I have some general thoughts about the IDs used with the GdaSqlBuilder functions:
1. I think we should have a GdaSqlBuilderID typedef. That will make things easier if this type is every more complex in some future API-breaking version of libgda. This would not break code that currently uses guint. 2. Some functions allow you to specify the value of the ID that will be returned, though you can leave it as 0 to make it auto-generated. For instance, gda_sql_builder_add_cond() http://library.gnome.org/devel/libgda/unstable/GdaSqlBuilder.html#gda-sql-builder-select-add-target However, 2.1. This adds an extra parameter, making the API more complicated, particularly when there are other parameters of exactly the same type, 2.2 This encourages application developers to use "magic numbers" in their code, making the code fragile and obscure. I don't like seeing parameters like "5, 1, 2" in code, but I don't mind "field_id, value_id". For this reason, I hide these parameters in libgdamm (The C++ binding). Application code then just saves the return values in temporary variables, whose names make the code clearer. -- [email protected] www.murrayc.com www.openismus.com _______________________________________________ gnome-db-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/gnome-db-list
