[Tutor] confused: classes containing dicts

[Bob Gailer]

At 08:04 PM 10/8/2004, Calum Mackay wrote:
>Kent Johnson wrote:
>>The first few pages of chapter 21 of Learning Python talk about this a 
>>little.
>
>Having re-read it, I'm not sure it doesn't perpetuate my confusion.
>
>They use the example:
>
>class SharedData:
>     spam = 42
>
>x = SharedData()
>y = SharedData()
>
>and then note that references:
>
>x.spam, y.spam, SharedData.spam
>
>all give the same value. Also, if SharedData.spam is changed, then 
>references of x.spam, y.spam see that change.
>
>However, it then goes onto note that if x.spam is changed, x.spam attaches 
>a name to x itself. i.e:
>
>x.spam = 88
>
>and y.spam, SharedData.spam still give 42.
>
>This is not what happens with dicts, at least in my example:
>
>class SharedData:
>     l = {}
>
>x = SharedData()
>
> >>> x.l["a"] = 123
> >>> x.l
>{'a': 123}
> >>> SharedData.l
>{'a': 123}
>
>Here, assignment to x *has* changed the attr in the class, unlike the 
>following:
>
> >>> x = SharedData()
> >>> x.l
>123
> >>> x.l = 456
> >>> x.l
>456
> >>> SharedData.l
>123
>
>There's a difference in behaviour here, depending on whether the attribute 
>is an int or a dict.

No - there is a difference in the target of the assignment:
 >>> x.l = 456 is an assignment to attribute l of instance x. SInce the 
instance attribute did not exist, it is created.
 >>> x.l["a"] = 123 is an assignment to a key in the dictionary x.l, not to 
the attribute x.
HTH

Bob Gailer
bgailer at alum.rpi.edu
303 442 2625 home
720 938 2625 cell