put a :here after the first there: in your rule.

still havent found a string that fails , trying all the combos of 
*'s at the beginning end , middle ...

In your case you might not have a crash, because you are replacing 
a short text with a longer one. Still, it's good to remember that 
bug for future reference. It really tripped me up when I first came 
across it, back when it still used to crash REBOL.

yes, shortening the string you are parsing would pull the rug out 
from under the interperter,
(and I was aware that the string was being lengthened) 

note: setting the parse pointer back to :here will position you before 
the "*"  

you may be better off  with :here skip to gaurentee progress in the 
case the change fails

OK, I tried this: parse "abc" [to "bc" a: "bc" (change/part a "b" 
2)]

It returns true on View 1.3 and Core 2.6, but false on View 1.2 and 
Core 2.5.0.

If the change fails it will throw an error. The trick is to put off 
the paren performing the change until you have gone through enough 
rules to ensure that the paren contents will succeed.

Remember, for many platforms, Core 2.5.0 is the current version.

Here's a simplified version of my example that can handle multiple 
instances of multiple markup types and be adapted to different end 
tags (thanks Tomc for the idea!):

markup-chars: charset "*~"
non-markup: complement markup-chars
tag1: ["*" "<strong>" "~" "<i>"]
tag2: ["*" "</strong>" "~" "</i>"]
parse/all data [
    any non-markup
    any [

        ; This next block can be generated if you have many markup types...

        [a: copy b "*" copy c to "*" copy d "*" e: | a: copy b "~" copy c 
        to "~" copy d "~" e: ]
        :a (change/part a rejoin [tag1/:b c tag2/:d] e)
        any non-markup
    ]
    to end
]

Tomc: "you may be better off  with :here skip to gaurentee progress"


Put the skip after the paren and I may agree with you there. Of course 
you would skip the number of chars in the replacement text then.

Wow, I'm off getting bored at meetings, come back and you've been 
working hard! Thanks, folks.

Here's what I have right now:

		markup-chars: charset "*_@"
		non-markup: complement markup-chars
		inline-tags: [
			"*" "strong"
			"_" "em"
			"@" "code"
		]

		markup-rule: [
			any non-markup
			any [
				[ a: "*" b: to "*" c: skip d: |
				  a: "_" b: to "_" c: skip d: | 
				  a: "@" b: to "@" c: skip d: ] :a (
					change/part a rejoin [ 
						"<" select inline-tags copy/part a b ">"
						copy/part b c 
						"</" select inline-tags copy/part a b ">"
					] d
				) any non-markup
			]
			to end
		]
		parse text markup-rule

you almost certinly want parse/all

whoops

If you want to guarantee progress with my and your examples (and 
better support multichar markup tags) change the last
  any non-markup
to
  any non-markup | skip
and that would do it.

okay, here's a slightly tweaked version that uses a multichar markup 
tag:

        markup-chars: charset "[*_-:---]"
        non-markup: complement markup-chars
        inline-tags: [
            "*" "strong"
            "_" "em"
            "@" "code"
            "--" "small"
        ]

        markup-rule: [
            any non-markup
            any [
                [ a: "*" b: to "*" c: skip d: |
                  a: "_" b: to "_" c: skip d: | 
                  a: "@" b: to "@" c: skip d: |
                  a: "--" b: to "--" c: skip skip d: ] :a (
                    change/part a rejoin [ 
                        "<" select inline-tags copy/part a b ">"
                        copy/part b c 
                        "</" select inline-tags copy/part a b ">"
                    ] d
                ) any non-markup | skip
            ]
            to end
        ]
        parse/all text markup-rule

Your first charset only needs one -

It passes my simple tests, now I need to throw more interesting tests 
in (multiple tags on the same line, nested tags, whatever)

Thanks BrianH, I'll fix that.

Nested tags of the same type won't work at all unless the start and 
end tags are different, and they won't work here without either recursion 
or a an algorythm that does nesting counts. Be careful with that 
because you'd have to update those counts in parens and you can't 
backtrack through parens.

Lucky for me, the rules don't support nested tags of the same type.


* *strong* text* would probably parse as <strong> </strong>strong 
<strong> text</strong>

Note that my last example keeps track of both the start and eng tags, 
even though I don't need to with the markup chars I used.

I need to test for *strong and _emphasized_ text.* (for example)

Yours will test for *strong and _emphasized* text._ as well right 
now (for example)

works beautifully, no changes needed. You guys rule.

The generated html might not be pretty though :)

Pretty comes after I know it works, if at all. :-)

awesome, it even works for "**bold**" text!

and won't touch ***********************************************

Really? The rules look like they'd translate that to <strong></strong> 
repeated many times. It doesn't?

Lemme write a test and see

ah, it turns them into <b></b>pairs

10 - Not OK - Rampant asterisks are usually ignored
Expected <<p>***********************************************</p>>

Got <<p><b></b><b></b><b></b><b></b><b></b><b></b><b></b><b></b><b></b><b></b><b></b><strong></
strong>*</p>

Part of me wants to just ignore it for now and get on to other stuff.

Well, if you want exceptions, you gotta code them in. In this case, 
before the block of your markup rules, as an alternate.

Like this (at the beginning of the any block):
    ["**" any "*"] |

(Be back, off to a class)

Well, I got things behaving by specifying " " at the beginning. It's 
a start.

ah the 2:46:49 post avoided that

with the opt[] block if what followed wan not a well formed marked 
up string

hm. Now I just need to figure out how to allow markup at the start 
of a line. I'll need to look at your code back there.

yay, got it working. It's ugly, but I got it working.

I'm not seeing a good solution to this at the moment, so I thought 
I would ask for help.  I'm working with parse and I want to create 
a new set of grammar rules dynamically based on the inputed grammar. 
 The simplest example I can present is starting with the following 
rule

	b: [19]

and in the end I want to have this.
        w: {}
	newb: [(append w "the b value is 19") 19 (append w newline)]


While I can insert an APPEND that doesn't contain values from the 
original rules into the block very easily, 

insert tail newb [(append w newline)]


I'm not sure of a way to insert a parenthesized APPEND with a string 
that contains info about the block.  I hope this is clear (or even 
possible), but please ask me questions if it is not.

I suppose the question I'm asking is there a way to force the arguments 
of an expression to evaluate without actually evaluating the expression. 
 Taking the aboe example

a: join "the b value is " first b

insert head newb [(append w a)]

where I want to evaluate 'a while ending up with the expression:

insert head newb [(append w "the b value is 19")]


This seems contrary to the nature of first order functions, but I 
just wanted to check.

Excuse me, first - class functions

interesting questions, unfortunately not having time to answer any

a: 19 append/only [] to-paren reduce ['append  'w join "the b value 
is " a]

or

a: 19 append/only [] to-paren compose [append w (join "the b value 
is " a)]