Boost logo

Boost-Build :

From: David Abrahams (gclbb-jamboost_at_[hidden])
Date: 2003-05-16 07:25:58


I've been doing some research on how to write a fully general
algorithm for deriving the best sequence of generators to produce
targets from arbitrary sources, and have convinced myself that the
problem is harder than I thought at first. At least, it doesn't map
well onto the domain I thought it did (natural language parsing)
because of

a) order-independence of sources (well OK, some NLP people do
know how to parse order-independent languages)

b) the fact that rules can produce more than one target (unlike
grammar productions). This one seems to be the killer for the
NLP analogy.

A little bit of research turned up these papers:

http://citeseer.nj.nec.com/heineman91rule.html
http://citeseer.nj.nec.com/tong94flexible.html

Which basically convinced me that we don't have time to try to solve
this problem correctly right now ;-)

I guess we can live with the strange "multiple" flag in the current
generators search for a while...

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com
 

Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk