Mongodb fulltext search problems

[Oleg Pavlov]

Hello!

I am using mongodb for full text search, but I've found one problem with it. I hava a database with ~300 000 documents.

When I make a query like this:

db.articles.find({ $text: { $search: {"word"} } }).pretty()

everything working well, but if I do

db.articles.find({ $text: { $search: {"\"10.5534/wjmh.2016.34.3.200\""} } }).pretty()

it works too, but very very slow (approximately 61 sec.). I add the method "explain()" and found this

 "winningPlan" : {
     "stage" : "TEXT",
     "indexPrefix" : {
 
 },
 "indexName" : "Article_TextIndex",
 "parsedTextQuery" : {
 "terms" : [
              "10",
              "200",
              "2016",
              "3",
              "34",
              "5534",
              "wjmh"
 ],
 "negatedTerms" : [ ],
 "phrases" : [
          "10.5534/wjmh.2016.34.3.200"
 ],
 "negatedPhrases" : [ ]
 },

It searches not only by pharses, it searches by terms too and I think this is a reason of the problem. 

Does someone know how to fix that? Or what can I do with it? Or maybe I do something wrong?

Thank you!

[Wan Bachtiar]

Does someone know how to fix that? Or what can I do with it? Or maybe I do something wrong?

Hi Oleg,

The text index stores one index entry for each unique stemmed term in each indexed field for each document in the collection. Although it’s a phrase search, the index entries are tokenised and stemmed version of the field value. See more info on Text Indexes

What’s your use case ? Looking at the value it looks like a code string. Would it be possibly to store in a separate field with a different index type ?

Could you provide an example document ? Also, which MongoDB version are you using ?

Regards,

Wan.

​

[Oleg Pavlov]

Hello, Wan. Thank you for your answer and I'm sorry it took me so long to reply. =)

Yes, the value is stored in the separated field, but I created a full text index for it, because my users would like to search by it in common way.

I think there is a solution to use separated endpoints for full text search and the value. It adds some over-engineering, but it will work.

среда, 25 января 2017 г., 4:08:46 UTC+3 пользователь Wan Bachtiar написал: