This works for me. :-) As for you point 1), if there is no tg_init_database, then the behaviour is exactly as before and SQLObject users will not be confused. That is to say, I have to do an explicit act to get the new behaviour and so I will know what will happen.
As for point 2), I am thinking that a counterpart method like tg_clear_database or something would be nice so that old data can be torn down. And alternative would be to make your tg_init_database, check for the existence of any data you are trying to insert. Further a reset or rebuild method is a simple as tg_clear_database(); tg_init_database(). If that is useful. And then there is the blue-sky-on-mars <wink> thinking that this could be the base for model versioning with upgrade functions in each model to that can migrate old data structures to new ones. I think this is a bit far off, though. (Can you tell I have been played with Zope and Twisted?) :-) Krys Fabian Neumann wrote: >Krys Wilken wrote: > > >>I think both ideas are of value. An exec command for running scripts, >>and a special module-level function in model.py that is run >>automatically by sql create. >> >> > >I might like it, too. What do you think of this: > >Index: turbogears/quickstart/projectname/model.py.source >=================================================================== >--- turbogears/quickstart/projectname/model.py.source (Revision 43) >+++ turbogears/quickstart/projectname/model.py.source (Arbeitskopie) >@@ -6,3 +6,6 @@ > > # class YourDataClass(SQLObject): > # pass >+ >+# def tg_init_database(): >+# pass >Index: turbogears/command/__init__.py >=================================================================== >--- turbogears/command/__init__.py (Revision 43) >+++ turbogears/command/__init__.py (Arbeitskopie) >@@ -104,6 +104,12 @@ > > from sqlobject.manager import command > command.the_runner.run(sys.argv) >+ if sqlobjcommand == "create": >+ mod = self.get_model() >+ if hasattr(mod, 'tg_init_database'): >+ print 'Filling database' >+ mod.tg_init_database() >+ > > def fix_egginfo(self, eggname): > print """ > > >Two issues with that: >1. This behaviour may be unexpected by people who are familiar with > "sqlobject-admin create". >2. If you run "sql create" twice or more, those init values are doubled. > > > >>An alternative to a single module-level function would be a special >>method on each model that is responsible for populating itself, perhaps >>"post_create(self)" or something. Sql create could just call that >>method of each model after all the tables have been created. >> >> > >I don't think this granularity is useful, as you can only create all >models at once (I think). > >Fabian > > > >
