[SAGA-RG] Comments on the actual specification rather than the words ...

[Andre Merzky]

Hi Steve,

On Tue, Oct 30, 2012 at 4:11 PM, Steve Fisher <dr.s.m.fisher at gmail.com> wrote:
> It is probably worth saying that all classes are of the new style -
> inheriting directly or indirectly from the standard "object".

+1


> We need to state the expected type of the input arguments

This is python, it does not matter!  Just kidding ;-)
Yes, we need that - I started adding that already.


> class SagaException the init method should allow no object to be specified:
>
> __init__ (self, message, object = None)

correct.


> class Url looks more complex than it is. It has a constructor which
> takes a string, a translate method and a bunch of properties.  I am
> not sure that the  string value should be available as a property
> because the string changes value any time that scheme, host etc. are
> modified and vice versa.

Ole and I have been discussing this.  As URL instances are always
local, we think it would suffice to only have the properties?


> What exactly does a line such as:
>
>      scheme = property (get_scheme, set_scheme )
>
> mean?

defines a 'scheme' property with explicit mapping of the setter/getter
to the respective get_scheme/set_scheme methods.


> class Context
> GFD90 does not mention the close() method. Is this here because you
> need some kind of destructor for a context?

Hmm, close should not be needed - I'll check.


> Now for the last page sd and isn
>
> as python takes named parameters we should have all parameters
> defaulting  where reasonable
>
> def __init__ (self, url = None, session=None) : pass # None

ok


> def list_services (self, service_filter = None, data_filter = None,
> authz_filter=None) : pass # [ServiceDescription]

ok


> and for the ISN
>
> def __init__ (self, model, name, filter = None, url=None,
> session=None) : pass # None

Makes sense, too - will fix.

Thanks, Andre.


> Steve



-- 
Nothing is really difficult...