Assigning to self.__class__

[Paul McGuire]

I have some places in pyparsing where I've found that the most
straightforward way to adjust an instance's behavior is to change its class.
I do this by assigning to self.__class__, and things all work fine.

(Converting to use of __new__ is not an option - in one case, the change is
temporary, and before the end of the function, I change it back again.)

Any comments on this practice? Is this intended capability for Python
objects, or am I taking advantage of a fortuitous accident, which may get
undone at a future time?

-- Paul

[bruno at modulix]

Paul McGuire wrote:

I have some places in pyparsing where I've found that the most
straightforward way to adjust an instance's behavior is to change its class.
Hooray ! You've just (re)discovered the state pattern... for which the
most stupid simple implementation in Python is to :
(snip) assigning to self.__class__, (snip)
!-)

Any comments on this practice?
It can be very confusing for newbies and peoples having no experience
with *dynamic* languages, and I guess control-freaks and
static-typing-addicts would runaway screaming. But I like it anyway !-)
Is this intended capability for Python
objects,
AFAIK, yes.
or am I taking advantage of a fortuitous accident, which may get
undone at a future time?

It's certainly not a fortuitous accident.
--
bruno desthuilliers
python -c "print '@'.join(['.'.join([w[::-1] for w in p.split('.')]) for
p in 'o****@xiludom.gro'.split('@')])"

[Heiko Wundram]

bruno at modulix wrote:

Paul McGuire wrote:

or am I taking advantage of a fortuitous accident, which may get
undone at a future time?

It's certainly not a fortuitous accident.

And even the (printed) cookbook has examples which assign to
self.__class__... I guess this means this feature isn't going to go away
soo. ;-)

--- Heiko.

[Alex Martelli]

Heiko Wundram <mo*******@bit-bukket.org> wrote:

bruno at modulix wrote:

Paul McGuire wrote:

or am I taking advantage of a fortuitous accident, which may get
undone at a future time?

It's certainly not a fortuitous accident.

And even the (printed) cookbook has examples which assign to
self.__class__... I guess this means this feature isn't going to go away

Hmmm, I'm the main decision-maker for what goes in the printed cookbook,
but it's Guido who decides what stays or goes in the language, and it's
not as if we always agree (though I will admit that when we disagree it
later usually turns out he's right, but that's another issue;-). Still,
in this case I don't think we disagree!-)
Alex

[Terry Reedy]

"Paul McGuire" <pt***@austin.rr._bogus_.com> wrote in message
news:_G***************@tornado.texas.rr.com...

I have some places in pyparsing where I've found that the most
straightforward way to adjust an instance's behavior is to change its
class.
I do this by assigning to self.__class__, and things all work fine.

(Converting to use of __new__ is not an option - in one case, the change
is
temporary, and before the end of the function, I change it back again.)

Any comments on this practice? Is this intended capability for Python
objects, or am I taking advantage of a fortuitous accident, which may get
undone at a future time?

Very little if any of the new class stuff is accidental. As I remember,
..__class__ was originally read-only and was intentionally changed when
practical use cases for rewriting it were developed and presented with a
request for change. Though I could be wrong, I would comfortably use this
feature myself.

Terry J. Reedy