df.query bug giving RuntimeWarning: divide by zero encountered in log10 in align.py", line 98, in _align_core ordm = np.log10(abs(reindexer_size - term_axis_size))

[aliabbasjp]

Getting the following bug:

RuntimeWarning: divide by zero encountered in log10
C:\Users\Nick\Anaconda3\lib\site-packages\pandas\computation\align.py:98:
ordm = np.log10(abs(reindexer_size - term_axis_size))
while running query().

dtypes:

CGI int64
SITEID int32
LONGITUDE float64
LATITUDE float32
dtype: object
Query:

towerData = towerData.query('CGI != 0')

once you get this runtime warning it just halts everything and you have to restart the program.

This bug happens only in multithreaded calls to query function. I am using the concurrent.futures module for the same.

runs fine for non threaded application

[sinhrks]

Thanks for the report. Do you have minimal copy-pastable example?

[rkern]

Note: I'll be avoiding that warning explicitly in my fix for #13109. The numpy error state is thread-local IIRC, so the current fire-and-forget np.seterr(all='ignore') probably only takes effect in the main thread. Why the OP's program halts after the warning, I don't know.

[grove80904]

I'm having a similar issue even without threading. See example code below.

`
import pandas as pd
import numpy as np
df = pd.DataFrame.from_items([('A', [1, 2, 3, 4, 5]),
('B', [14, 15, 16, 17, 18]),
('C', list(np.random.randn(5)))])
df.set_index(['A', 'B'], inplace = True)
df.query('15 <= B <= 17')

C:\Anaconda3\lib\site-packages\pandas\computation\align.py:98:
RuntimeWarning: divide by zero encountered in log10
ordm = np.log10(abs(reindexer_size - term_axis_size))
Out[1]:
C
A B
2 15 -0.852411
3 16 -0.665470
4 17 0.132162
`

Windows 7 64 bit
Spyder 3.0.0.dev0
Python 3.5.2
Pandas 0.18.1
Numpy 1.11.1