> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
> Jason van Zyl
>
> .....
>
> Ok, now I remeber why I did it as well. Looking at
>
> 1)
> resource.loader = file
> resource.loader = classpath
>
> resource.loader.file.<prop> = <value>
>
> versus
>
> 2)
> resource.loader = file
> resource.loader = classpath
>
> file.resource.loader.<prop> = <value>
>
> If I want to grab just the resource loaders then I will use
>
> Configuration. subset("resource.loader") I'm going to get all the config
> stuff for the resource loaders too, I don't want that.
>
> Where in 2) performing Configuration.subset("resource.loader")
> will simply give me back the two configuration entries I expect.
> Then I can use those values ('file', 'classpath') to retrieve the
> values for each individual loader.
>
> So I think we have to keep it in the form
>
> <id>.resource.loader.<prop> = value
>
> In order to make it easy to parse internally, I think it reads more
> like a sentence that way too.
>
> Is this ok?
Well, you should always check the output of
Configuration.subset("resource.loader")
to be sure that all returned nodes have just a "resource.loader" key and
not a "resource.loader.something" like key.
IMO, the clear advantage of the ordering you defend is a slight gain in
average performance during initialization. But I think that we all agree
that initialization performance is not such an important issue.
The other advantage - reading more like a sentence - is not so clear for
me. Most guys reading/writing these files are programmers used to coded
stuff (instead of English teachers or so).
Taking myself as an example, it takes me much less effort to check
code/configurations when identifiers in its text are aligned and a
logical group of identifiers has the same prefix - that way it is easier
to follow the "different" part of the identifier.
All this to say that my personal preference is still with 1).
=:o)
Have fun,
Paulo Gaspar
