Storage of norms array

[Narendra yadala <na...@gmail.com>]

Hi All

Currently I am using document.setBoost for setting a boost proportionate to
some other parameter.
For ex: p is my parameter whose value is 5.0f
I am using document.setBoost(p) for making this p as a boost for this
document.
But since document.boost gets encoded inside norms array, so I am forced to
store
the norms array for all the fields in my document and as a result
I am wasting space of 2 bytes for two fields for which I dont need any
length normalization and stuff.
And also my parameter p does not exist for all the documents in the index.
So now if I have million documents and only 2% has this parameter p then I
am storing
980,000*3 extra bytes considering there are 3 fields in my document.
So what should I do considering the space considerations in my context.

Regards
Narendra

Re: Storage of norms array

[Erick Erickson <er...@gmail.com>]

I think you shouldn't worry about it until you can demonstrate that
you actually have a space problem. Worrying about storing
an extra < 3M is almost certainly a waste of time.

What evidence do you have that this is any kind of problem
at all?

Best
Erick

On 10/9/07, Narendra yadala <na...@gmail.com> wrote:
>
> Hi All
>
> Currently I am using document.setBoost for setting a boost proportionate
> to
> some other parameter.
> For ex: p is my parameter whose value is 5.0f
> I am using document.setBoost(p) for making this p as a boost for this
> document.
> But since document.boost gets encoded inside norms array, so I am forced
> to
> store
> the norms array for all the fields in my document and as a result
> I am wasting space of 2 bytes for two fields for which I dont need any
> length normalization and stuff.
> And also my parameter p does not exist for all the documents in the index.
> So now if I have million documents and only 2% has this parameter p then I
> am storing
> 980,000*3 extra bytes considering there are 3 fields in my document.
> So what should I do considering the space considerations in my context.
>
> Regards
> Narendra
>