多索引搜索（Autocomplete）咨询：跨多索引实现自动补全功能

Got it, let's tackle this problem step by step since Elasticsearch 6.x's single-type-per-index constraint is throwing a wrench in your original plan. Here's how you can replicate that single-index autocomplete behavior across your four separate indices:

1. Ensure Each Index Has a Consistent Autocomplete Field #

First, you need to add a completion-type field to every one of your indices (cities, regions, countries, hotels). This is the field Elasticsearch uses for fast, optimized autocomplete lookups.

Example Mapping Update (Repeat for All Indices) #

For each index, run a mapping update to add a name_suggest field (keep the name consistent across all indices to simplify queries):

PUT /cities/_mapping
{
  "properties": {
    "name": {
      "type": "text"
    },
    "name_suggest": {
      "type": "completion",
      "analyzer": "simple",
      "preserve_separators": true,
      "preserve_position_increments": true,
      "max_input_length": 50
    }
  }
}

• The simple analyzer ensures basic, consistent tokenization (lowercasing, stripping special characters) across all indices.
• Adjust max_input_length or other completion settings based on your data's typical name length.

2. Sync Data to Populate the Autocomplete Field #

Since you're using Logstash JDBC or the Reindex API, you need to make sure every document has the name_suggest field populated with the value from the name field.

Logstash Filter Example #

Add this to your Logstash config's filter section to copy the name value to name_suggest:

filter {
  mutate {
    copy => { "name" => "name_suggest" }
  }
}

Reindex API Example #

If using Reindex to migrate data into these indices, include a script to populate the autocomplete field:

POST _reindex
{
  "source": {
    "index": "your_source_index"
  },
  "dest": {
    "index": "cities"
  },
  "script": {
    "source": "ctx._source.name_suggest = ctx._source.name"
  }
}

3. Run Cross-Index Autocomplete Searches #

Now you can query all four indices at once to get unified autocomplete results. The syntax is almost identical to your single-index search—you just specify multiple indices separated by commas.

Example Autocomplete Query #

POST /cities,regions,countries,hotels/_search
{
  "suggest": {
    "global_name_suggest": {
      "prefix": "par", // Replace with user's input
      "completion": {
        "field": "name_suggest",
        "size": 10, // Number of results to return
        "skip_duplicates": true, // Remove duplicate names across indices
        "fuzzy": {
          "fuzziness": 1 // Optional: Allow minor typos for better user experience
        }
      }
    }
  }
}

• skip_duplicates is useful if, for example, "Paris" exists in both cities and regions—it will only return one instance.
• The fuzzy parameter adds resilience to user typos (adjust fuzziness based on how lenient you want to be).

4. Optional: Use Index Templates for Consistency #

To avoid repeating the mapping setup for each index, create an index template that applies the name_suggest field to any new indices matching your pattern:

PUT _template/location_autocomplete_template
{
  "index_patterns": ["cities", "regions", "countries", "hotels"],
  "mappings": {
    "properties": {
      "name": {
        "type": "text"
      },
      "name_suggest": {
        "type": "completion",
        "analyzer": "simple"
      }
    }
  }
}

This ensures any future updates or new indices matching the pattern will automatically have the correct autocomplete field configured.

内容的提问来源于stack exchange，提问作者Alireza Bashiri