World: r3wp

Hmm... How to do that?
We need to know where a particular
Maybe:
1. Read script *and* Load script
2. Visit each item in the loaded block, recursively.
3. As each item is visited, check its type.

4. Depending somewhat on type, parse (in the READed script) to the 
molded item:
4.1  If it's a series, search for the "opener", eg. block! -> "["
4.2  If it's a non-series, search for it molded.
4.3

(oops.. incomplete post, sorry)

We need to know where a particular .. value is in the original source 
string (read from file).

4.3 (Cater for variations in molding, eg. strings in the source written 
as "..." can be molded {...} and vice versa, depending on length.)

Interesting idea. Brian. We could certainly weight the results so 
those with words in the body rather than strings etc are listed higher....In 
fact, we already do that as one of the ranking factors.
Let me think about it....

When I am making changes to the mezzanines, I use the script library 
for research on existing usage of these mezzanines to see whether 
anything would break. When the name of a function is a common word 
that would show up in comments or strings, this is much more difficult. 
I would also like to look for third-party functions that I can't 
remember where they are defined or their author.

Would it be useful to index the files? That way you could extract 
keywords on posting rather than at search time.

We do index the files, and we have several indexes, some built by 
parsing parts of the script. So we can, usually, search for special 
parts of scripts -- like strings or comments.

Part of the difficulty is that there are some scripts that we cannot 
do a 'load or a 'load/header on - they are broken or have an incompatible 
'needs: header. So we cannot add those scripts to the sepcialised 
indexes. They need specialised handling :-)

Watch this space -- I'm fiddling with some code that may partially 
need your need.

Can't load ? Just skip it with with a warning message :) You still 
get your 99% solution.

Oh, sorry, I don't know how many are not loadable in rebol.org. I 
just loaded my entire source base after fixing one that was unloadable.

but I am obviously not taking public submissions like rebol.org.

If we can't load, we try a few tricks (like commenting out the needs: 
header and reloading).

But still, there are one ot two that we fail with. So they don't 
get as well indexed as the others...And (for the same reasons) they 
are the ones that appear in black and white even if you click the 
"view in color" link.

How easy is it to edit your scripts after they have been submitted? 
Do you feel comfortable adding the appropriate needs header where 
needed, provided you know enough about REBOL versions to determine? 
Can you flag the appropriate files as problematic?

For that matter, do you index line comments separately, as LOAD filters 
them out?

It's pretty easy to update a script after you've contributed it.....Use 
the [update] link to add a new version.

Yes, we also index comment lines separately. Basically, we use the 
same logic as Carl's color-code.r to analyze the parts of a script:

  http://www.rebol.org/cgi-bin/cgiwrap/rebol/view-script.r?script=color-code.r

Brianh: <I am only interested in searching the various word and path 
types.>

Try this quick concept code. In REBOL.org's search box, use [b] as 
a sort-of inline refinement to limit script searches to just the 
body of a script, eg
     sing

  -- finds nearly 300 scripts which contain the string "sing"    but
    sing [b]

-- finds just three scripts that have the word-part "sing" in their 
body

.

Even better script searching at REBOL.org -- we've extended the "[b]" 
notation to include other parts of a script and any tags the script 
has. The search can now be highly tuned to what you are looking for:
http://www.rebol.org/boiler.r?display=introduction.html

BTW: Is it possible to search by script comments?

Do you mean comments _in_ a script; or comments about a script (ie 
the discussion threads)?

Either way, yes -- check the link in my previous post for the syntax 
and some examples.

In the script. But cool that it's supported.

Is there something going on with packages?  I uploaded a new one 
several days ago and can't easily redownload it (for testing) - repack.r 
keeps giving me "ERROR 306:  not available right now"

nothing happens if I click on "help", either..

It's not a clear message -- sorry......I think it is because the 
package is not set as being available for downloading:

http://www.rebol.org/package-information.r?script-name=arff-datamining.r

Yeah, 'tis my own mistake.. oops

Not really, it's a default which is set backwards to that which many 
people would expect. Plus an unclear error message. So my apologies 
again.

I found some wrong links in http://www.rebol.org/documentation.r?script=makedoc2.r

The document is uploaded and owned by Carl. Is it only him, who can 
change the links?

Any of the Librarians can change any document and/or script.....But 
we tend not to do that without the owner's permission.

Please PM me the changes, and I'll get it sorted one way or another.
Thanks!

Welcome back...

Thanks, Reichart.

Geomol -- Thanks....I've made the changes and notified Carl that 
his documentation has been updated.

Doing a search, "Author: Maxim" doesn't briing up his newest script, 
Blood.r   The author field in blood.r is the same as other scripts 
which do show up with the above search.  A search for "Liquid" does 
return blood.r.  Is this a case of "hasn't been indexed yet"?

probably.... use the "latest scripts" link.

if you just put maxim in the search field, blood is in the list

Maxim, I have the script.  Playing with it now.  If I was directing 
this question at you I would have put it in the !Liquid group.  ;-) 
 This is a question about the behind the scenes opperations of rebol.org...

I know   hehe

Ammon -- some of the indexing happens in the background, so can trail 
12 hours or so behind a script being updated.

In the specific case of searching for an author name, the syntax 
you need is different:
    [author//maxim olivier-adlhoch]
Help for searches is here:
   http://www.rebol.org/boiler.r?display=introduction.html

Of course my syntax was wrong.   It probably fairly common for someone 
to search the library for a specific set-word! and even if it's not, 
we definitely want to allow for it.

I expected granular indexing as well...

Thanks for the info!

Guys, at the moment, rebol.org is soooo slow here, something like 
10 sec to appear. Is it just my experience here, or are there any 
problems? ...

Nope.  I'm actually getting rebol.org in about 2 seconds but I'm 
seeing about 10 seconds of lag on this world.

Sunanada, I have another question for you.  While I was poking around 
the library earlier I noticed that I have named some of my scripts 
very poorly and would like to change the FILE: value in the header, 
which apparently just creates a new copy of the script.  Is there 
a way to remove the old script such that I really am renaming the 
script rather than uploading a new one?

Also, the link to the list of license tags from the [license] link 
of a script is broken.  You have an extra slash in there.

OMG!! I just discovered there is an online DIFF engine on rebol.org 
!!! WOW... sunanda... you must provide that functionality upon script 
update, this is too awesome to hide away in an easy to miss set of 
links in the script statistics.

I mean, if we could view the diff before pressing "update" we would 
have something starting to look like an integrated VCS  :-)

I second the motion!

the interface is already ubber clean  :-)  


if I had this on rebol.org update, I think It would make the update 
process even more appealing for potential new users.

Ammon -- I can rename existing scripts. It's a moderator-only function.
If you want it done, please let me know which ones.

Ammon -- thanks for bug report.

Maxim -- thanks....the diff function code is an embarrassing collection 
of random hacks. I'm hoping someone else will write a better version 
that could be made public :-)

well what I am seeing impressed me ... I wasn't expecting that on 
rebol.org... really you should allow us to compare right in the update 
page.  I know I'd use it every single time  :-)