Re: [SQLObject] [ 1489030 ] Fix relative paths for connections

[Ian B. <ia...@co...>]

Matthew Good wrote:
> On Thu, 2006-05-18 at 18:58 +0100, Andy Kilner wrote:
> 
>>On Thu, 2006-05-18 at 10:47 -0500, Ian Bicking wrote:
>>
>>>Oleg Broytmann wrote:
>>>
>>>>https://sourceforge.net/tracker/index.php?func=detail&aid=1489030&group_id=74338&atid=540674
>>>>
>>>>   I'd like to discuss the patch. I am personally against relative database
>>>>URIs. What if you do os.chdir() in the middle of your program?
>>>>   What do people think about it?
>>>
>>>I don't like relative-to-cwd paths generally.  But I'd be comfortable 
>>>with the more explicit sqlite:$CWD/foo
>>
>>Relative paths are useful in my case for a TurboGears project using a
>>small sqlite database (for testing in the early stages). The db uri is
>>stored in a config file which lives in a subversion repository and
>>therefore having to store the absolute path is a pain as it differs per
>>user. Yes we could store the database under /tmp but at the moment it's
>>convenient to include the database in the repository as well.
> 
> 
> In the 0.9 previews of TG (not sure about 0.8) there is a config
> variable "top_level_dir" which points to your top package directory, so
> in my dev.cfg I use this for my DB URI:
> sqlobject.dburi="sqlite:///%(top_level_dir)s/../dev.db"

I think the config file is the right place to handle this, and similarly 
in Paste it would be sqlite://%(here)s/../dev.db.  But I also appreciate 
that this might not be available for every config format, so variable 
substitution might be okay; that it doesn't look that pretty doesn't 
seem like a very meaningful criticism.

Though honestly SQLObject shouldn't really be in the business of doing 
variable substitutions; more layers of substitution and quoting gets 
annoying if you aren't using them.  So maybe it is best just to allow 
relative paths; at least it's pretty obvious what it means.

-- 
Ian Bicking  /  ia...@co...  /  http://blog.ianbicking.org