[Configobj-develop] Re: Validate - List Default Values

[j v. <jgv...@co...>]

I think indicating list values by surrounding with square brackets :

    checkname(default="[ 'val 1', 'val 2', 'val 3']")

is preferable.

-- jv


On Tue, 2005-12-13 at 10:14 +0000, Fuzzyman wrote:
> Hello all,
> 
> There is a bug in the current validate code.
> 
> It doesn't correctly handle lists as default values - how did that get
> past the tests ?
> 
> In fact commas in default values *seem* to break it completely. No worry
> - I can fix this, so far only with a *truly horrible* set of regular
> expressions, but I'm sure they can be optimised.
> 
> The question is how do we want to specify list values as defaults in the
> configspec ?
> 
> The current rationale (which doesn't work) uses the same syntax as
> ConfigObj. This might sound logical - but it has an unfortunate side effect.
> 
> ConfigObj syntax is to treat any value with commas in as a list - 
> unless it is quoted. Because configspec values are part of an argument
> list they need to be quoted anyway. this means the current syntax for a
> normal value with commas is
> 
>     checkname(default="'a value, with commas'")
> 
> Notice the *double* quotes ! Truly horrible IMO.
> 
> The alternative would be to indicate list values by surrounding with
> square brackets :
> 
>     checkname(default="[ 'val 1', 'val 2', 'val 3']")
> 
> The advantage is that it means you don't need to double quote default
> values with commas. The disadvantage is that we now have a different
> syntax for lists in the configspec and in the ConfigObj.
> 
> I don't think this is such a problem.
> 
> I wouldn't (initially) add any escape values - this would mean you
> couldn't use quotes or square brackets inside list members.
> 
> The alternative is to build a 'full scale' parser for the configspec -
> which I am reluctant to do. An alternative way of specifying defaults is
> to use merge (the recursive update which will be in the new release of
> ConfigObj).
> 
> Feedback appreciated - this changes the syntax of the configspec, so
> it's a (relatively) important decision.
> 
> Thanks
> 
> Fuzzyman
> http://www.voidspace.org.uk/python/index.shtml