[PEPr] Comment on RFC::VectorIndex

Denny Shimkoski (http://pear.php.net/user/denny) has commented on the proposal for RFC::VectorIndex. Comment: I misunderstood the RFC designation. I'll remember this for any future proposals. In the meantime, I'm going to move this proposal to the Math category. I just took a deeper look into the Vector classes. The only problem I see is that the Tuple class doesn't preserve array keys. This turns out to be a deal breaker, since VectorIndex depends on the keys to define the so-called vector space (each new word is a new dimension). For each new word encountered, the vector space grows accordingly. Frequency data is stored on two levels -- 1. Globally $wordCounts['word']['master_id'] 2. Locally (document-level) $wordCounts['word']['document_id'] The global vector ensures that all local vectors share the same "space". The local vectors cut down on storage requirements. Could you please elaborate on the comment about limit cases? Do you have any relevant links to share? Any pointers would be much appreciated. The current class includes a toArray() method to account for precalculation needs. The output can then be loaded into another VectorIndex object using the fromArray() method. Thanks much for the feedback! Proposal information: http://pear.php.net/pepr/pepr-proposal-show.php?id=414 -- Sent by PEPr, the automatic proposal system at http://pear.php.net