Re: [Pytables-users] __getattr__ vs. __getitem__ (Natural naming)

[Norbert N. <Nor...@gm...>]

Am Sonntag, 16. Januar 2005 16:35 schrieb Ivan Vilata i Balaguer:
> [...]
> Please note I that intentionally left out 
> __set{item,attr}__ because this is a greater change.  File.create*()
> calls would be changed to direct constructor invocation for leaves and
> groups, a thing which I definetly think Right and perfectly plausible
> (but I might be wrong).

I don't quite understand: it already is possible to do
 mygroup.somename = Array(somevalue)
to create an array. Where is the "greater change"? Actually, I never liked the 
File.create*() method to create elements.

> [...]
> (but they would be kept for compatibility).

I would recommend spicing as much as possible with DeprecatedWarning messages. 
Offering too many alternatives is more confusing than helpful.

> Imagine issuing: 
>   >>> # h5file.root.detector.readout <=>
>   >>> tbl = h5file.root['detector/readout']

or
 tbl = h5file.root['detector']['readout']

Anyhow: I don't see much gain here. Just the question whether you do the '/' 
parsing or not. Don't have much of an opinion there.

>   >>> # h5file.root.column['name column'].attrs.temperature <=>
>   >>> attr = h5file.root['column/name column/attrs/temperature']

I have no idea what that's supposed to mean...

>     I think this should be quite easy to implement and would make
> building pathnames and traversing them a pleasure, keeping interactive
> behaviour the same.  Maybe even File.__???{item,attr}__ could be
> tailored to defer to the root group object invocations using an item /
> attribute starting with '/', so that using h5file['/detector/readout']
> would be possible...



-- 
_________________________________________Norbert Nemec
         Bernhardstr. 2 ... D-93053 Regensburg
     Tel: 0941 - 2009638 ... Mobil: 0179 - 7475199
           eMail: <No...@Ne...>