gourmet team mailing list archive

[Bernhard Reiter <ockham@xxxxxxxxx>]

Tom,

Am 2013-06-28 15:07, schrieb Thomas Mills Hinkle:
> Bernhard et al,
I'm the one to blame...
> I've been a bit busy and haven't followed all the development leading 
> up to the 0.16 release, but it looks like 0.16.0 went out with changes 
> to the database model but *without* code to update existing databases.
Actually, I tried to make sure existing databases *are* updated, see 
https://github.com/thinkle/gourmet/blob/master/gourmet/backends/db.py#L510 
through line 554. I thought I tested this stuff -- does it break your DB?
<https://github.com/thinkle/gourmet/blob/master/gourmet/backends/db.py#L510> 

> Specifically, it seems like primary keys have been changed in the 
> pantry and other columns.
>
> This is pretty bad -- I've always tried to ensure that existing users 
> get treated well. I'm wondering, have bug reports been flowing in 
> about this? Is 0.16 not out generally?
Yes it is, and the Ubuntu package has even been part of 13.04 already. 
(Other distributions lag behind a bit, though, and I still haven't 
managed to roll out a Windows release.) I really tried to make sure not 
to break anything -- please describe what specifically is broken (any 
errors, or messed up DB layouts?) Anyway, there have been no bug reports 
yet that I was able to attribute to this change. Actually, not that many 
new bug reports after 0.16 at all...
> For the future, changes to the DB structure are a pain in the ass and 
> shouldn't be made for simple "code cleanup" reasons -- I doubt there 
> is any functional benefit to introducing new primary key columns named 
> 'id' in place of or in addition to the old ones?
Yes there is -- it wasn't possible to connect gourmet to a MySQL 
database before that change, see 
https://github.com/thinkle/gourmet/issues/681 and the blog article 
linked therein that made me aware of the issue, 
http://blog.henrygis.com/2011/02/setting-up-gourmet-recipe-manager-part.html

In a nutshell, MySQL requires length parameters to VARCHARs if they're 
supposed to serve as primary keys. So I had the choice between adding 
such parameters (with some arbitrarily chosen value for their maximum 
lengths, thus possibly breaking longer values in the affected columns), 
or introduce new auto-incremented integer primary keys -- I went for the 
latter. (This is also described in the comment at 
https://github.com/thinkle/gourmet/blob/master/gourmet/backends/db.py#L535 )
<https://github.com/thinkle/gourmet/blob/master/gourmet/backends/db.py#L535> 

> I would tend to be inclined to just roll back the changes (that's what 
> I'll do locally),
Please don't (at least not in the public git repo) -- as described 
above, this wasn't just for cleanup's sake. If it's really broken, we'd 
better provide an actual fix with 0.16.1.
> but since the code is now in the wild and new users will get the new 
> columns, I think we have to write the update code anyway.
Ironically, the one bug report I know of related to this is 
https://github.com/thinkle/gourmet/issues/711 -- seems like it still 
doesn't work with MySQL, but I haven't had the time yet to look into it.
<https://github.com/thinkle/gourmet/issues/711>
> Tom
Bernhard