Found this while working on #12136.
Rounding errors occur for timedeltas initialized using fractional microsecond values:
>>> pd.Timedelta('12.312us') == pd.Timedelta(12312)
False
>>> pd.Timedelta(12.312, unit='us') == pd.Timedelta(12312)
False
>>> pd.Timedelta('12.312us') == pd.Timedelta(12311)
True
Additionally, precision is lost when initializing timedeltas with fractional millisecond (and probably second) values:
>>> pd.Timedelta(0.012312, unit='ms') == pd.Timedelta(12311)
False
>>> pd.Timedelta(0.012311, unit='ms') == pd.Timedelta(12311)
False
>>> pd.Timedelta(0.012311, unit='ms') == pd.Timedelta(12000)
True
I believe the problem is here: https://github.com/pydata/pandas/blob/288059adfb751d52be57e444198198518d256c05/pandas/tslib.pyx#L3399
bumping up the values of p should probably solve this.
output of pd.show_versions()
commit: None python: 2.7.11.final.0 python-bits: 64 OS: Windows OS-release: 10 machine: AMD64 processor: Intel64 Family 6 Model 78 Stepping 3, GenuineIntel byteorder: little LC_ALL: None LANG: None
pandas: 0+unknown nose: 1.3.7 pip: 8.1.0 setuptools: 20.2.2 Cython: 0.23.4 numpy: 1.10.4 scipy: 0.17.0 statsmodels: None xarray: None IPython: 4.1.2 sphinx: 1.3.5 patsy: 0.4.0 dateutil: 2.4.2 pytz: 2015.7 blosc: None bottleneck: 1.0.0 tables: 3.2.2 numexpr: 2.4.6 matplotlib: 1.5.1 openpyxl: 2.3.2 xlrd: 0.9.4 xlwt: 1.0.0 xlsxwriter: 0.8.4 lxml: 3.5.0 bs4: 4.3.2 html5lib: 0.999 httplib2: None apiclient: None sqlalchemy: 1.0.12 pymysql: 0.6.7.None psycopg2: None jinja2: 2.8 boto: None
Comment From: jreback
you can try that, though might break other things.
Comment From: jorisvandenbossche
Closing in favor of https://github.com/pandas-dev/pandas/issues/14156