scalar references

[John Hunter]

I have some (potentially mutable) scalar values that I want to share
among many class instances. Changes in one instance should be
reflected across the many instances. A classic Borg is not
appropriate because I have more than one instance.

Use case: in a plotting library I have several figures. Each figure
has a DPI parameter, possibly different between figures, which I want
to share with all the figure components (Axes, Ticks, etc...). I want
to update the DPI in one place and have it shared across all the
relevant components of the figure.

My current approach (which works) is

class RRef:
'A read only ref'
def __init__(self, val):
self._val = val

def get(self):
return self._val

class RWRef(RRef):
'A readable and writable ref'
def set(self, val):
self._val = val

Then I can instantiate a DPI instance as

dpi = RWRef(72)

and __init__ all the figure components with it, which store

self.dpi = dpi

When a figure changes it's DPI via the set method, all the components
have a reference to the mutated value.

I don't really have a problem with this approach, but am wondering if
this is the best (most pythonic!) way to share a mutable scalar value
among several objects.

In addition, I want to do basic binary op arithmetic on read only
references, eg

def bintimes(x,y): return x*y
def binadd(x,y): return x+y
def binsub(x,y): return x-y

class RRef:
'A read only ref'
def __init__(self, val):
self._val = val

def get(self):
return self._val

def __add__(self, other):
return BinOp(self, other, binadd)

def __mul__(self, other):
return BinOp(self, other, bintimes)

def __sub__(self, other):
return BinOp(self, other, binsub)

class BinOp(RRef):
'A read only ref that handles binary ops of refs'
def __init__(self, ref1, ref2, func=binadd):
self.ref1 = ref1
self.ref2 = ref2
self.func = func

def get(self):
return self.func(self.ref1.get(), self.ref2.get())

Thanks for any suggestions!
John Hunter

[Erik Max Francis]

John Hunter wrote:

My current approach (which works) is

class RRef:
'A read only ref'
def __init__(self, val):
self._val = val

def get(self):
return self._val

class RWRef(RRef):
'A readable and writable ref'
def set(self, val):
self._val = val

Yep, this is the kind of thing I usually do. If you'd like to write
shorter, but less self-documenting code, you can just use a one-element
list.

--
Erik Max Francis && ma*@alcyone.com && http://www.alcyone.com/max/
__ San Jose, CA, USA && 37 20 N 121 53 W && &tSftDotIotE
/ \
\__/ Perfect situations must go wrong
-- Florence, _Chess_

[Peter Otten]

John Hunter wrote:

I have some (potentially mutable) scalar values that I want to share
among many class instances. Changes in one instance should be
reflected across the many instances. A classic Borg is not
appropriate because I have more than one instance.

Below is a slight variation of your approach. I think both the beauty and
the danger is that, for read access, attributes defined in the client
instance are indistinguishable from attributes defined in the template
object. I've separated them here, but you could even use a UseTemplate
instance as the template for another instance.

class Template:
def __init__(self, **kwd):
self.__dict__.update(kwd)
german = Template(color="blue", language="german")
french = Template(color="yellow", language="french")

class UseTemplate:
def __init__(self, name, template=german):
self.name = name
self.template = template
def __getattr__(self, name):
"Attributes not found in UseTemplate are looked up in the template"
return getattr(self.template, name)

def printit(u):
print "name=%s, color=%s, language=%s" % (u.name, u.color, u.language)

first = UseTemplate("first", german)
second = UseTemplate("second", french)
special = UseTemplate("special", french)

def printThem():
for u in [first, second, special]:
printit(u)
print

printThem()

# instance attributes shade template attributes
special.color = "black"

# changes in template affect all instances using it
french.color = "red"

printThem()

del special.color # the template attr will reappear
printThem()

Peter