Timestamp.floor() rounds up when a few nanoseconds before a round time

[frexvahi]

Code Sample, a copy-pastable example if possible

>>> pd.Timestamp('2001-02-03 04:05:06.99999999').floor('1s')
Timestamp('2001-02-03 04:05:07')

>>> pd.Timestamp('2004-05-06 07:08:59.9999999').floor('1min')
Timestamp('2004-05-06 07:09:00')

Problem description

For times very close to but below a round second, pd.Timestamp.floor() will round up instead of down. The threshold for where this happens varies from second to second but seems to be between 50 and 200 nanoseconds (i.e. .9999998 will floor() correctly, .99999995 will floor() to the next second upwards).

Expected Output

Timestamp('2001-02-03 04:05:06')
Timestamp('2004-05-06 07:08:00')

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.7.final.0
python-bits: 64
OS: Linux
OS-release: 4.15.0-42-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: en_GB.UTF-8

pandas: 0.23.4
pytest: 4.0.1
pip: 18.1
setuptools: 40.6.2
Cython: 0.29
numpy: 1.15.4
scipy: 1.0.0
pyarrow: None
xarray: 0.11.0
IPython: 7.2.0
sphinx: 1.8.2
patsy: 0.5.0
dateutil: 2.7.5
pytz: 2018.7
blosc: None
bottleneck: 1.2.1
tables: 3.4.4
numexpr: 2.6.4
feather: None
matplotlib: 2.1.2
openpyxl: 2.5.11
xlrd: 1.1.0
xlwt: 1.3.0
xlsxwriter: 1.1.2
lxml: 4.2.5
bs4: 4.6.3
html5lib: 1.0.1
sqlalchemy: 1.2.14
pymysql: None
psycopg2: 2.6.2 (dt dec pq3 ext lo64)
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[chris-b1]

Thanks for the report - this looks to be fixed in master - probably should add some confirming tests

In [2]: ts = pd.Timestamp('2001-02-03 04:05:06.99999999')

In [3]: ts.floor('1s')
Out[3]: Timestamp('2001-02-03 04:05:06')

In [4]: pd.Timestamp('2004-05-06 07:08:59.9999999').floor('1min')
Out[4]: Timestamp('2004-05-06 07:08:00')

[chris-b1]

Fixed by #22802 (test coverage is good), duplicate of #22591