Closed
Description
Overview
I think I found a bug in texttable's integer conversion logic. Specifically, the intermediate conversion to a floating-point produces an incorrect output for large integers due to floating-point imprecision.
Minimum example:
from texttable import Texttable
tt = Texttable()
tt.set_cols_dtype(['i']) # dtype 'i' is the problem here
tt.add_rows([['hello'], [18014398509481983]])
print(tt.draw())Expected output:
+-------------------+
| hello |
+===================+
| 18014398509481983 |
+-------------------+
Actual output:
+-------------------+
| hello |
+===================+
| 18014398509481984 |
+-------------------+
Explanation of bug
This problem is with texttable's integer (see dtype(['i']) above) conversion code in texttable.py:
@classmethod
def _fmt_int(cls, x, **kw):
"""Integer formatting class-method.
- x will be float-converted and then used.
"""
return str(int(round(cls._to_float(x))))The cls._to_float(x) converts the input to a float, which produces the wrong value for large ints. E.g., float(18014398509481983) = 18014398509481984.
Possible solution idea:
Is the float conversion really necessary? If not, str(int(round(18014398509481983))) = 18014398509481983 produces the right answer.
Environment configuration:
Texttable version: 1.6.2
Python version: 3.8.1
Metadata
Metadata
Assignees
Labels
No labels