The Query Wizard is in need of a serious redesign. Its implementation is a confusing mess, and it gets itself confused over far too many things:

 - There is a Java SQLQueryComposer class, which confusingly is *not*
   an implementation of the XSQLQueryComposer interface (nor the
   corresponding service). The latter uses an XSingleSelectQueryComposer,
   but in very partial ways, and sometimes resetting it, blurring what
   is the responsibility of the Java SQLQueryComposer and what of the
   XSingleSelectQueryComposer.

 - Confused columns with same name but in different tables

 - Confused about aliasing between result columns and table columns
   (shows e.g. in the ORDER BY clause)

 - Confused about the table column that is being aggregated
   and the *result* of the aggregation. Actually, does not
   have a clear notion of result column of aggregation.

   For example, aggregate both the SUM and the AVG of column "cl"
   (in that order).
   The "aliases" step proposes the "cl" column only once, and
   the generated query has only the sum, with the alias chosen
   for column "cl". It should generate *both* the SUM and AVG
   and the alias step should show SUM("cl") *and* AVG("cl"),
   but *not* "cl" itself.

 - Does not include the GROUP BY columns in the result, giving a
   generally not-that-useful result.

   E.g. imagine we want the SUM of the amounts of the invoices
   per customer (that is, the turnover per customer).
   The wizard makes a query that gives the turnover, but does not
   say for which customer each number is.

 - Gives only utterly non-useful choices for HAVING

   It proposes only the (unadulterated) GROUP BY columns. Selecting
   those in the WHERE clause gives *exactly* the same result.

   It should propose the RESULTS of the aggregations.

 - mishandles dates (see bug 59978)


The wizard itself is in wizards/com/sun/star/wizards/query, but a lot of the work is done in wizards/com/sun/star/wizards/db and wizards/com/sun/star/wizards/ui.