Re: [bdbxml] Regular expressions in bdbxml

Dan Brian wrote,

> It's impossible for any regular expression index to provide lookup
> speed close to the current index types.

True, but then even moderately close is better than a long way off. From my
perspective as an incurably docucentric XML-er the current indexing
facilities in bdbxml stop at a threshold which my applications need to
cross.  I know that soon they will be extended to the sub-document level I
need, and it seems to me that it would be hard to justify not beefing up
regex execution by working it into the new indexation.

>  And even then, it's usually
> just for POSIX 1003.2 regexes, as opposed to PCRE

I think users would expect the indexing support to be on W3C schema regexes
as per the XPath2 and XQuery specs.
http://www.w3.org/TR/2005/WD-xpath-functions-20050404/#regex-syntax
Little point in the indexed-enabled regex support being a different flavour
from the standard-compliant one already supported, though not currently
index-assisted.

> I use BDB XML for a similar application to yours, and we don't need
> regexes. Substring are adequate to do most of our queries (where we
> are looking up words). The exception is if you want to look up
> misspelled queries.

Ah well, you're obviously lucky with the natural language(s) you
applications need to support. But it's a different game with medieval
insular French (massively variable orthography, so every query has to be
treated as a potential "misspelling" since there is no canonical form) or
Tagalog/Pilipino (where the lexical items you need to match in your data are
not the derived forms found "in the wild" and input by the user but roots
subject to sometimes extensive morpho-phonological transformations -
affixation often munges the root and modifies the affix(es) at the same
time). From what I picked up nearly 40 years ago in Linguistics 101, I
suspect Turkish may be even more of a moving target where matching input
terms against lexical items is concerned, and that's a language that a lot
of EU contracts are going to start requiring programmers to handle Real Soon
Now. Some of this can indeed be managed via subqueries, but I wouldn't like
to do that without support from a reasonably speedy regex implementation.

Michael Beddow



------------------------------------------
To remove yourself from this list, send an
email to [email protected]