BUG: DatetimeIndex has become unhashable in 1.3.1?

[yohplala]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Hi,
Following code below was working with pandas 1.2.5.
It raises now an error with pandas 1.3.1.

Code Sample, a copy-pastable example

import pandas as pd
import random

# Right data
ts_open = pd.DatetimeIndex([
                           pd.Timestamp('2021/01/01 00:37'),
                           pd.Timestamp('2021/01/01 00:40'),
                           pd.Timestamp('2021/01/01 01:00'),
                           pd.Timestamp('2021/01/01 03:45'),
                           pd.Timestamp('2021/01/01 03:59'),
                           pd.Timestamp('2021/01/01 05:20')])
length = len(ts_open)
random.seed(1)
volume = random.sample(range(1, length+1), length)
df_smpl = pd.DataFrame({'volume' : volume, 'ts_open': ts_open})

# Left data
ts_full = pd.date_range(start=pd.Timestamp('2021/01/01 00:00'),
                        periods=7, freq='1h')
ts_event = ts_full[:-1]
length_ipr = len(ts_event)
random.seed(2)
volume = random.sample(range(20, length_ipr+20), length_ipr)
df_ipr = pd.DataFrame({'volume' : volume, 'timestamp': ts_event})

at = ts_full[1:]
df_smpl.index = df_smpl['ts_open']

# Test
df_full = pd.merge_asof(df_ipr, df_smpl,
                        left_on=at, right_index=True,
                        allow_exact_matches=False, direction='backward')

Error message:

Traceback (most recent call last):

  File "<ipython-input-152-b5845a330a4b>", line 30, in <module>
    df_full = pd.merge_asof(df_ipr, df_smpl,

  File "/home/yoh/anaconda3/lib/python3.8/site-packages/pandas/core/reshape/merge.py", line 581, in merge_asof
    op = _AsOfMerge(

  File "/home/yoh/anaconda3/lib/python3.8/site-packages/pandas/core/reshape/merge.py", line 1736, in __init__
    _OrderedMerge.__init__(

  File "/home/yoh/anaconda3/lib/python3.8/site-packages/pandas/core/reshape/merge.py", line 1619, in __init__
    _MergeOperation.__init__(

  File "/home/yoh/anaconda3/lib/python3.8/site-packages/pandas/core/reshape/merge.py", line 682, in __init__
    self._validate_specification()

  File "/home/yoh/anaconda3/lib/python3.8/site-packages/pandas/core/reshape/merge.py", line 1782, in _validate_specification
    if left_on in self.left.columns

  File "/home/yoh/anaconda3/lib/python3.8/site-packages/pandas/core/indexes/base.py", line 4572, in __contains__
    hash(key)

TypeError: unhashable type: 'DatetimeIndex'

Output of pd.show_versions()

INSTALLED VERSIONS

commit : c7f7443
python : 3.8.8.final.0
python-bits : 64
OS : Linux
OS-release : 5.8.0-63-generic
Version : #71~20.04.1-Ubuntu SMP Thu Jul 15 17:46:08 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.3.1
numpy : 1.20.3
pytz : 2021.1
dateutil : 2.8.2
pip : 21.1.3
setuptools : 52.0.0.post20210125
Cython : 0.29.24
pytest : 6.2.4
hypothesis : None
sphinx : 4.0.2
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.3
IPython : 7.22.0
pandas_datareader: None
bs4 : 4.9.3
bottleneck : None
fsspec : 2021.07.0
fastparquet : 0.7.0
gcsfs : None
matplotlib : 3.3.4
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 3.0.0
pyxlsb : None
s3fs : None
scipy : 1.6.2
sqlalchemy : None
tables : None
tabulate : 0.8.9
xarray : 0.19.0
xlrd : None
xlwt : None
numba : 0.53.1

[simonjayhawkins]

Thanks @yohplala for the report

Following code below was working with pandas 1.2.5.

first bad commit: [cd0224d] BUG: Raise ValueError for non numerical join columns in merge_asof (#34488)

but looks like the traceback has since changed. was..

2021-08-02T13:13:55.0632373Z 1.3.0.dev0+356.gcd0224d26c
2021-08-02T13:13:55.0654274Z Traceback (most recent call last):
2021-08-02T13:13:55.0655519Z   File "../42844.py", line 37, in <module>
2021-08-02T13:13:55.0656179Z     df_full = pd.merge_asof(
2021-08-02T13:13:55.0657055Z   File "/home/runner/work/pandas/pandas/pandas/core/reshape/merge.py", line 557, in merge_asof
2021-08-02T13:13:55.0657893Z     op = _AsOfMerge(
2021-08-02T13:13:55.0658704Z   File "/home/runner/work/pandas/pandas/pandas/core/reshape/merge.py", line 1670, in __init__
2021-08-02T13:13:55.0659566Z     _OrderedMerge.__init__(
2021-08-02T13:13:55.0660833Z   File "/home/runner/work/pandas/pandas/pandas/core/reshape/merge.py", line 1564, in __init__
2021-08-02T13:13:55.0661751Z     _MergeOperation.__init__(
2021-08-02T13:13:55.0662624Z   File "/home/runner/work/pandas/pandas/pandas/core/reshape/merge.py", line 656, in __init__
2021-08-02T13:13:55.0663522Z     self._validate_specification()
2021-08-02T13:13:55.0664509Z   File "/home/runner/work/pandas/pandas/pandas/core/reshape/merge.py", line 1713, in _validate_specification
2021-08-02T13:13:55.0666096Z     self.left[self.left_on[0]].dtype
2021-08-02T13:13:55.0666970Z   File "/home/runner/work/pandas/pandas/pandas/core/frame.py", line 3037, in __getitem__
2021-08-02T13:13:55.0668021Z     indexer = self.loc._get_listlike_indexer(key, axis=1, raise_missing=True)[1]
2021-08-02T13:13:55.0669128Z   File "/home/runner/work/pandas/pandas/pandas/core/indexing.py", line 1269, in _get_listlike_indexer
2021-08-02T13:13:55.0670220Z     self._validate_read_indexer(keyarr, indexer, axis, raise_missing=raise_missing)
2021-08-02T13:13:55.0671554Z   File "/home/runner/work/pandas/pandas/pandas/core/indexing.py", line 1311, in _validate_read_indexer
2021-08-02T13:13:55.0672552Z     raise KeyError(f"None of [{key}] are in the [{axis_name}]")
2021-08-02T13:13:55.0675081Z KeyError: "None of [DatetimeIndex(['2021-01-01 01:00:00', '2021-01-01 02:00:00',\n               '2021-01-01 03:00:00', '2021-01-01 04:00:00',\n               '2021-01-01 05:00:00', '2021-01-01 06:00:00'],\n              dtype='datetime64[ns]', freq='H')] are in the [columns]"

cc @phofl

[phofl]

@yohplala Could you please narrow your example? It should be minimal and reproducible

[phofl]

@simonjayhawkins I don't think that this is supposed to work. Docs say

left_onlabel

    Field name to join on in left DataFrame.

while an Index is provided here.

[yohplala]

@yohplala Could you please narrow your example? It should be minimal and reproducible

Hi, yes, will see to that this evening then.

@simonjayhawkins I don't think that this is supposed to work.

:( (being sad here) It was working with 1.2.5 (the failure happens with test cases of mine that are double-checked: results were correct then)

[yohplala]

@yohplala Could you please narrow your example? It should be minimal and reproducible

Hi Patrick,
please, find enclosed reduced example.

import pandas as pd

# Right data
ts_right = pd.DatetimeIndex([pd.Timestamp('2021/01/01 00:37'),
                             pd.Timestamp('2021/01/01 01:40')])
right = pd.DataFrame({'val' : [2,6], 'ts': ts_right}, index=ts_right)

# Left data
ts_left = pd.date_range(start=pd.Timestamp('2021/01/01 00:00'),
                        periods=3, freq='1h')
left = pd.DataFrame({'val' : [4,8,7], 'timestamp': ts_left})

# Test
merged = pd.merge_asof(left, right, left_on = ts_left, right_index=True,
                       allow_exact_matches=False, direction='backward')

You are right about the root cause (DatetimeIndex beoing provided). Using directly a column of the left DataFrame works ok.

[simonjayhawkins]

changing milestone to 1.3.5

[jbrockmendel]

@phofl looks you you worked on this code recently-ish. Is this a usage we should or do support?

[phofl]

Not sure, I have something in the pipeline to restore but I would be in favor of deprecating for 2.0. we have on tests for passing an array into merge_asof and only two for merge, which were added ages ago