You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
259 lines
8.2 KiB
259 lines
8.2 KiB
# $Id: errorbar.py,v 1.2 2010-12-01 17:11:46 wirawan Exp $
|
|
#
|
|
# Module wpylib.math.stats.errorbar
|
|
# Errorbar text handling for Python
|
|
#
|
|
# Created: 20081118
|
|
# Wirawan Purwanto
|
|
#
|
|
# Moved to wpylib, 20101007
|
|
|
|
import math
|
|
from math import sqrt
|
|
import os
|
|
import os.path
|
|
import re
|
|
import wpylib.shell_tools as sh
|
|
|
|
class regexp__aux(object):
|
|
'''Auxiliary objects for routines below. Compiled into regexp objects
|
|
for speed.'''
|
|
# CHECKME FIXME: This class is NOT originally designed to be multithread-safe
|
|
# since regex objects are shared between all threads.
|
|
# To be thread-safe, the regex objects or aux classes may need to be
|
|
# instantiated separately for each thread instance, or whatever else way
|
|
# possible.
|
|
init = False
|
|
|
|
@staticmethod
|
|
def initialize():
|
|
R = regexp__aux
|
|
# Each of the regex stuff below is a 2-tuple containing the regex
|
|
# object and its associated function to extract the value/errbar pair
|
|
# (as a tuple) from the re.Match object.
|
|
#
|
|
# Standard errorbar matcher (errorbar in parantheses), in the form of a tuple
|
|
# The first element of the matcher is a regexp matcher object, and the
|
|
# second element is a function to extract the (value,error) tuple from the
|
|
# successful matching process.
|
|
R.errbar = \
|
|
(
|
|
re.compile(r"([-+]?\d+)" # leading digits with optional sign
|
|
r"((?:\.\d*)?)" # optional digits after decimal point
|
|
r"\((\d+\.?\d*[Ee]?[-+]?\d*)\)" # errorbar in paranthesis
|
|
r"((?:[Ee][-+]?\d+)?)$"),
|
|
lambda M: ( float(M.group(1) + M.group(2) + M.group(4)), # value = simple composition
|
|
# for errorbar:
|
|
# - must multiply with value's exponent
|
|
# - and account for decimal digits (if any)
|
|
float(M.group(3)) * float("1"+M.group(4)) * 10**(-max(len(M.group(2))-1, 0)) )
|
|
)
|
|
# Later additional matchers can be added here
|
|
R.init = True
|
|
|
|
@staticmethod
|
|
def aux():
|
|
R = regexp__aux
|
|
if not R.init: R.initialize()
|
|
return R
|
|
|
|
@staticmethod
|
|
def match(matcher, Str, rslt, flatten=False):
|
|
'''Matches the string `Str' against the errorbar regexp pattern in `matcher[0]'.
|
|
If it matches, the value and error are added to the `rslt' list.
|
|
Depending on whether `flatten' is True or not, it is added as a tuple or as
|
|
two elements, respectively, into the `rslt' list.'''
|
|
# Note: matcher is an object like R.errbar above.
|
|
m = matcher[0].match(Str)
|
|
if m:
|
|
if flatten:
|
|
rslt.extend(matcher[1](m))
|
|
else:
|
|
rslt.append(matcher[1](m))
|
|
return True
|
|
else:
|
|
return False
|
|
|
|
def expand(obj, convert_float=False, flatten=False):
|
|
'''Expands compressed errorbar notation to two consecutive numbers
|
|
(returned as a tuple).
|
|
|
|
Input: The input can be a string, or a list of strings.
|
|
|
|
Output:
|
|
The list element that has the format of "VAL(ERR)" (or its scientific
|
|
format twist) will be expanded into two numbers in the output list.
|
|
All other elements will be passed "as is" to the output.
|
|
Optionally, the non-float items can be force-converted to floats if
|
|
the convert_float is set to True.
|
|
'''
|
|
if getattr(obj, "__iter__", False):
|
|
iterable_inp = True
|
|
objs = obj
|
|
else:
|
|
iterable_inp = False
|
|
objs = ( obj, )
|
|
|
|
rgx = regexp__aux.aux()
|
|
|
|
rslt = []
|
|
for o in objs:
|
|
t = type(o)
|
|
if t == int or t == float or t == complex:
|
|
rslt.append(o)
|
|
else:
|
|
# Assume a string!
|
|
o = o.strip()
|
|
#m = rgx.errbar.match(o)
|
|
if (rgx.match(rgx.errbar, o, rslt, flatten)):
|
|
#print "match: errbar"
|
|
pass
|
|
elif convert_float:
|
|
# Convert to float right away, store into the `rslt' list
|
|
rslt.append(float(o))
|
|
#rslt.append( (float(o),) )
|
|
else:
|
|
# Unless otherwise requested, the object will not be converted
|
|
# to float:
|
|
rslt.append(o)
|
|
return rslt
|
|
|
|
|
|
COMPRESS_ERRORBAR_EXE = os.path.dirname(__file__) + "/compress_errorbar.exe"
|
|
|
|
def compress_errorbar_cxx(v, e, errdigits=2):
|
|
"""Temporary plug-hole measure to get python compress errorbars.
|
|
Using a small C++ executable to perform the task."""
|
|
#perl_lib_dir = sh.getenv("WORK_DIR", "HOME") + "/scripts"
|
|
return sh.pipe_out((COMPRESS_ERRORBAR_EXE, str(v), str(e), str(errdigits))).strip()
|
|
|
|
|
|
class float_decomp(object):
|
|
"""Floating-point decomposer.
|
|
We are assuming IEEE double precision here."""
|
|
|
|
def __init__(self, val, decdigits=16):
|
|
self.val = val
|
|
V = "%+.*e" % (decdigits, val)
|
|
self.sign = V[0]
|
|
self.digits = V[1] + V[3:3+16]
|
|
self.exp = int(V[20:])
|
|
|
|
set = __init__
|
|
|
|
|
|
|
|
class errorbar(object):
|
|
"""A simple class holding a scalar value with an error bar.
|
|
When converted to a float, its mean is returned.
|
|
When converted to a string, its string representation is returned.
|
|
which usually is meant to be a value with errorbar in parenthesis.
|
|
This value is custom-made."""
|
|
|
|
ebproc = staticmethod(compress_errorbar_cxx)
|
|
|
|
def __init__(self, val, err, eb=None, ebproc=None):
|
|
self.val = val
|
|
self.err = err
|
|
if ebproc != None:
|
|
self.ebproc = ebproc
|
|
if eb == None:
|
|
self.eb = self.ebproc(val, err)
|
|
else:
|
|
self.eb = eb
|
|
|
|
def __float__(self):
|
|
return self.val
|
|
def __float__(self):
|
|
return self.val
|
|
value = __float__
|
|
mean = __float__
|
|
def error(self):
|
|
return self.err
|
|
def __str__(self):
|
|
if getattr(self, "eb", None):
|
|
return self.eb
|
|
else:
|
|
return "%g +- %g" % (self.val, self.err)
|
|
display = __str__
|
|
def __repr__(self):
|
|
return "errorbar(%s,%s,'%s')" % (self.val, self.err, self.display())
|
|
def ebupdate(self):
|
|
self.eb = self.ebproc(self.val, self.err)
|
|
def copy(self):
|
|
return self.__class__(self.val, self.err, self.eb, self.ebproc)
|
|
# Some algebraic operations with scalars are defined here:
|
|
def __mul__(self, y):
|
|
"""Scales by a scalar value."""
|
|
return self.__class__(self.val*y, self.err*y, ebproc=self.ebproc)
|
|
__rmul__ = __mul__
|
|
def __imul__(self, y):
|
|
"""Scales itself by a scalar value."""
|
|
self.val *= y
|
|
self.err *= y
|
|
self.ebupdate()
|
|
def __div__(self, y):
|
|
"""Divides by a scalar value."""
|
|
return self.__class__(self.val/y, self.err/y, ebproc=self.ebproc)
|
|
def __idiv__(self, y):
|
|
"""Scales itself by a scalar value (division)."""
|
|
self.val /= y
|
|
self.err /= y
|
|
self.ebupdate()
|
|
|
|
def __add__(self, y):
|
|
"""Adds by a scalar value or another errorbar value.
|
|
In the latter case, the uncertainty is assumed to be uncorrelated,
|
|
thus we can use a simple formula to update the errorbar."""
|
|
if isinstance(y, errorbar):
|
|
return self.__class__(self.val+y.val,
|
|
sqrt(self.err**2 + y.err**2),
|
|
ebproc=self.ebproc)
|
|
else:
|
|
return self.__class__(self.val+y, self.err, ebproc=self.ebproc)
|
|
__radd__ = __add__
|
|
|
|
def __sub__(self, y):
|
|
"""Subtracts by a scalar value or another errorbar value.
|
|
In the latter case, the uncertainty is assumed to be uncorrelated,
|
|
thus we can use a simple formula to update the errorbar."""
|
|
if isinstance(y, errorbar):
|
|
return self.__class__(self.val-y.val,
|
|
sqrt(self.err**2 + y.err**2),
|
|
ebproc=self.ebproc)
|
|
else:
|
|
return self.__class__(self.val-y, self.err, ebproc=self.ebproc)
|
|
|
|
def __rsub__(self, y):
|
|
"""Subtracts this errorbar value from a scalar value or another
|
|
errorbar value.
|
|
In the latter case, the uncertainty is assumed to be uncorrelated,
|
|
thus we can use a simple formula to update the errorbar."""
|
|
if isinstance(y, errorbar):
|
|
return self.__class__(y.val-self.val,
|
|
sqrt(self.err**2 + y.err**2),
|
|
ebproc=self.ebproc)
|
|
else:
|
|
return self.__class__(y-self.val, self.err, ebproc=self.ebproc)
|
|
|
|
@staticmethod
|
|
def create_str(s):
|
|
"""Creates an errorbar object from an errorbar string."""
|
|
eb = expand(s, convert_float=True)[0]
|
|
return errorbar(*eb)
|
|
|
|
|
|
class errorbar_compressor(object):
|
|
"""Compressor for errorbar string."""
|
|
def __init__(self):
|
|
self.errdigits = 2
|
|
def __call__(self, val, err, **args):
|
|
errdigits = args.get("errdigits", self.errdigits)
|
|
v = float_decomp(val)
|
|
e = float_decomp(err, decdigits=errdigits)
|
|
|
|
|
|
|
|
#def errorbar_algebra
|
|
|
|
|
|
|