A small, complete example of the issue
from io import StringIO
from pandas import Timedelta, merge_asof, read_csv, to_datetime
file1 = '''DATE_TIME,Label
2015-03-02 08:36:51,0
2015-03-02 08:36:54,1'''
df1 = read_csv(StringIO(file1))
df1.DATE_TIME = to_datetime(df1.DATE_TIME)
print(df1)
file2 = '''DATE_TIME,OtherData
2015-03-02 08:36:50,0
2015-03-02 08:36:51,1
2015-03-02 08:36:52,2
2015-03-02 08:36:53,3
2015-03-02 08:36:54,4
2015-03-02 08:36:55,5
2015-03-02 08:36:56,6
2015-03-02 08:36:57,7
2015-03-02 08:36:58,8
2015-03-02 08:36:59,9
2015-03-02 08:37:00,10'''
df2 = read_csv(StringIO(file2))
df2.DATE_TIME = to_datetime(df2.DATE_TIME)
print(df2)
print(merge_asof(df2, df1, on='DATE_TIME'))
print(merge_asof(df2, df1, on='DATE_TIME', tolerance=Timedelta(seconds=1)))
print(merge_asof(df2, df1, on='DATE_TIME', tolerance=Timedelta(seconds=5)))
Current Output
DATE_TIME Label
0 2015-03-02 08:36:51 0
1 2015-03-02 08:36:54 1
DATE_TIME OtherData
0 2015-03-02 08:36:50 0
1 2015-03-02 08:36:51 1
2 2015-03-02 08:36:52 2
3 2015-03-02 08:36:53 3
4 2015-03-02 08:36:54 4
5 2015-03-02 08:36:55 5
6 2015-03-02 08:36:56 6
7 2015-03-02 08:36:57 7
8 2015-03-02 08:36:58 8
9 2015-03-02 08:36:59 9
10 2015-03-02 08:37:00 10
DATE_TIME OtherData Label
0 2015-03-02 08:36:50 0 NaN
1 2015-03-02 08:36:51 1 0.0
2 2015-03-02 08:36:52 2 0.0
3 2015-03-02 08:36:53 3 0.0
4 2015-03-02 08:36:54 4 1.0
5 2015-03-02 08:36:55 5 1.0
6 2015-03-02 08:36:56 6 1.0
7 2015-03-02 08:36:57 7 1.0
8 2015-03-02 08:36:58 8 1.0
9 2015-03-02 08:36:59 9 1.0
10 2015-03-02 08:37:00 10 1.0
DATE_TIME OtherData Label
0 2015-03-02 08:36:50 0 NaN
1 2015-03-02 08:36:51 1 0.0
2 2015-03-02 08:36:52 2 0.0
3 2015-03-02 08:36:53 3 NaN
4 2015-03-02 08:36:54 4 1.0
5 2015-03-02 08:36:55 5 1.0
6 2015-03-02 08:36:56 6 NaN
7 2015-03-02 08:36:57 7 NaN
8 2015-03-02 08:36:58 8 NaN
9 2015-03-02 08:36:59 9 NaN
10 2015-03-02 08:37:00 10 NaN
DATE_TIME OtherData Label
0 2015-03-02 08:36:50 0 NaN
1 2015-03-02 08:36:51 1 0.0
2 2015-03-02 08:36:52 2 0.0
3 2015-03-02 08:36:53 3 0.0
4 2015-03-02 08:36:54 4 1.0
5 2015-03-02 08:36:55 5 1.0
6 2015-03-02 08:36:56 6 1.0
7 2015-03-02 08:36:57 7 1.0
8 2015-03-02 08:36:58 8 1.0
9 2015-03-02 08:36:59 9 1.0
10 2015-03-02 08:37:00 10 NaN
Expected Output
DATE_TIME Label
0 2015-03-02 08:36:51 0
1 2015-03-02 08:36:54 1
DATE_TIME OtherData
0 2015-03-02 08:36:50 0
1 2015-03-02 08:36:51 1
2 2015-03-02 08:36:52 2
3 2015-03-02 08:36:53 3
4 2015-03-02 08:36:54 4
5 2015-03-02 08:36:55 5
6 2015-03-02 08:36:56 6
7 2015-03-02 08:36:57 7
8 2015-03-02 08:36:58 8
9 2015-03-02 08:36:59 9
10 2015-03-02 08:37:00 10
DATE_TIME OtherData Label
0 2015-03-02 08:36:50 0 NaN
1 2015-03-02 08:36:51 1 0.0
2 2015-03-02 08:36:52 2 0.0
3 2015-03-02 08:36:53 3 0.0
4 2015-03-02 08:36:54 4 1.0
5 2015-03-02 08:36:55 5 1.0
6 2015-03-02 08:36:56 6 1.0
7 2015-03-02 08:36:57 7 1.0
8 2015-03-02 08:36:58 8 1.0
9 2015-03-02 08:36:59 9 1.0
10 2015-03-02 08:37:00 10 1.0
DATE_TIME OtherData Label
0 2015-03-02 08:36:50 0 **0.0**
1 2015-03-02 08:36:51 1 0.0
2 2015-03-02 08:36:52 2 0.0
3 2015-03-02 08:36:53 3 NaN
4 2015-03-02 08:36:54 4 1.0
5 2015-03-02 08:36:55 5 1.0
6 2015-03-02 08:36:56 6 NaN
7 2015-03-02 08:36:57 7 NaN
8 2015-03-02 08:36:58 8 NaN
9 2015-03-02 08:36:59 9 NaN
10 2015-03-02 08:37:00 10 NaN
DATE_TIME OtherData Label
0 2015-03-02 08:36:50 0 **0.0**
1 2015-03-02 08:36:51 1 0.0
2 2015-03-02 08:36:52 2 0.0
3 2015-03-02 08:36:53 3 0.0
4 2015-03-02 08:36:54 4 1.0
5 2015-03-02 08:36:55 5 1.0
6 2015-03-02 08:36:56 6 1.0
7 2015-03-02 08:36:57 7 1.0
8 2015-03-02 08:36:58 8 1.0
9 2015-03-02 08:36:59 9 1.0
10 2015-03-02 08:37:00 10 NaN
Output of pd.show_versions()
# Paste the output here pd.show_versions() here
INSTALLED VERSIONS
------------------
commit: None
python: 3.5.1.final.0
python-bits: 64
OS: Darwin
OS-release: 16.1.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None
pandas: 0.19.1
nose: None
pip: 9.0.0
setuptools: 28.0.0
Cython: None
numpy: 1.11.2
scipy: 0.18.1
statsmodels: None
xarray: None
IPython: 4.0.1
sphinx: None
patsy: None
dateutil: 2.5.3
pytz: 2016.6.1
blosc: None
bottleneck: None
tables: 3.3.0
numexpr: 2.6.1
matplotlib: 1.5.0
openpyxl: 2.3.5
xlrd: 1.0.0
xlwt: 1.1.2
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
httplib2: 0.9.2
apiclient: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.8
boto: None
pandas_datareader: None
None
Comment From: jreback
this by definition is only backwards looking http://pandas.pydata.org/pandas-docs/stable/generated/pandas.merge_asof.html?highlight=merge_asof#pandas.merge_asof
if you need both directions you can do a reindex with tolerance
is that not clear from the docs? this is an ordered time operation
Comment From: jreback
see docs for .reindex here
Comment From: chrisaycock
@VelizarVESSELINOV I just saw this old issue. A "direction" parameter has just been added. The code is in the master branch and will be released as part of pandas 0.20.0.
Comment From: VelizarVESSELINOV
@chrisaycock Thank you 👍