Mariadb Syntax SQL Ddl

MariaDB SQL DDL

Definition statements that change the schema : CREATE, ALTER, DROP, RENAME, plus sequences, generated columns, invisible columns, and partitioning. Read this before any schema migration on a production-size table.

Quick Reference

• ALWAYS specify ALGORITHM= and LOCK= on every ALTER TABLE so the server fails fast instead of silently falling back to COPY.
• ALWAYS prefer ALGORITHM=INSTANT for ADD COLUMN (10.3+) and DROP COLUMN (10.4+) when the table allows it. Metadata-only change, no rewrite.
• ALWAYS use ALGORITHM=INPLACE, LOCK=NONE for adding non-INSTANT-able indexes on big tables.
• NEVER default to ALGORITHM=COPY on production-size tables : rebuilds the whole table, blocks DML in earlier versions. From 11.2+ COPY supports LOCK=NONE but still rewrites everything.
• NEVER add an INSTANT column to a ROW_FORMAT=COMPRESSED table or to a table with a FULLTEXT index : both reject INSTANT.
• ALWAYS choose VIRTUAL for generated columns unless you need an FK pointing at the column, replication of the computed value, or skip the recompute cost on heavy reads. PERSISTENT/STORED is synonyms but PERSISTENT is the canonical MariaDB keyword.
• ALWAYS use atomic RENAME TABLE old TO retired, new TO old; for blue-green deployments. Multi-table RENAME is atomic within a single statement (10.6+ atomic DDL).
• NEVER use ALTER VIEW and expect rebuild : views require CREATE OR REPLACE VIEW.
• Sequences are MariaDB-only (10.3+), Oracle/PostgreSQL-compatible syntax. MySQL has no sequences.
• Partitioned tables CANNOT have foreign keys and CANNOT be referenced by foreign keys : hard architectural boundary.
• Atomic DDL is rollback-safe on server crash (10.6+) for single-table DDL. Multi-table DROP is crash-safe but not fully atomic.

Decision Trees

Picking an ALTER algorithm

Is the change metadata-only (ADD COLUMN at any position 10.4+, DROP COLUMN 10.4+, RENAME COLUMN/INDEX 10.5.3+, modify ENUM by appending, set DEFAULT) ?
   YES : ALGORITHM=INSTANT, LOCK=NONE
   NO  : continue

Is the change an ADD INDEX, DROP INDEX, or column modification that does not rewrite clustered-index keys ?
   YES : ALGORITHM=INPLACE, LOCK=NONE
   NO  : continue

Does the change require rebuilding the clustered index (change PK, change column type, change ROW_FORMAT) ?
   YES on small table  : ALGORITHM=COPY, LOCK=SHARED, run during maintenance window
   YES on big table    : use pt-online-schema-change or gh-ost ; never inline COPY on prod

Generated column : VIRTUAL or PERSISTENT

Do you need an FK pointing at the generated column ?
   YES : PERSISTENT (FK only allowed on stored)
   NO  : continue

Do you replicate row-based and need the computed value applied identically on the replica ?
   YES : PERSISTENT
   NO  : continue

Is the read path heavy and the expression expensive ?
   YES : PERSISTENT (trade disk for CPU)
   NO  : VIRTUAL (default ; no disk cost)

Partitioning : type selection

Filter is a date / numeric range with predictable boundaries ?
   YES : RANGE or RANGE COLUMNS (multi-column variant)

Filter is a finite set of discrete values (region codes, tenant ids) ?
   YES : LIST or LIST COLUMNS

You want even distribution by hashing a key, no semantic boundary ?
   YES : HASH (LINEAR HASH for less-expensive partition addition)

You want hash-partitioning by primary key columns ?
   YES : KEY (or LINEAR KEY)

You need automatic time-period partitions tied to system-versioning ?
   YES : SYSTEM_TIME (MariaDB-only ; requires WITH SYSTEM VERSIONING)

Patterns

CREATE TABLE (modern default)

-- 10.6+ : production-grade table template
CREATE TABLE invoice (
  id           BIGINT UNSIGNED AUTO_INCREMENT PRIMARY KEY,
  customer_id  BIGINT UNSIGNED NOT NULL,
  amount       DECIMAL(12,2)   NOT NULL,
  status       ENUM('draft','sent','paid','void') NOT NULL DEFAULT 'draft',
  metadata     JSON            NULL CHECK (JSON_VALID(metadata)),
  created_at   TIMESTAMP(6)    NOT NULL DEFAULT CURRENT_TIMESTAMP(6),
  updated_at   TIMESTAMP(6)    NOT NULL DEFAULT CURRENT_TIMESTAMP(6) ON UPDATE CURRENT_TIMESTAMP(6),
  KEY ix_customer_status (customer_id, status),
  CONSTRAINT fk_invoice_customer FOREIGN KEY (customer_id) REFERENCES customer(id) ON DELETE RESTRICT
) ENGINE=InnoDB
  ROW_FORMAT=DYNAMIC
  DEFAULT CHARSET=utf8mb4
  COLLATE=utf8mb4_uca1400_ai_ci;

Always set CHARSET explicitly per table : 10.6 and 10.11 LTS stock builds still default to latin1, which silently breaks emoji and accented strings.

Fast schema change : INSTANT ADD COLUMN

-- 10.4+ : metadata-only column add, no rewrite, no DML block
ALTER TABLE invoice
  ADD COLUMN paid_at TIMESTAMP(6) NULL AFTER created_at,
  ALGORITHM=INSTANT,
  LOCK=NONE;

The server errors out (HY000 1845) if INSTANT is impossible (compressed row format, FULLTEXT index present). No silent COPY fallback.

Fast index add : INPLACE

-- 10.6+ : online index build, DML allowed during build
ALTER TABLE invoice
  ADD KEY ix_status_created (status, created_at),
  ALGORITHM=INPLACE,
  LOCK=NONE;

ADD INDEX is INPLACE-only ; it cannot be INSTANT because the index data structure must be populated.

Multi-operation ALTER in one statement

-- 10.4+ : combine compatible operations to reduce one-touch overhead
ALTER TABLE invoice
  ADD COLUMN tax_amount DECIMAL(12,2) NOT NULL DEFAULT 0 AFTER amount,
  ADD COLUMN currency   CHAR(3)       NOT NULL DEFAULT 'EUR' AFTER tax_amount,
  ALGORITHM=INSTANT,
  LOCK=NONE;

All columns added in one metadata change. If any single sub-change rejects INSTANT, the whole statement errors out and nothing changes.

Atomic table swap for blue-green deploys

-- 10.6+ : single-statement atomic swap, no client sees the wrong shape
RENAME TABLE
  invoice           TO invoice_v1,
  invoice_v2        TO invoice;
-- After verification :
DROP TABLE invoice_v1;

Within a single RENAME TABLE statement, all renames either succeed atomically or roll back together. Atomic for InnoDB, MyRocks, MyISAM, Aria.

Sequence as MySQL-incompatible ID generator

-- 10.3+ MariaDB-only
CREATE SEQUENCE invoice_seq
  START WITH 1000
  INCREMENT BY 1
  MINVALUE   1000
  MAXVALUE   9999999999
  CACHE      1000
  NOCYCLE;

INSERT INTO invoice (id, customer_id, amount) VALUES
  (NEXT VALUE FOR invoice_seq, 42, 99.95);

-- Inspect the last value generated in this session
SELECT PREVIOUS VALUE FOR invoice_seq;

NEXT VALUE FOR seq is SQL-standard ; NEXTVAL(seq) is the legacy function form, both work. CACHE 1000 pre-allocates 1000 ids per request : restart loses unused cached ids, which is a gap-creator. CACHE 1 maximises consistency at the cost of per-row overhead.

Generated column with functional index (JSON path)

-- 10.2+ : VIRTUAL column over JSON path, indexed
ALTER TABLE event
  ADD COLUMN customer_id BIGINT UNSIGNED
    AS (JSON_VALUE(payload, '$.customer.id')) VIRTUAL,
  ADD KEY ix_event_customer (customer_id),
  ALGORITHM=COPY;

VIRTUAL re-evaluates on read but the index is materialised. ADD INDEX on a VIRTUAL column requires ALGORITHM=COPY (KB-documented limitation).

Invisible column for soft-deprecation

-- 10.3+ : hide a legacy column from SELECT * without breaking explicit queries
ALTER TABLE customer
  MODIFY legacy_phone VARCHAR(20) INVISIBLE,
  ALGORITHM=INSTANT,
  LOCK=NONE;

SELECT * skips the column. SELECT legacy_phone still works. Marker for "remove in next major release" without immediate ALTER + app-deploy coordination.

Range partition by date

-- 10.6+ : RANGE COLUMNS supports multi-column and date types
CREATE TABLE event_log (
  id          BIGINT UNSIGNED AUTO_INCREMENT,
  occurred_at DATETIME(6)     NOT NULL,
  payload     JSON            NULL CHECK (JSON_VALID(payload)),
  PRIMARY KEY (id, occurred_at)
) ENGINE=InnoDB
  PARTITION BY RANGE COLUMNS (occurred_at) (
    PARTITION p2024 VALUES LESS THAN ('2025-01-01'),
    PARTITION p2025 VALUES LESS THAN ('2026-01-01'),
    PARTITION p2026 VALUES LESS THAN ('2027-01-01'),
    PARTITION pmax  VALUES LESS THAN (MAXVALUE)
  );

The PK must include the partitioning column (occurred_at) : a hard MariaDB rule. Foreign keys are forbidden on partitioned tables.

Drop table safely

-- IF EXISTS suppresses the no-such-table error ; useful in idempotent migrations
DROP TABLE IF EXISTS staging_import_2024;

-- Multi-table drop : crash-safe but not fully atomic in 10.6+
DROP TABLE IF EXISTS tmp_a, tmp_b, tmp_c;

If any table is referenced by a FK from another table, the DROP errors out (Cannot delete or update a parent row : a foreign key constraint fails). Drop child tables first or SET FOREIGN_KEY_CHECKS=0 during controlled maintenance only.

Atomic DDL (10.6+)

Single-statement DDL is rollback-safe across server crash : CREATE TABLE (except CREATE OR REPLACE), ALTER TABLE, single-table DROP TABLE, RENAME TABLE, CREATE VIEW, CREATE SEQUENCE, CREATE TRIGGER, DROP TRIGGER. The ddl_recovery.log tracks in-progress operations ; on restart MariaDB completes or reverses each entry up to 3 retries.

NOT fully atomic (crash-safe but partial-completion possible) : multi-table DROP TABLE, CREATE OR REPLACE TABLE, DROP DATABASE.

Storage-engine support : InnoDB, MyRocks, MyISAM, Aria. NOT supported : S3 engine, partitioning engine (some operations).

Cross-References

• mariadb-syntax-sql-dml for INSERT/UPDATE/DELETE patterns including RETURNING and ON DUPLICATE KEY UPDATE.
• mariadb-syntax-indexing for index-type details, leftmost-prefix, descending indexes, ignored indexes.
• mariadb-syntax-json for JSON validation, JSON_VALUE/JSON_QUERY, functional indexes.
• mariadb-impl-schema-design for naming conventions, PK choice (BIGINT vs UUID), charset strategy.
• mariadb-impl-migration for online schema-change tools (pt-online-schema-change, gh-ost).
• mariadb-core-storage-engines for ROW_FORMAT trade-offs and engine selection.

Reference Files

• references/methods.md : complete CREATE/ALTER/DROP/RENAME grammar, sequence functions, partitioning expressions.
• references/examples.md : 12+ working DDL examples covering every category listed in this skill.
• references/anti-patterns.md : 9 anti-patterns from real production incidents with root-cause and fix.

Sources

Verified against : mariadb.com/kb/en/alter-table/, mariadb.com/kb/en/instant-add-column-for-innodb/, mariadb.com/kb/en/create-sequence/, mariadb.com/kb/en/generated-columns/, mariadb.com/kb/en/invisible-columns/, mariadb.com/kb/en/partitioning-types-overview/, mariadb.com/kb/en/partition-pruning-and-selection/, mariadb.com/kb/en/rename-table/, mariadb.com/kb/en/atomic-ddl/.