Re: find and replace, and other questions

On 12/13/10 6:26 AM, "[email protected]" <[email protected]> wrote:

> Ahhh, it's been a long time, eh?

Yea, and back again.  Maybe not to be Blamed this time.  We'll see.  ;)

> First off, the "\-constants" mentioned in the manual aren't the same as
> the back reference \1, \2, etc.

I was just hoping the writing had been sloppy--since a digit looks like a
"constant".  ;)
 
> The \-constants are things like \n (newline), \t (tab) and \xHH (hex
> char) and \oOOO (octal char) are constants in the "C" language sense.
> They exist solely so that it's easy to put in a single byte wide
> character of whatever value desired easily.
> 
> But- the \1, \2, etc. values are REGEX-engine back-references.  They
> aren't constants; the regex engine is constantly rebinding them as it
> tries to do the first-longest REGEX match.  Thus, they are in a state of
> flux until the regex completes.  Further, the regex engine doesn't
> explictly export them back; rather the parenthesized variables _are_
> exported back, explicitly, and Ville Laurikari (the author of TRE) is
> very careful not to break that part of the interface..
> 
> Note also that \1, \2 etc. are _usually_ but not _always_ mapped to the
> "bound variables" you get out of a match statement.  That's because in
> TRE (and thus in CRM114) it's possible to set a "no bind" on a
> parenthesized subexpression (I'm trying to find that; I recall seeing it
> but can't find it now).

Just to be complete and clear, in spite of your following admonition, the
value in ALTER's // is a "var­expanded string" which I think is none of
TRE's business, right?

> So- the bottom line is that it might be possible to do your \1 modifier,
> but it would be fraught with peril, and not stable with respect to
> Ville's work.
>
> Thus, I'd say "find a different problem to work on".  :(

However I think I already have my solution.  I never cared about the syntax,
I just wanted a solution.  It couldn't be done with *no* changes to
mailfilter, but the changes were less than I thought when I used the word
"hack".

I did it for non-line-spanners only.  The prototype fix required changes to
two lines:

    match [:m_text:] <nomultiline> (:place:) /:*:fr:/

  BECOMES:

    match [:m_text:] <nomultiline> (:place: :_1: :_2:) /:*:fr:/

and

    alter (:place:) /:*:to:/

  BECOMES:

    eval (:place:) /:*:to:/

And then I can freely use () groups on the "LHS" of >-> and :*:_1: etc. (as
far as I care to support) on the "RHS".  Not sure if using eval there has
any gotchas, but trivial test cases work.

Voila!

And probably I can (and really should) make it more efficient by evaluating
:to: only once outside the inner loop of the rewrite processing, but haven't
tried yet.

And a clean way to do it might be to provide something like >=> instead of
>-> to support "RHS" evaluation and maybe :_1: thru :_9: .
 
> re: the library:
> 
> And yes, our major effort this year has been to create / improve the
> C-callable library.  For a lot of users (especially the user who's
> paying for this work to be done!), the ability to integrate tightly (and
> the speedup when you change from the CRM114 JIT language compiler to a
> C-callable subroutine call) is really important.
> 
> The library is LGPLed, and on the download page of the web site.  Right
> now I'm making an improved tenfold-validate program to test yet another
> new classifier- a boosted RULAX (Are You Like An X) classifier.  The
> idea is to use boosting (an array of very simple but weak classifiers)
> to build a strong classifier by voting.

Sounds good!

-Kurt



------------------------------------------------------------------------------
Lotusphere 2011
Register now for Lotusphere 2011 and learn how
to connect the dots, take your collaborative environment
to the next level, and enter the era of Social Business.
http://p.sf.net/sfu/lotusphere-d2d
_______________________________________________
Crm114-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/crm114-general