I'm quite new to ElasticSearch/OpenSearch, so if I overlook something basic, please forgive me.

I'm making a service which performs full-text search on novels. novels index has the following fields:

• title keyword. 10-40 characters long in average.
• conetnt text type. in most documents, far longer than 100,000.

As an experiment, I change the type of content type into keyword, bacause I wanted to know the difference between text type and keyword type. I was not sure if documents which contains super long keyword field would be indexed, but no problem occurred.

And tried the following wildcard query on it.

{
  "query": {
    "wildcard": {
      "content": "*あらゆる透明な幽霊の複合体*"
    }
  }
}

However, it didn't hit the document which contains the above phrase. After that, I tried wildcard query on title field, and it worked nicely, and then I tried match_phrase query, like the following, and it hit the proper document as well:

{
  "query": {
    "match_phrase": {
      "content": "あらゆる透明な幽霊の複合体"
    }
  }
}

So, the only reason I could come up with why wildcard query didn't work on content field is, simply, it's super long.

Now, I have two questions.

1. Is my hypothesis--wildcard query doesn't work on super long texts--correct?
2. If my hypothesis is true, why wildcard query doesn't work on super long texts? What's the mechanism under the hood?

I would deeply appreciate it if you would shed some lights on it.

Incidentally, I'm using opensearch1.2.4 on docker.

CodePudding user response：

The reason might be that your keyword field has an ignore_above parameter set to some value smaller than your content length, so any string of text longer than that will not be indexed.

It's usually not a good idea to store long strings of text as keyword. You should prefer either match_only_text or the wildcard field type depending on your exact use case.

I think, however, that those two field types are not available on Opensearch, so your only option is to play with the ignore_above parameter or simply use the text field type and use a match_phrase query.