double_const.py
来自「mallet是自然语言处理、机器学习领域的一个开源项目。」· Python 代码 · 共 31 行
PY
31 行
from test_support import TestFailed# A test for SF bug 422177: manifest float constants varied way too much in# precision depending on whether Python was loading a module for the first# time, or reloading it from a precompiled .pyc. The "expected" failure# mode is that when test_import imports this after all .pyc files have been# erased, it passes, but when test_import imports this from# double_const.pyc, it fails. This indicates a woeful loss of precision in# the marshal format for doubles. It's also possible that repr() doesn't# produce enough digits to get reasonable precision for this box.PI = 3.14159265358979324TWOPI = 6.28318530717958648PI_str = "3.14159265358979324"TWOPI_str = "6.28318530717958648"# Verify that the double x is within a few bits of eval(x_str).def check_ok(x, x_str): assert x > 0.0 x2 = eval(x_str) assert x2 > 0.0 diff = abs(x - x2) # If diff is no larger than 3 ULP (wrt x2), then diff/8 is no larger # than 0.375 ULP, so adding diff/8 to x2 should have no effect. if x2 + (diff / 8.) != x2: raise TestFailed("Manifest const %s lost too much precision " % x_str)check_ok(PI, PI_str)check_ok(TWOPI, TWOPI_str)
⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?