SQL/MX 2.x Database and Application Migration Guide (G06.23+, H06.04+, J06.03+)

Migrating SQL/MX Release 1.8 Applications to

SQL/MX Release 2.x

HP NonStop SQL/MX Database and Application Migration Guide—540435-005

7-4

Stable Access

changing the source code (that is, embedding a CONTROL QUERY DEFAULT

statement). For more information, see the

SQL/MX Programming Manual for C and

COBOL.

For dynamically compiled applications, set UPD_ABORT_ON_ERROR in a CONTROL

QUERY DEFAULT statement in the source code and recompile the application.

For more information about UPD_ABORT_ON_ERROR, see the

SQL/MX Reference

Manual.

Stable Access

Stable access provides cursor stability by preventing a row at the current cursor

position from being modified by another program yet allows concurrent access to other

rows in the table.

SQL/MX Release 1.8 Stable Access

In SQL/MX Release 1.8, STABLE access is equivalent to READ COMMITTED access,

which requires that data accessed by a DML statement be from committed rows.

SQL/MX Release 2.x Stable Access

In SQL/MX Release 2.x, STABLE access is equivalent to the SQL/MP form of STABLE

access, which provides cursor stability.

Effect of the New Behavior on SQL/MX Applications

Applications that were statically compiled in SQL/MX Release 1.8 will continue using

STABLE to mean READ COMMITTED in SQL/MX Release 2.x. However, SQL/MX

applications that are compiled dynamically at run time, statically recompiled, or

automatically recompiled in SQL/MX Release 2.x will get true STABLE access

behavior and cursor stability.

Retaining SQL/MX Release 1.8 Stable Access

To retain READ COMMITTED behavior for dynamic applications and recompiled

applications in SQL/MX Release 2.x, change STABLE to READ COMMITTED in the

source code and recompile the application.

For more information about access options, see the

SQL/MX Reference Manual.

IEEE Floating-Point Host Variables

SQL/MX Release 2.x introduces IEEE floating-point format. The storage and precision

of IEEE floating-point data types are different from those of Tandem floating-point data

types.