How to set precision without rounding? - python

I have big integers in database (numeric(24,0)). And I want to convert this number to "human readable" format. I have function:
def from_int(value, precision):
fprec = "%." + str(precision) + "f"
return fprec % (Decimal(value) * (Decimal(10)**(-1*precision)))
And it works:
from_int(1000000, 6)
'1.000000'
from_int(1000000, 8)
'0.01000000'
from_int(1000000, 12)
'0.000001000000'
from_int(1000000, 2)
'10000.00'
but for:
from_int(19999999999999999, 2)
'200000000000000.00'
How to set precision without rounding?

Formatting with %f converts the number to floating point losing precision. Instead use str:
def from_int(value, precision):
return str(Decimal(value) * Decimal(10)**(-precision))

The problem is in the %f construct that expects a float. So even if you compute everything in decimal, you have a nasty float conversion at the end => so the rounding issue. I think you could just write :
def from_int(value, precision):
return str((Decimal(value) * (Decimal(10)**(-1*precision))))
I tested it on your values and it gives correct results.

If all you want is a string representation, you may be able to get away with something like this:
>>> ('%%.%df' % 2) % 2.445
'2.44'
so in your original code:
def from_int(value, precision):
return ('%%.%df' % precision) % value
EDIT:
To deal with precision issue, it's indeed simpler to use Decimal type, and the quantize method:
>>> d = decimal.Decimal(19999999999999999)
>>> prec = 2
>>> str(d.quantize(decimal.Decimal(10) ** -prec))
'19999999999999999.00'
>>>
ref: How can I format a decimal to always show 2 decimal places?
EDIT: Using the Decimal type, you can shift decimal places without rounding:
>>> str(d * (decimal.Decimal(10) ** -prec))
'199999999999999.99'

Related

Round up to 2 decimal in Python

I have a beginner problem. How can I round up to 2 decimal?
Here is what I tried and what I want to achieve:
import math
var_1 = 14.063 # expected = 14.06
var_2 = 10.625 # expected = 10.63
print(round(14.063, 2))
print(round(10.625, 2))
print('===========================')
def round_up(n, decimals=0):
multiplier = 10 ** decimals
return math.ceil(n * multiplier) / multiplier
print(round_up(var_1, 2))
print(round_up(var_2, 2))
And the Output is:
14.06
10.62
===========================
14.07
10.63
So neither of those wroks for me...
The Decimal class, quantize() method, and ROUND_HALF_UP rule from the decimal module can handle this:
from decimal import Decimal, ROUND_HALF_UP
var_1 = 14.063 # expected = 14.06
var_2 = 10.625 # expected = 10.63
# a Decimal object with an explicit exponent attribute/property (to be interpreted by quantize)
Two_places = Decimal("1e-2")
for var in [var_1, var_2]:
rounded = Decimal(var).quantize(Two_places, rounding=ROUND_HALF_UP)
print(f"decimal: {rounded}")
print(f"float: {float(rounded)}")
and I get:
decimal: 14.06
float: 14.06
decimal: 10.63
float: 10.63
Keep in mind that when you're dealing with floats, you're always manipulating a less-than-precise representation of what you probably (naturally) have in mind:
Decimal(1.65) # Decimal('1.649999999999999911182158029987476766109466552734375')
Decimal('1.65') # Decimal('1.65')
In the first case, 1.65 was first turned into an IEEE-754 float, which has precision errors going from base-10 to base-2, then passed to Decimal. In the second case, Decimal interpreted the number as "one, and 65 100-ths" which equates to "165 times 10 raised to the minus 2", or 165e-2.
Try this. This finds the nearest one and if not, then round up -
import math
v1 = 14.063
v2 = 10.625
def round_up(n, decimals=0):
multiplier = 10 ** decimals
var_down = round(n, 2)
var_up = math.ceil(n * multiplier) / multiplier
if n - var_down >= var_up - n:
return var_up
else:
return var_down
v1_round = round_up(v1, 2)
v2_round = round_up(v2, 2)
print (v1_round) # 14.06
print (v2_round) # 10.63
this should work, although there is probebly a more efficient way of doing it. I just took your code and determined which one was closer, and if they are the same to round up.
Edit: It seems that PCM has made such version.
import math
decimals = 2
var_1 = 14.063
var_2 = 10.625
var_1down = round(var_1, decimals)
var_2down = round(var_2, decimals)
def round_up(n, decimals=0):
multiplier = 10 ** decimals
return math.ceil(n * multiplier) / multiplier
var_1up = round_up(var_1, decimals)
var_2up = round_up(var_2, decimals)
if var_1 - var_1down >= var_1up - var_1:
var_1round = var_1up
else:
var_1round = var_1down
if var_2 - var_2down >= var_2up - var_2:
var_2round = var_2up
else:
var_2round = var_2down
print (var_1round)
print (var_2round)
If you check the docs you will see that "values are rounded to the closest multiple of 10 to the power minus ndigits; if two multiples are equally close, rounding is done toward the even choice (so, for example, both round(0.5) and round(-0.5) are 0, and round(1.5) is 2)".
So 10.625 rounds to 10.62. You may try adding a very small value, e.g. 0.00001, but even so, since the way float numbers work, you may have some surprise in a few cases.

Format decimal without trailing zeros [duplicate]

I have a long list of Decimals and that I have to adjust by factors of 10, 100, 1000,..... 1000000 depending on certain conditions. When I multiply them there is sometimes a useless trailing zero (though not always) that I want to get rid of. For example...
from decimal import Decimal
# outputs 25.0, PROBLEM! I would like it to output 25
print Decimal('2.5') * 10
# outputs 2567.8000, PROBLEM! I would like it to output 2567.8
print Decimal('2.5678') * 1000
Is there a function that tells the decimal object to drop these insignificant zeros? The only way I can think of doing this is to convert to a string and replace them using regular expressions.
Should probably mention that I am using python 2.6.5
EDIT
senderle's fine answer made me realize that I occasionally get a number like 250.0 which when normalized produces 2.5E+2. I guess in these cases I could try to sort them out and convert to a int
You can use the normalize method to remove extra precision.
>>> print decimal.Decimal('5.500')
5.500
>>> print decimal.Decimal('5.500').normalize()
5.5
To avoid stripping zeros to the left of the decimal point, you could do this:
def normalize_fraction(d):
normalized = d.normalize()
sign, digits, exponent = normalized.as_tuple()
if exponent > 0:
return decimal.Decimal((sign, digits + (0,) * exponent, 0))
else:
return normalized
Or more compactly, using quantize as suggested by user7116:
def normalize_fraction(d):
normalized = d.normalize()
sign, digit, exponent = normalized.as_tuple()
return normalized if exponent <= 0 else normalized.quantize(1)
You could also use to_integral() as shown here but I think using as_tuple this way is more self-documenting.
I tested these both against a few cases; please leave a comment if you find something that doesn't work.
>>> normalize_fraction(decimal.Decimal('55.5'))
Decimal('55.5')
>>> normalize_fraction(decimal.Decimal('55.500'))
Decimal('55.5')
>>> normalize_fraction(decimal.Decimal('55500'))
Decimal('55500')
>>> normalize_fraction(decimal.Decimal('555E2'))
Decimal('55500')
There's probably a better way of doing this, but you could use .rstrip('0').rstrip('.') to achieve the result that you want.
Using your numbers as an example:
>>> s = str(Decimal('2.5') * 10)
>>> print s.rstrip('0').rstrip('.') if '.' in s else s
25
>>> s = str(Decimal('2.5678') * 1000)
>>> print s.rstrip('0').rstrip('.') if '.' in s else s
2567.8
And here's the fix for the problem that #gerrit pointed out in the comments:
>>> s = str(Decimal('1500'))
>>> print s.rstrip('0').rstrip('.') if '.' in s else s
1500
Answer from the Decimal FAQ in the documentation:
>>> def remove_exponent(d):
... return d.quantize(Decimal(1)) if d == d.to_integral() else d.normalize()
>>> remove_exponent(Decimal('5.00'))
Decimal('5')
>>> remove_exponent(Decimal('5.500'))
Decimal('5.5')
>>> remove_exponent(Decimal('5E+3'))
Decimal('5000')
Answer is mentioned in FAQ (https://docs.python.org/2/library/decimal.html#decimal-faq) but does not explain things.
To drop trailing zeros for fraction part you should use normalize:
>>> Decimal('100.2000').normalize()
Decimal('100.2')
>> Decimal('0.2000').normalize()
Decimal('0.2')
But this works different for numbers with leading zeros in sharp part:
>>> Decimal('100.0000').normalize()
Decimal('1E+2')
In this case we should use `to_integral':
>>> Decimal('100.000').to_integral()
Decimal('100')
So we could check if there's a fraction part:
>>> Decimal('100.2000') == Decimal('100.2000').to_integral()
False
>>> Decimal('100.0000') == Decimal('100.0000').to_integral()
True
And use appropriate method then:
def remove_exponent(num):
return num.to_integral() if num == num.to_integral() else num.normalize()
Try it:
>>> remove_exponent(Decimal('100.2000'))
Decimal('100.2')
>>> remove_exponent(Decimal('100.0000'))
Decimal('100')
>>> remove_exponent(Decimal('0.2000'))
Decimal('0.2')
Now we're done.
Use the format specifier %g. It seems remove to trailing zeros.
>>> "%g" % (Decimal('2.5') * 10)
'25'
>>> "%g" % (Decimal('2.5678') * 1000)
'2567.8'
It also works without the Decimal function
>>> "%g" % (2.5 * 10)
'25'
>>> "%g" % (2.5678 * 1000)
'2567.8'
I ended up doing this:
import decimal
def dropzeros(number):
mynum = decimal.Decimal(number).normalize()
# e.g 22000 --> Decimal('2.2E+4')
return mynum.__trunc__() if not mynum % 1 else float(mynum)
print dropzeros(22000.000)
22000
print dropzeros(2567.8000)
2567.8
note: casting the return value as a string will limit you to 12 significant digits
Slightly modified version of A-IV's answer
NOTE that Decimal('0.99999999999999999999999999995').normalize() will round to Decimal('1')
def trailing(s: str, char="0"):
return len(s) - len(s.rstrip(char))
def decimal_to_str(value: decimal.Decimal):
"""Convert decimal to str
* Uses exponential notation when there are more than 4 trailing zeros
* Handles decimal.InvalidOperation
"""
# to_integral_value() removes decimals
if value == value.to_integral_value():
try:
value = value.quantize(decimal.Decimal(1))
except decimal.InvalidOperation:
pass
uncast = str(value)
# use exponential notation if there are more that 4 zeros
return str(value.normalize()) if trailing(uncast) > 4 else uncast
else:
# normalize values with decimal places
return str(value.normalize())
# or str(value).rstrip('0') if rounding edgecases are a concern
You could use :g to achieve this:
'{:g}'.format(3.140)
gives
'3.14'
This should work:
'{:f}'.format(decimal.Decimal('2.5') * 10).rstrip('0').rstrip('.')
Just to show a different possibility, I used to_tuple() to achieve the same result.
def my_normalize(dec):
"""
>>> my_normalize(Decimal("12.500"))
Decimal('12.5')
>>> my_normalize(Decimal("-0.12500"))
Decimal('-0.125')
>>> my_normalize(Decimal("0.125"))
Decimal('0.125')
>>> my_normalize(Decimal("0.00125"))
Decimal('0.00125')
>>> my_normalize(Decimal("125.00"))
Decimal('125')
>>> my_normalize(Decimal("12500"))
Decimal('12500')
>>> my_normalize(Decimal("0.000"))
Decimal('0')
"""
if dec is None:
return None
sign, digs, exp = dec.as_tuple()
for i in list(reversed(digs)):
if exp >= 0 or i != 0:
break
exp += 1
digs = digs[:-1]
if not digs and exp < 0:
exp = 0
return Decimal((sign, digs, exp))
Why not use modules 10 from a multiple of 10 to check if there is remainder? No remainder means you can force int()
if (x * 10) % 10 == 0:
x = int(x)
x = 2/1
Output: 2
x = 3/2
Output: 1.5

How to print negative zero in Python

I am converting decimal degrees to print as DMS. The conversion algorithm is what you would expect, using modf, with the addition that sign is taken out of the MS portion and left in only for the D portion.
Everything is fine except for the case where the Degree is negative zero, -0. An example is -0.391612 which should print as -0°23'29".
"%d" drops the negative sign. What format string can I use to print -0?
I've worked around it with a kludge that converts the numbers to strings and prepends a "-" if negative, then uses "%s" as the format. It's awkward and feels inelegant.
Here's the code:
def dec_to_DMS(decimal_deg):
deg = modf(decimal_deg)[1]
deg_ = fabs(modf(decimal_deg)[0])
min = modf(deg_ * 60)[1]
min_ = modf(deg_ * 60)[0]
sec = modf(min_ * 60)[1]
return deg,min,sec
def print_DMS(dms): # dms is a tuple
# make sure the "-" is printed for -0.xxx
format = ("-" if copysign(1,dms[0]) < 0 else "") + "%d°%02d'%02d\""
return format % dms
print print_DMS(dec_to_DMS(-0.391612))
>>> -0°23'29"
deg_ is to prevent the function returning (-0,-23,-29); it returns the correct (-0,23,29).
Use format()
>>> format(-0.0)
'-0.0'
>>> format(0.0)
'0.0'
>>> print '''{: g}°{}'{}"'''.format(-0.0, 23, 29)
-0°23'29"
You must print the degrees as a floating point value, not as an integer, as 2 complement integers (used on most platforms) do not have a negative zero. %d is for formatting integers.
print u'{:.0f}°{:.0f}\'{:.0f}\"'.format(deg, fabs(min), fabs(sec)).encode('utf-8')
deg, min and sec are all floats, the fabs calls are there so that the respective signs are not printed.
Alternatively using the old format string style:
print (u'%.0f°%.0f\'%.0f\"' % (deg, fabs(min), fabs(sec))).encode('utf-8')

Convert base-2 binary number string to int

I'd simply like to convert a base-2 binary number string into an int, something like this:
>>> '11111111'.fromBinaryToInt()
255
Is there a way to do this in Python?
You use the built-in int() function, and pass it the base of the input number, i.e. 2 for a binary number:
>>> int('11111111', 2)
255
Here is documentation for Python 2, and for Python 3.
Just type 0b11111111 in python interactive interface:
>>> 0b11111111
255
Another way to do this is by using the bitstring module:
>>> from bitstring import BitArray
>>> b = BitArray(bin='11111111')
>>> b.uint
255
Note that the unsigned integer (uint) is different from the signed integer (int):
>>> b.int
-1
Your question is really asking for the unsigned integer representation; this is an important distinction.
The bitstring module isn't a requirement, but it has lots of performant methods for turning input into and from bits into other forms, as well as manipulating them.
Using int with base is the right way to go. I used to do this before I found int takes base also. It is basically a reduce applied on a list comprehension of the primitive way of converting binary to decimal ( e.g. 110 = 2**0 * 0 + 2 ** 1 * 1 + 2 ** 2 * 1)
add = lambda x,y : x + y
reduce(add, [int(x) * 2 ** y for x, y in zip(list(binstr), range(len(binstr) - 1, -1, -1))])
If you wanna know what is happening behind the scene, then here you go.
class Binary():
def __init__(self, binNumber):
self._binNumber = binNumber
self._binNumber = self._binNumber[::-1]
self._binNumber = list(self._binNumber)
self._x = [1]
self._count = 1
self._change = 2
self._amount = 0
print(self._ToNumber(self._binNumber))
def _ToNumber(self, number):
self._number = number
for i in range (1, len (self._number)):
self._total = self._count * self._change
self._count = self._total
self._x.append(self._count)
self._deep = zip(self._number, self._x)
for self._k, self._v in self._deep:
if self._k == '1':
self._amount += self._v
return self._amount
mo = Binary('101111110')
Here's another concise way to do it not mentioned in any of the above answers:
>>> eval('0b' + '11111111')
255
Admittedly, it's probably not very fast, and it's a very very bad idea if the string is coming from something you don't have control over that could be malicious (such as user input), but for completeness' sake, it does work.
A recursive Python implementation:
def int2bin(n):
return int2bin(n >> 1) + [n & 1] if n > 1 else [1]
If you are using python3.6 or later you can use f-string to do the
conversion:
Binary to decimal:
>>> print(f'{0b1011010:#0}')
90
>>> bin_2_decimal = int(f'{0b1011010:#0}')
>>> bin_2_decimal
90
binary to octal hexa and etc.
>>> f'{0b1011010:#o}'
'0o132' # octal
>>> f'{0b1011010:#x}'
'0x5a' # hexadecimal
>>> f'{0b1011010:#0}'
'90' # decimal
Pay attention to 2 piece of information separated by colon.
In this way, you can convert between {binary, octal, hexadecimal, decimal} to {binary, octal, hexadecimal, decimal} by changing right side of colon[:]
:#b -> converts to binary
:#o -> converts to octal
:#x -> converts to hexadecimal
:#0 -> converts to decimal as above example
Try changing left side of colon to have octal/hexadecimal/decimal.
For large matrix (10**5 rows and up) it is better to use a vectorized matmult. Pass in all rows and cols in one shot. It is extremely fast. There is no looping in python here. I originally designed it for converting many binary columns like 0/1 for like 10 different genre columns in MovieLens into a single integer for each example row.
def BitsToIntAFast(bits):
m,n = bits.shape
a = 2**np.arange(n)[::-1] # -1 reverses array of powers of 2 of same length as bits
return bits # a
For the record to go back and forth in basic python3:
a = 10
bin(a)
# '0b1010'
int(bin(a), 2)
# 10
eval(bin(a))
# 10

How to get numbers after decimal point?

How do I get the numbers after a decimal point?
For example, if I have 5.55, how do i get .55?
5.55 % 1
Keep in mind this won't help you with floating point rounding problems. I.e., you may get:
0.550000000001
Or otherwise a little off the 0.55 you are expecting.
Use modf:
>>> import math
>>> frac, whole = math.modf(2.5)
>>> frac
0.5
>>> whole
2.0
What about:
a = 1.3927278749291
b = a - int(a)
b
>> 0.39272787492910011
Or, using numpy:
import numpy
a = 1.3927278749291
b = a - numpy.fix(a)
Using the decimal module from the standard library, you can retain the original precision and avoid floating point rounding issues:
>>> from decimal import Decimal
>>> Decimal('4.20') % 1
Decimal('0.20')
As kindall notes in the comments, you'll have to convert native floats to strings first.
An easy approach for you:
number_dec = str(number-int(number))[1:]
Try Modulo:
5.55%1 = 0.54999999999999982
To make it work with both positive and negative numbers:
try abs(x)%1. For negative numbers, without with abs, it will go wrong.
5.55 % 1
output 0.5499999999999998
-5.55 % 1
output 0.4500000000000002
import math
orig = 5.55
whole = math.floor(orig) # whole = 5.0
frac = orig - whole # frac = 0.55
similar to the accepted answer, even easier approach using strings would be
def number_after_decimal(number1):
number = str(number1)
if 'e-' in number: # scientific notation
number_dec = format(float(number), '.%df'%(len(number.split(".")[1].split("e-")[0])+int(number.split('e-')[1])))
elif "." in number: # quick check if it is decimal
number_dec = number.split(".")[1]
return number_dec
>>> n=5.55
>>> if "." in str(n):
... print "."+str(n).split(".")[-1]
...
.55
Just using simple operator division '/' and floor division '//' you can easily get the fraction part of any given float.
number = 5.55
result = (number/1) - (number//1)
print(result)
Sometimes trailing zeros matter
In [4]: def split_float(x):
...: '''split float into parts before and after the decimal'''
...: before, after = str(x).split('.')
...: return int(before), (int(after)*10 if len(after)==1 else int(after))
...:
...:
In [5]: split_float(105.10)
Out[5]: (105, 10)
In [6]: split_float(105.01)
Out[6]: (105, 1)
In [7]: split_float(105.12)
Out[7]: (105, 12)
Another example using modf
from math import modf
number = 1.0124584
# [0] decimal, [1] integer
result = modf(number)
print(result[0])
# output = 0124584
print(result[1])
# output = 1
This is a solution I tried:
num = 45.7234
(whole, frac) = (int(num), int(str(num)[(len(str(int(num)))+1):]))
Float numbers are not stored in decimal (base10) format. Have a read through the python documentation on this to satisfy yourself why. Therefore, to get a base10 representation from a float is not advisable.
Now there are tools which allow storage of numeric data in decimal format. Below is an example using the Decimal library.
from decimal import *
x = Decimal('0.341343214124443151466')
str(x)[-2:] == '66' # True
y = 0.341343214124443151466
str(y)[-2:] == '66' # False
Use floor and subtract the result from the original number:
>> import math #gives you floor.
>> t = 5.55 #Give a variable 5.55
>> x = math.floor(t) #floor returns t rounded down to 5..
>> z = t - x #z = 5.55 - 5 = 0.55
Example:
import math
x = 5.55
print((math.floor(x*100)%100))
This is will give you two numbers after the decimal point, 55 from that example. If you need one number you reduce by 10 the above calculations or increase depending on how many numbers you want after the decimal.
import math
x = 1245342664.6
print( (math.floor(x*1000)%1000) //100 )
It definitely worked
Another option would be to use the re module with re.findall or re.search:
import re
def get_decimcal(n: float) -> float:
return float(re.search(r'\.\d+', str(n)).group(0))
def get_decimcal_2(n: float) -> float:
return float(re.findall(r'\.\d+', str(n))[0])
def get_int(n: float) -> int:
return int(n)
print(get_decimcal(5.55))
print(get_decimcal_2(5.55))
print(get_int(5.55))
Output
0.55
0.55
5
If you wish to simplify/modify/explore the expression, it's been explained on the top right panel of regex101.com. If you'd like, you can also watch in this link, how it would match against some sample inputs.
Source
How to get rid of additional floating numbers in python subtraction?
You can use this:
number = 5.55
int(str(number).split('.')[1])
This is only if you want toget the first decimal
print(int(float(input()) * 10) % 10)
Or you can try this
num = float(input())
b = num - int(num)
c = b * 10
print(int(c))
Using math module
speed of this has to be tested
from math import floor
def get_decimal(number):
'''returns number - floor of number'''
return number-floor(number)
Example:
n = 765.126357123
get_decimal(n)
0.12635712300004798
def fractional_part(numerator, denominator):
# Operate with numerator and denominator to
# keep just the fractional part of the quotient
if denominator == 0:
return 0
else:
return (numerator/ denominator)-(numerator // denominator)
print(fractional_part(5, 5)) # Should be 0
print(fractional_part(5, 4)) # Should be 0.25
print(fractional_part(5, 3)) # Should be 0.66...
print(fractional_part(5, 2)) # Should be 0.5
print(fractional_part(5, 0)) # Should be 0
print(fractional_part(0, 5)) # Should be 0
Easier if the input is a string, we can use split()
decimal = input("Input decimal number: ") #123.456
# split 123.456 by dot = ['123', '456']
after_coma = decimal.split('.')[1]
# because only index 1 is taken then '456'
print(after_coma) # '456'
if you want to make a number type
print(int(after_coma)) # 456
a = 12.587
b = float('0.' + str(a).split('.')[-1])
What about:
a = 1.234
b = a - int(a)
length = len(str(a))
round(b, length-2)
Output:
print(b)
0.23399999999999999
round(b, length-2)
0.234
Since the round is sent to a the length of the string of decimals ('0.234'), we can just minus 2 to not count the '0.', and figure out the desired number of decimal points. This should work most times, unless you have lots of decimal places and the rounding error when calculating b interferes with the second parameter of round.
You may want to try this:
your_num = 5.55
n = len(str(int(your_num)))
float('0' + str(your_num)[n:])
It will return 0.55.
number=5.55
decimal=(number-int(number))
decimal_1=round(decimal,2)
print(decimal)
print(decimal_1)
output: 0.55
See what I often do to obtain numbers after the decimal point in python
3:
a=1.22
dec=str(a).split('.')
dec= int(dec[1])
If you are using pandas:
df['decimals'] = df['original_number'].mod(1)

Categories

Resources