World: r3wp

*

Steeve;  Sunanda uses it in his world class index builder SKIMP. 
 rebol.org  skimp.r  It's a heady read I might add.  After giving 
it a glance ask yourself if Sunanda cut his teeth on personal computers 
or Big Iron.  :)

'alter is short for  'alternate (ie alternate between two states).

It's a shorthand way to add an item to a series if it is not there, 
or remove it if it is.

I may create strange data structures, but I use 'alter all the time.

Steeve, I actually used it once the other day.

I think I used it like this:
	if not find block item [alter block item]

But it seems an inefficient way of doing things.

I always found ALTER less than useful. Never used it.

Hi Anton, did you 'source 'alter?

alter: func [

    {If a value is not found in a series, append it; otherwise, remove 
    it.} 
    series [series! port!] 
    value 
    /local temp
][

    either temp: find series value [remove temp] [append series value]
]

So, you'd better off with:

if not find block item [append block item]

interesting

But I think, that the return value of alter is less than usefull, 
There's no way of knowing, whether the value has been added or removed 
... other than doing a 'find on it afterwards ... ;-)

true.

Sorry, I did not use the example I gave above. Actually, I used alter 
like this:
	alter opened-rows count

That was probably my first use of alter ever.

Now, that seems more in line, with how it's meant to be used.

a proposal of a refinement for alter:
alter/count 

increment a value instead of add/remove it, return the total count 
for this value.
>> alter/count [] 'a
== 1
>> alter/count [a 1] 'a
== 2

Instead of ALTER functionality I use this quite a lot... but I'm 
not sure I would use funtion with refinement for this as I use it 
in loops where speed is important.

But if you need it, you can use something like that:
alter-count: func[
	block [any-block!]
	value [any-type!]
	/local temp count
][
	either temp: find/tail/skip block :value 2 [
		change temp count: temp/1 + 1
		count
	][
		append block reduce [:value 1]
		1
	]
]
(the name of the function should be probably different)

It cannot be part of ALTER function as ALTER can be used with any 
series. I'm not sure how you could strore the counts in bitsets or 
strings;-)

one-liner version:
alter-count: func[
	block [any-block!]
	value [any-type!]
][

 pick change block: any [find/tail/skip block :value 2 insert tail 
 block :value] 1 + any [block/1 0] -1
]

More than ALTER, I would like
	append-if-necessary
	remove-if-necessary
(but probably with better, shorter names)

huh ?

yeah! i posted my first modification on devbase

even if i have don"t access to alpha T_T

hum * i haven't an access

i have some difficultes to construc devbase.r , missing icons in 
the title, any idea ?

sorry , i got no problem in fact (i just have to open my eyes)

everything should download automatically

These alter-count functions aren't taking R3 into account. Most of 
these key/value usage patterns will be handled by the map! type in 
R3.
; alter-count, just code since it's too simple for a function
key-counts/:key: 1 + all [key-counts/:key 0]

Whoops: all -> any

not agree, how do add new key with your code ?

Like that.

Doesn't work in R2, but this is the !REBOL3 group, isn't it? :)

The trick is that you can't remove keys, but if you assign none to 
the map at a given key the effect is the same.

i see

The general pattern in R3 is that none is the equivalent of missing 
data. Pick off the end of a series returns none too.

more tolerant functions

throw less errors

Recovering from a none is easier and more efficient than recovering 
from an error. Series bounds are just an implementation detail anyways, 
when you have series that can autoexpand.

Does that mean I won't be able to build the equivalent of an R2 hash! 
with map! in R3?


I use a hash! to build a unique list of words as it was far quicker 
than any other method I could come up with.

Yeah, hash! has been replaced with map!, which is faster but more 
specialized. You could either use block!, assign true to the map! 
at the word key, or keep a word count.

this works in R2

 key-counts/:key: 1 + any [select key-counts :key  do [repend key-counts 
  [:key 0] 0 ] ]

key-counts/:key: 1 + any [select key-counts :key do [repend key-counts 
:key 0] ]

forget the second

Thanks for the suggestions: I think that I'll have to use  a map! 
with a dummy value as using a block! was too slow.

I'm missing hash! in R3 because with map! I cannot use integers as 
keys... I hope it will be changed.

It's hard in hash! as well, but I agree. There is some missing infrastructure 
around the map! type as well, like iteration and searching.

has the conversation moved elsewhere?

no, it is just some kind of slow-down period. Activities around R3 
should be renewed as we speak, it seems Gabriele and Cyphre will 
be back on R3 full-time soon. There is also draft of release strategy. 
It seems RT decided (upon suggestions) to divide R3 release, so we 
should see Core like R3 alpha in one month, View will come later 
(2 - 3 months)

I should have figured... December is a slow month

I gleaned some good news from Docbase: work has started on implementing 
unicode support :-)