Skip to content

The SAP HANA Backend — Peculiarities

Hana supports SAP HANA 2.0 and SAP HANA Cloud. This document collects what deviates from the other backends and what the model must take into account.

The backend is selected by the JDBC URL of SAP's driver, jdbc:sap://host:port/... (com.sap.db.jdbc.Driver, which the application has to provide).

Minimum version. Hana.validateVersion() requires SAP HANA 2.0. SAP HANA Cloud reports the major version 4 and passes as well.

Type Mapping

SqlType HANA type Note
VARCHAR(n) NVARCHAR(n) n is capped at 5000 (getMaxSize); VARCHAR holds ASCII only on HANA 2.0
VARCHAR NVARCHAR(5000) an attribute without a size
CHAR NCHAR(1) for the same reason
CLOB NCLOB
XML NCLOB no native XML type; isSqlxmlSupported() is false
JSON NCLOB no native JSON column type; isJsonSupported() is false
BLOB BLOB
BIT BOOLEAN defaults are written as TRUE/FALSE, not as strings
TINYINT SMALLINT HANA's TINYINT is unsigned (0–255) and cannot hold a negative Java byte
FLOAT REAL DOUBLE stays DOUBLE
DECIMAL DECIMAL without a size, a decimal floating-point number that keeps the scale of each value
UUID — isUUIDSupported() is false, so a UUID is stored as a string

The generated CREATE TABLE does not name a table type (COLUMN or ROW), so the database's configured default applies. HANA Cloud knows column tables only.

Migration

HANA alters a column with ALTER TABLE ... ALTER (<column declaration>), and the new declaration replaces the old one. There is no statement that changes only the default or only the null constraint, which shapes the migration:

  • sqlAlterColumnType() always emits the whole declaration, including the default and an explicit NULL or NOT NULL: ALTER TABLE foo ALTER (name NVARCHAR(64) DEFAULT 'x' NOT NULL).
  • sqlAlterColumnNullConstraint() emits the same statement, and sqlAlterColumnDefault() returns null.
  • getMigrationStrategy() therefore never returns DEFAULT. A changed null constraint yields NULL — preceded by the UPDATE that replaces existing NULLs if the column becomes NOT NULL — which carries the type and the default along. Otherwise, a changed type, size, scale or default yields TYPE.
  • Renaming a column and changing its comment remain separate steps: RENAME COLUMN foo.name TO label and COMMENT ON COLUMN.

As on MySQL, dropping the DEFAULT from a generated ALTER statement when hand-editing a migration does not leave the default alone — it removes it.

Tables and indexes are renamed with RENAME TABLE and RENAME INDEX. Columns are added and dropped with ALTER TABLE ... ADD (...) and ALTER TABLE ... DROP (...).

Other Peculiarities

  • Paging needs a limit. SELECT ... LIMIT ? OFFSET ? FOR UPDATE — the locking clause comes last, and since there is no OFFSET without a LIMIT, a query with only an offset is given a limit of Integer.MAX_VALUE.
  • The default comes before the null constraint in a column declaration: DEFAULT 0 NOT NULL.
  • No filtered and no function-based indexes (isFilteredIndexSupported(), isFunctionBasedIndexSupported()). A model index using either is commented out if it is optional, and rejected otherwise.
  • No deferrable constraints. Deferrable unique indexes and foreign keys are created as plain ones.
  • Comments are supported on tables and columns (COMMENT ON), but not on indexes and sequences.
  • Sequences are read with SELECT seq.NEXTVAL FROM DUMMY, and the keep-alive select is SELECT 1 FROM DUMMY.
  • 127-character identifiers (getMaxNameLength()), and reserved words beyond SQL:2003 such as LIMIT, TOP, MINUS, ROWID, NEXTVAL and CURRVAL are rejected at build time (RESERVED_WORDS_HANA). The schemas SYS and PUBLIC are reserved.
  • Vendor error codes. Besides the standard SQLStates, the backend checks HANA's error codes: 131 (lock wait timeout), 133 (deadlock) and 146 (resource busy with NOWAIT) are transient, 301 (unique constraint violated) is a constraint violation, and -10709 and -10807 (connection failed, connection down) mean the communication link is lost.
  • No IF EXISTS on generated drops.