OpenSearch:Configure an index table

Schema structure

The top-level structure of an index table schema:

{
    "table_name": "sample",       // Index table name — also used in the generated schema filename
    "fields": [],                 // Field definitions (field name, type, analyzer)
    "indexs": [],                 // Inverted index configurations
    "attributes": [],             // Forward index (attribute) configurations
    "summarys": {},               // Summary index configuration
    "dictionaries": [],           // Dictionaries for bitmap indexes
    "adaptive_dictionaries": [],  // Rules for generating adaptive bitmap indexes
    "enable_ttl": true,           // Whether to enable TTL; default: false
    "ttl_field_name": "ttl_field", // Field that holds per-document TTL values
    "default_ttl": 86400          // Default TTL in seconds
}

fields configuration

The fields array declares every field available for indexing, attribute storage, or summary retrieval.

"fields": [
    {
        "field_name": "title",
        "field_type": "TEXT",
        "analyzer": "chn_standard"           // Required for TEXT fields; not allowed for other types
    },
    {
        "field_name": "dup_title",
        "field_type": "TEXT",
        "analyzer": "fuzzy",
        "user_defined_param": {
            "copy_from": "title"             // Copies data from the "title" field
        }
    },
    {
        "field_name": "category",
        "field_type": "INTEGER",
        "multi_value": true,
        "compress_type": "uniq|equal"        // Remove duplicates + compress offset data
    },
    {
        "field_name": "mlr_features",
        "field_type": "INTEGER",
        "multi_value": true,
        "updatable_multi_value": true        // Allows in-place updates to multi-value data
    },
    {
        "field_name": "feature",
        "field_type": "float",
        "multi_value": true,
        "fixed_multi_value_count": 32,       // Fixed-length multi-value: exactly 32 values per document
        "compress_type": "uniq|fp16",        // fp16: ~50% compression, accepts slight precision loss
        "updatable_multi_value": true
    },
    {
        "field_name": "price",
        "field_type": "INTEGER",
        "enable_null": true,
        "default_null_string": "default_null" // String representing a null value in the source document
    },
    {
        "field_name": "timestamp",
        "field_type": "TIMESTAMP",
        "default_time_zone": "+0800"          // UTC+8; applied when source document omits timezone
    }
]

indexs configuration

The indexs array holds a list of inverted index configurations. Each element is a complete index definition:

"indexs": [
    { /* index 1 configuration */ },
    { /* index 2 configuration */ }
]

For supported index types and configuration details, see Overview of inverted indexes.

attributes configuration

The attributes array lists the field names to store as forward indexes. All field types except TEXT can be used as attributes. Fields must be declared in fields.

"attributes": [
    "user_id",
    "product_id",
    "category"
]

summarys configuration

The summarys object specifies which fields to store verbatim for retrieval at query time. All field types can be included. Fields must be declared in fields.

"summarys": {
    "summary_fields": ["id", "company_id", "subject", "cat_id"],
    "compress": false
}

dictionaries configuration

The dictionaries array defines static high-frequency term dictionaries for bitmap indexes. Bitmap indexes reduce index space and improve retrieval performance for high-frequency terms. Omit this section if no bitmap index is needed.

"dictionaries": [
    {
        "dictionary_name": "bitmap1",
        "content": "a;an"         // Terms separated by semicolons
    },
    {
        "dictionary_name": "bitmap2",
        "content": "of;and"
    }
]

adaptive_dictionaries configuration

The adaptive_dictionaries array defines rules for automatically generating high-frequency term bitmap indexes based on document statistics. Use this when you want the system to determine which terms qualify as high-frequency rather than specifying them manually.

"adaptive_dictionaries": [
    {
        "adaptive_dictionary_name": "df",
        "dict_type": "DOC_FREQUENCY",
        "threshold": 1500000       // Terms with df >= 1,500,000 become high-frequency
    },
    {
        "adaptive_dictionary_name": "percent",
        "dict_type": "PERCENT",
        "threshold": 30            // Terms where (df / totalDocCount) * 100 >= 30 become high-frequency
    },
    {
        "adaptive_dictionary_name": "size",
        "dict_type": "INDEX_SIZE"  // Compares bitmap index size vs original; uses smaller option
    }
]

Related topics

• Overview of inverted indexes

• Built-in field types in OpenSearch Retrieval Engine Edition

• Analyzers