Re: Searching + Sorting in 3 milion documents

2005-11-03 Thread Manoj Kr. Sheoran
Dear Otis, Thanks for prompt reply. > - state of the index (optimized vs. unoptimized) Which one will be best for these sort of scenario? Optimized? > - amount of RAM (can your index fit into RAM?) and speed of disk > - desired response time Is there any OutOfMemory issue in case if RAM amout is

Re: Searching + Sorting in 3 milion documents

2005-11-03 Thread Otis Gospodnetic
Redirecting to java-user (-dev is for developers of Lucene). @daffodildb.com, interesting :) It really depends on a number of factors: - types of Fields in your Documents - number of indexed Fields - types of queries - state of the index (optimized vs. unoptimized) - amount of RAM (can your index

Re: Searching the contents

2005-11-03 Thread Manoj Kr. Sheoran
Hi Chris, Thanks for sharing your views with us. Is sorting works properly. Is there any idea regarding time estimation for sort on 4-5 columns with 1 million records? Regards, Manoj - Original Message - From: "Chris Lu" <[EMAIL PROTECTED]> To: Sent: Friday, November 04, 2005 12:30 PM Su

Re: Searching the contents

2005-11-03 Thread Chris Lu
One of my case is using PIII 450MHz 256M RAM, with 1million records, the search is around 3~4 seconds for the very first search, and sub-second, usually under 0.5 second. Sorting on one column will be longer, like 7~8 seconds for the first sort. And subseconds with cache. Chris Lu ---

Searching the contents

2005-11-03 Thread Manoj Kr. Sheoran
Hi, I am having 1,00,000 documents in a index but in near future it will be 3 million and more. I am somewhat concerned about the searhing process with this much number of document. I am giving order on some fields of documents. Could anybody tell be the expected result from lucene engine wit

Re: Reverse sorting by index order

2005-11-03 Thread Oren Shir
Hi, I've found the answer. Thanks to Andy Lee for giving me the idea to look inside Lucene's code. The line is new Sort(new SortField(null, SortField.DOC, true)) What got me confused is the null in the constructor, but I see that's the way it is used in SortField. Oren Shir On 11/3/05, Oren Sh

Re: Reverse sorting by index order

2005-11-03 Thread Oren Shir
I use 1.4.3 . Any other options? On 11/3/05, Andy Lee <[EMAIL PROTECTED]> wrote: > > On Nov 3, 2005, at 10:22 AM, Oren Shir wrote: > > There is no constructor for Sort(SortField, boolean) in Lucene API. > > Which > > version are you using? > > I think 1.9rc1. I have a pretty recent svn checkout --

Re: Reverse sorting by index order

2005-11-03 Thread Andy Lee
On Nov 3, 2005, at 10:22 AM, Oren Shir wrote: There is no constructor for Sort(SortField, boolean) in Lucene API. Which version are you using? I think 1.9rc1. I have a pretty recent svn checkout -- maybe this constructor is new. --Andy -

Re: Reverse sorting by index order

2005-11-03 Thread Oren Shir
Hi, There is no constructor for Sort(SortField, boolean) in Lucene API. Which version are you using? Thanks, Oren Shir On 11/3/05, Andy Lee <[EMAIL PROTECTED]> wrote: > > On Nov 3, 2005, at 9:37 AM, Oren Shir wrote: > > If I understand correctly, when sorting by Sort.INDEXORDER the oldest > > do

Re: Reverse sorting by index order

2005-11-03 Thread Andy Lee
On Nov 3, 2005, at 9:37 AM, Oren Shir wrote: If I understand correctly, when sorting by Sort.INDEXORDER the oldest documents that were added to the index will be returned first. I want the reverse, because I'm more interested in newer documents. Looking at the source, I see that Sort.INDEXOR

Reverse sorting by index order

2005-11-03 Thread Oren Shir
Hi, If I understand correctly, when sorting by Sort.INDEXORDER the oldest documents that were added to the index will be returned first. I want the reverse, because I'm more interested in newer documents. 1) Is there a simple way? 2) Will going over the Hits object in reverse order be much more

RE: Lucene or DB?

2005-11-03 Thread mark harwood
[ignore last message - result of a caffeine-starved keyboard slip] Doesn't sound like you have much in the way of Lucene-related criteria here other than the possibility of a fuzzy match on filename. The other stuff you mention isn't "relevance ranked" criteria like most text searches so it sounds

RE: Lucene or DB?

2005-11-03 Thread mark harwood
Doesn --- [EMAIL PROTECTED] wrote: > Thanks. I will take a look at those classes. > I do need to support search queries like: > - Find all files that are named foo.doc. > - Find all the files that have not been accessed in > last 6 > months(atime). > - Find all PDF files with size > 2 MB > > The