Skip to main content
Version: Nightly

STATISTICS

The STATISTICS table provides index metadata in a MySQL-compatible shape. It powers SHOW INDEX and returns one row per indexed column for primary keys, time indexes, inverted indexes, fulltext indexes, and skipping indexes.

USE INFORMATION_SCHEMA;
DESC STATISTICS;

The output is as follows:

+---------------------+--------+------+------+---------+---------------+
| Column              | Type   | Key  | Null | Default | Semantic Type |
+---------------------+--------+------+------+---------+---------------+
| table_catalog       | String |      | NO   |         | FIELD         |
| table_schema        | String |      | NO   |         | FIELD         |
| table_name          | String |      | NO   |         | FIELD         |
| non_unique          | Int64  |      | NO   |         | FIELD         |
| index_schema        | String |      | NO   |         | FIELD         |
| index_name          | String |      | NO   |         | FIELD         |
| seq_in_index        | Int64  |      | NO   |         | FIELD         |
| column_name         | String |      | NO   |         | FIELD         |
| collation           | String |      | YES  |         | FIELD         |
| cardinality         | Int64  |      | YES  |         | FIELD         |
| sub_part            | Int64  |      | YES  |         | FIELD         |
| packed              | String |      | YES  |         | FIELD         |
| nullable            | String |      | NO   |         | FIELD         |
| index_type          | String |      | NO   |         | FIELD         |
| comment             | String |      | NO   |         | FIELD         |
| index_comment       | String |      | NO   |         | FIELD         |
| is_visible          | String |      | NO   |         | FIELD         |
| expression          | String |      | YES  |         | FIELD         |
| greptime_index_type | String |      | YES  |         | FIELD         |
+---------------------+--------+------+------+---------+---------------+

Create tables with different index types and query the metadata:

CREATE TABLE IF NOT EXISTS system_metrics (
    host STRING,
    idc STRING FULLTEXT INDEX INVERTED INDEX,
    cpu_util DOUBLE,
    memory_util DOUBLE,
    disk_util DOUBLE,
    desc1 STRING,
    desc2 STRING FULLTEXT INDEX,
    desc3 STRING FULLTEXT INDEX,
    ts TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    PRIMARY KEY(host, idc),
    TIME INDEX(ts)
);

CREATE TABLE IF NOT EXISTS test (
    a STRING,
    b STRING SKIPPING INDEX,
    c DOUBLE,
    ts TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    PRIMARY KEY(a, b),
    TIME INDEX(ts)
);

SELECT table_schema,
       table_name,
       non_unique,
       index_name,
       seq_in_index,
       column_name,
       index_type,
       greptime_index_type
FROM information_schema.statistics
WHERE table_schema = 'public'
  AND table_name IN ('system_metrics', 'test')
ORDER BY table_name, index_name, seq_in_index, column_name;

The output is as follows:

+--------------+----------------+------------+----------------------+--------------+-------------+------------+---------------------+
| table_schema | table_name     | non_unique | index_name           | seq_in_index | column_name | index_type | greptime_index_type |
+--------------+----------------+------------+----------------------+--------------+-------------+------------+---------------------+
| public       | system_metrics | 1          | FULLTEXT_INDEX_desc2 | 1            | desc2       | FULLTEXT   | fulltext_bloom      |
| public       | system_metrics | 1          | FULLTEXT_INDEX_desc3 | 1            | desc3       | FULLTEXT   | fulltext_bloom      |
| public       | system_metrics | 1          | FULLTEXT_INDEX_idc   | 1            | idc         | FULLTEXT   | fulltext_bloom      |
| public       | system_metrics | 1          | INVERTED_INDEX_idc   | 1            | idc         | INVERTED   | inverted            |
| public       | system_metrics | 0          | PRIMARY              | 1            | host        | PRIMARY    | dense               |
| public       | system_metrics | 0          | PRIMARY              | 2            | idc         | PRIMARY    | dense               |
| public       | system_metrics | 1          | TIME INDEX           | 1            | ts          | TIME       |                     |
| public       | test           | 0          | PRIMARY              | 1            | a           | PRIMARY    | dense               |
| public       | test           | 0          | PRIMARY              | 2            | b           | PRIMARY    | dense               |
| public       | test           | 1          | SKIPPING_INDEX_b     | 1            | b           | SKIPPING   | bloom_filter        |
| public       | test           | 1          | TIME INDEX           | 1            | ts          | TIME       |                     |
+--------------+----------------+------------+----------------------+--------------+-------------+------------+---------------------+

The following statements are equivalent:

SELECT table_name, non_unique, index_name, seq_in_index, column_name
FROM information_schema.statistics
WHERE table_schema = '<db_name>' AND table_name = '<table_name>';

SHOW INDEX FROM <table_name> IN <db_name>;

The description of columns in the STATISTICS table is as follows:

  • table_catalog: The catalog of the table. The value is always def.
  • table_schema: The database of the table.
  • table_name: The name of the table.
  • non_unique: Whether the index can contain duplicate values. 0 means unique, 1 means non-unique.
  • index_schema: The database that contains the index metadata. It is the same as table_schema.
  • index_name: The index name. PRIMARY and TIME INDEX keep their names, while each secondary index is named <TYPE>_INDEX_<column>, such as INVERTED_INDEX_idc, FULLTEXT_INDEX_desc2, or SKIPPING_INDEX_b.
  • seq_in_index: The position of the column inside the index, starting from 1. Secondary indexes are single-column and always 1; only a multi-column primary key produces 1, 2, and so on.
  • column_name: The indexed column name.
  • collation: The column sort direction in the index. The value is currently always A.
  • cardinality: The estimated number of unique values in the index. Currently NULL.
  • sub_part: The indexed prefix length. Currently NULL.
  • packed: How the key is packed. Currently NULL.
  • nullable: Whether the indexed column can be NULL. The value is YES or an empty string.
  • index_type: The index category, one of PRIMARY, TIME, INVERTED, FULLTEXT, or SKIPPING.
  • comment: Additional index metadata. Currently an empty string.
  • index_comment: User-facing index comments. Currently an empty string.
  • is_visible: Whether the index is visible. The value is currently always YES.
  • expression: The indexed expression for functional indexes. Currently NULL.
  • greptime_index_type: The GreptimeDB internal index type identifier: dense or sparse for primary keys, inverted, fulltext_bloom or fulltext_tantivy, and bloom_filter for skipping indexes. It is empty for the time index.