[DDLUTILS-214] Primary Key Column order lost (a problem if there are multiple Primary Keys)

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.1
• Fix Version/s: 1.1
• Component/s: Core (No specific database)
• Environment: Any
• Assignee: Thomas Dudziak
• Reporter: Cordeo
• Created: 17/Jul/08 11:13
• Updated: 11/Aug/08 05:18
• Resolved: 11/Aug/08 05:18

Description

Symptom:
14558 [btpool0-5] DEBUG org.apache.ddlutils.platform.postgresql.PostgreSqlPlatform - About to execute SQL DELETE FROM "a" WHERE "p" = ? AND "q" = ?
The delete sometimes works, sometimes I get this:
14569 [btpool0-5] WARN org.apache.ddlutils.platform.postgresql.PostgreSqlPlatform - Attempted to delete a single row "a": q = 6, p = 3 in table "a" but changed 0 row(s).

Symptom analysis:
The query
DELETE FROM "a" WHERE "p" = ? AND "q" = ?
is filled in order (6,3), without regarding the field names (q = 6, p = 3) resulting in
DELETE FROM "a" WHERE "p" = 6 AND "q" = 3

Code analysis
1. SqlBuilder.getDeleteSql
SqlBuilder.getDeleteSql uses the Map pkValues iterator to generate the prepared statement.

2. PlatformImplBase.toColumknValues
This has a major problem: Map iterator order is not defined. Java documentation: "Some map implementations, like the TreeMap class, make specific guarantees as to their order; others, like the HashMap class, do not." In this case, the Map used is created by PlatformImplBase.toColumknValues:
protected HashMap toColumnValues(SqlDynaProperty[] properties, DynaBean bean)

{ HashMap result = new HashMap(); ... }

so iterator order is undefined.

3. PlatformImplBase.delete (for example)
a)
String sql = createDeleteSql(model, dynaClass, primaryKeys, null);
createDeleteSql uses getDeleteSql (1) and due to (2), the order the columns in the generated String sql is undefined. So, this could generate either
DELETE FROM "a" WHERE "p" = ? AND "q" = ?
DELETE FROM "a" WHERE "q" = ? AND "p" = ?
which one cannot be determined
b) setObject uses a deterministic array order sqlIndex to fill the ? - marks
SqlDynaProperty[] primaryKeys = dynaClass.getPrimaryKeyProperties();
for (int idx = 0; idx < primaryKeys.length; idx++)

{ setObject(statement, idx + 1, dynaBean, primaryKeys[idx]); }

so there is a chance of 1/(number of PK columns) this will run correctly.

Solution:
use LinkedHashMap or TreeMap: they retain put-order, which is based on the SqlDynaProperty[] so by definition has the same order as the setObject loop:
protected HashMap toColumnValues(SqlDynaProperty[] properties, DynaBean bean)

{ HashMap result = new LinkedHashMap(); ... }