I happened across the recent discussion and found it very interesting as I have
been dusting off and trying to get ready a module that I had made that created
a number class that handled numeric values and uncertainties in computations by
computing (via overloaded operators) a new value and uncertainty after every
operation. It treats all errors as uncorrelated, however.
Can you use a dictionary of values that gets initialized in the function where
you are doing the calculation to act as your modifiable float? The idea is:
* collect the variable, value and uncertainties in a dictionary
* for some number of iterations,
--tweak all the values
--send them to your function
--have the function start by evaluating those values in its own context then...
--perform the computation as normal and return the result...
--which is accumulated
* compute the average and deviation of the results that were accumulated.
Here's a working example.
Lot's to learn from the recipe that you posted. Thanks!
/chris
###
def meanstdv(x):
from math import sqrt
mean = std = 0
y=x[:]
y.sort()
while len(y)>1:
y.append(y.pop(0)+y.pop(0))
mean = y[0] / float(len(x))
for a in x:
std += (a - mean)**2
std = sqrt(std / float(len(x)-1))
return mean, std
def makecontext( cmds=''):
#execute cmds in clean context (e.g. from math import *)
context = {}
for c in cmds:
if c:
exec(c, context)
return context
def func(context):
#initialize vars in context into this function's context
for _k,_v in context.items():
try:
exec('%s=%s'%(_k,str(_v)))
except:
pass
#now do the calculation and return a result
z=x+y
return z
# these are the input variables with values and uncertainties
vuexpr='''
x=1.2,.1
y=2.3,.2
'''
# load them into a dictionary
vudict={}
for li in vuexpr.strip().splitlines():
expr,u=li.split(',')
vname,vval=expr.split('=')
vudict[vname]=(eval(vval),eval(u))
import random
# accumulate the results
computed_vals=[]
for i in range(1000):
init=[]
for vname,(vval,u) in vudict.items():
expr='%s=%s'%(vname,random.normalvariate(vval,u))
init.append(expr)
computed_vals.append(func(makecontext(init)))
# report the ave, dev of computed value
print meanstdv(computed_vals)
###### output
>>> (3.4963200469919222, 0.22218528738467364)
>>> # using my own routines...
>>> a=phys(1.2,.1);b=phys(2.3,.2); a+b
(3.5,0.22360679775)
>>> print _
3.5(2)
--
http://mail.python.org/mailman/listinfo/python-list
