I just wanted to give an update on this. I rebuilt the index using
storeOffsetsWithPositions=true for the relevant highlight fields,
and I did see the size of the index increase, but I didn't see any
significant improvement in query response time.

Thank you,
maria

On Tue, Mar 19, 2024 at 9:46 AM Maria Muslea <maria.mus...@gmail.com> wrote:

> Thank you for your suggestions. I did read the reference guide and tried
> everything that I thought might help.
> I am using hl.method=unified and I do not have wildcards in my query, BUT
> I did not try storeOffsetsWithPositions=true.
>
> I will try this and see how it goes
>
> Thank you again,
> Maria
>
> On Thu, Mar 14, 2024 at 5:54 PM David Smiley <dsmi...@apache.org> wrote:
>
>> Hi Maria,
>>
>> Did you read the reference guide?:
>>
>> https://solr.apache.org/guide/8_11/highlighting.html#schema-options-and-performance-considerations
>> In particular, you're going to get the best performance with
>> hl.method=unified and storeOffsetsWithPositions=true.  I wouldn't
>> bother with termPositions or termOffsets options but termVectors is
>> great if you have wildcards in your queries; otherwise remove this too
>> as it'll be needless bloat.
>>
>> Try setting hl.fragsize=0 which can short-circuit some expensive
>> sentence fragmentation that I suspect might not be useful if your
>> values are only 200 characters.
>>
>> 2-3M values -- wow that's something.  Solr has to pull all that data
>> back per doc that's matched.
>>
>> Good luck,
>>   ~ David Smiley
>>
>> On Mon, Mar 11, 2024 at 5:53 PM Maria Muslea <maria.mus...@gmail.com>
>> wrote:
>> >
>> > Hi,
>> >
>> > I am having trouble with highlighting being very slow. I tried all
>> > suggestions that I could find online, but it is still much slower than I
>> > would like it to be.
>> >
>> > This happens on multi-valued fields that can sometimes have ~2-3M
>> values of
>> > ~200 characters each value.
>> >
>> > I am using hl.method=unified with SOLR 8.11.3, and the highlighting can
>> > take 10-20 seconds.
>> >
>> > I tried indexing all the fields that I highlight with:
>> termVectors="true"
>> > termPositions="true" termOffsets="true", but this didn't seem to help
>> with
>> > hl.method.unified or hl.method=fastVector.
>> >
>> > I tried hl.fragAlignRation=0.0 and also hl.snippets=1.
>> >
>> > Can you suggest anything else that I could try?
>> >
>> > I appreciate your help.
>> >
>> > Thank you,
>> > Maria
>>
>

Reply via email to