Incorrect result, regardless of precision settings. emax didn't help either

[ghost]

I get a result i don't expect when i add a small number to larger number. In the decimal system it is obvious that the result is mathematically wrong.
I tried a higher precision, and this didn't work. After reading issue #112 i tried to set emin but this didn't change anything.

Any idea how to get the number i expected?

#!/usr/bin/env python3
from bigfloat import *

precision(300)

#Doesn't change anything if this line is here or not
Context(precision=300, emin=-9999999)


#########################
# add 21, works somewhat#
#########################
r = BigFloat(1)
a = BigFloat(10**33)
r = r + 21*(1/a)
print(r)
print(int(r*10**40))

########################
# add 22, doesn't work #
########################
r = BigFloat(1)
a = BigFloat(10**35)
r = r + 22*(1/a)
print(r)
print(int(r*10**40))

What i expected:

1.00000000000000000000000000000002100
10000000000000000000000000000000210000000
1.00000000000000000000000000000000022
10000000000000000000000000000000002200000

What i got:

1.00000000000000000000000000000002099
10000000000000000000000000000000209715200
1.00000000000000000000000000000000019
10000000000000000000000000000000002097152

Other things i tried:

• Setting emax=999999
• Using import bigfloat instead of from bigfloat import * and then using r = bigfloat.BigFloat("1",bigfloat.precision(4000))

No success so far.

Version:

>>> import bigfloat
>>> bigfloat.__version__
'0.4.0'

[mdickinson]

Thanks for the report. precision is a context manager, which needs to be used with a with statement. Here's an example:

>>> from bigfloat import *
>>> r = BigFloat(1)
>>> a = BigFloat(10**35)
>>> with precision(300):
...     r = r + 22 * (1 / a)
...     print(r)
...     print(int(r*10**40))
... 
1.0000000000000000000000000000000002200000000000000000000000000000000000000000000000000000000
10000000000000000000000000000000002200000

See this page of the tutorial for more.