Thanks for the responses. For fault tolerance, we have decided to set the
"mergeFactor" to 1, ie, we want to update the on-disk index every time a
document is added via addDocument(). My question was that will this create a
new segment every time? Or is there a separate "segmentFactor", such that an
incremental index update updates an existing segment *on disk*, only
starting a new segment once a given number of documents have been added to
the index?

Thanks,
John

On 4/23/06, Erik Hatcher <[EMAIL PROTECTED]> wrote:
>
> If you use the compound format, all "files" are kept inside a single
> filesystem file.
>
>         Erik
>
>
> On Apr 23, 2006, at 2:13 PM, John Paige wrote:
>
> > So, if I use one indexwriter instance to index one document, will
> > it create
> > a segment per document?
> > How many files per segment get added if I use compound index file
> > format?
> >
> > Thanks,
> > John
> >
> > On 4/23/06, Erik Hatcher <[EMAIL PROTECTED]> wrote:
> >>
> >> If you use the compound index file format (the default since Lucene
> >> 1.4) you'll avoid the file descriptors issue.  If you add 10
> >> documents at one time with a single IndexWriter, you will not create
> >> 10 segments, only one segment (generally speaking, based on the
> >> default segment factors).
> >>
> >>         Erik
> >>
> >>
> >> On Apr 23, 2006, at 9:48 AM, John Paige wrote:
> >>
> >>> Hello all,
> >>>    In my application it is required to build an index for each
> >>> user. We need
> >>> to add documents to the existing index frequently.
> >>> We cannot use RAMDirectory to create a RAM index and merge it
> >>> with the
> >>> FSDirectory index later on based on the mergefactor. We need to add
> >>> each
> >>> document in the FSDirectory based index.
> >>> We are afraid of getting into the "out of file descriptors" issue ,
> >>> so my
> >>> question is, does lucene create a segment for each document that
> >>> gets added
> >>> to the FS based index i.e. will adding 10 documents to the existing
> >>> index
> >>> create 10 segments?
> >>>
> >>> Thanks,
> >>> John
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: [EMAIL PROTECTED]
> >> For additional commands, e-mail: [EMAIL PROTECTED]
> >>
> >>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>

Reply via email to