Mongo Lucene Regex Search

MongoDB Atlas Atlas Search
,
1

Hi,

I’m running following query on atlas cloud

[
  {
    $search: {
      compound: {
        must: [
          {
            text: {
              path: "site",
              query: "test",
            },
          },
          {
            text: {
              path: "objectKey",
              query: "product",
            },
          },
          {
            regex: {
              path: ["name"],
              query: "(.*)yearly(.*)",
            },
          },
        ],
      },
    },
  },
  { $sort: { displayOrder: -1 } },
  { $skip: 0 },
  { $limit: 10 },
]

Here I’m trying to match the name using regular expression. “name” field is indexed using lucene.keyword. I’m expecting that it will do case insensitive search but it’s not. If I do “Yearly” then it works but with “yearly”, it’s not working.
Do we need to do any changes in index so it can work with case insensitive?

Thanks,

2

lucene.keyword indexes the original value as-is, case-specific. To achieve case-insensitive on a full string like that, create a custom analyzer that uses the keyword tokenizer followed by a lowercase filter.

I would caution you about using a regex with leading wildcard characters - as that will do effectively an index scan. If you can tokenize the text such that “yearly” is its own word/token that would be a more performant query.

3

Also, I see you’re doing a $sort after $search. I recommend, instead, that you map displayOrder to a token field type (or numeric if that is a better fit) and use the sort parameter of $search. Operations after $search consume ALL of the results in order to do sorting, whereas using the $search.sort option will be vastly more performant.