in solrconfig.xml:
10000
This is a max number of words in a field. How many words does one have in a long issue thread with comments? A Word count on this full issue suggests it has > 30 k on the first page: http://drupal.org/node/29706
solr will truncate longer texts, making the truncated part invisible to searching.
| Comment | File | Size | Author |
|---|---|---|---|
| #7 | 348832-maxfieldlength-7.patch | 1.64 KB | pwolanin |
Comments
Comment #1
JacobSingh commentedmaxFieldLength has serious impacts on RAM usage for Solr and hence performance.
I'm not sure what is the reasonable default here, but this is a consideration we can't ignore.
Comment #2
robertdouglass commentedSo maybe we eventually need to index comments separately? The disadvantage of that would be the overall relevancy of a node should be calculated on the basis of its entire text and all the comments. On the other hand, it might be beneficial to searchers to be able to link directly to the exact comment that matches the search.
Of course, for nodes with comments with pagination, this would still lead to the "can't link to a comment on page >2" problem.
Comment #3
JacobSingh commentedI don't see a course of action here. Please re-open if it becomes relevant again.
Comment #4
pwolanin commentedlet's reconsider, at least for 7
Comment #5
pwolanin commentedThis is also quite relevant w.r.t. attachment indexing.
Comment #6
pwolanin commentedDiscussed with Solr experts in IRC, and reading in the bookLucene in Action, there seems to be no real performance hit from increasing this. Obviosuly if you have many docs with huge fields, it will take more memory to index them. Lucene in Action suggests that you usually want to use unlimited for field length. A quick scan doesn't show me how to do that for Solr, but doubling to 20k is certainly feasible.
Comment #7
pwolanin commentedPatch to increase that value - please bump the version number to something in sync w. the release number
Comment #8
pwolanin commentedapplied this patch to 6.x-1.x and 5.x-2.x. Also just changed maxfield length in 6.x-2.x and HEAD without adjusting the version names yet, since we should likely wait for 6.x-2.0 for that?
Comment #10
sjeandroz commentedhello,
I had this problem, my nodes who are too long were truncated.
but when I increased this parameters, my nodes where indexed but for the part who was truncated before, the snippet's mecanism don't works...
My nodes appears in the search result but snippets were not built (only for the part who whas truncated before I increase this param)...
If anyone could help me please... (and sorry for my bad level in english...)
Thanks
Comment #11
sjeandroz commentedOk I solved my problem.
I fix the "hl.maxAnalyzedChars" to a very hight number and it's Ok.
Thanks