New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Boolean comparison vs tuple fails in 0.17.0 #11339

Closed
svidela opened this Issue Oct 15, 2015 · 6 comments

Comments

Projects
None yet
4 participants
@svidela

svidela commented Oct 15, 2015

Just updated to 0.17.0 (python 2.7.9 on OS X Yosemite) and I found the following:

>>> import pandas as pd
>>> pd.__version__
u'0.17.0'
>>> s = pd.Series([(1,1),(1,2)])
>>> s==(1,2)
0    False
1    False
dtype: bool

Meanwhile in 0.16.2:

>>> import pandas as pd
>>> pd.__version__
'0.16.2'
>>> s = pd.Series([(1,1),(1,2)])
>>> s==(1,2)
0    False
1     True
dtype: bool
@shoyer

This comment has been minimized.

Show comment
Hide comment
@shoyer

shoyer Oct 15, 2015

Member

Can you check that you're using the same version of numpy? I think this could actually be an issue with numpy 1.10 adjusting its casting rules...

Member

shoyer commented Oct 15, 2015

Can you check that you're using the same version of numpy? I think this could actually be an issue with numpy 1.10 adjusting its casting rules...

@svidela

This comment has been minimized.

Show comment
Hide comment
@svidela

svidela Oct 16, 2015

Hi,

In both cases I have latest numpy (1.10.1). I don't know if it's related or not but I actually got to this issue because I have a Series where values are frozensets. In that case not only fails but there is also an exception. Maybe the traceback helps you to know if this is numpy- or just pandas-related

>>> s = pd.Series([frozenset([1]), frozenset([1,2])])
>>> s
0       (1)
1    (1, 2)
dtype: object
>>> s == frozenset([1,2])
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-40-2d7a371a1a3d> in <module>()
----> 1 s == frozenset([1,2])

/Users/sancho/.virtualenvs/test/lib/python2.7/site-packages/pandas/core/ops.pyc in wrapper(self, other, axis)
    724                 other = np.asarray(other)
    725
--> 726             res = na_op(values, other)
    727             if isscalar(res):
    728                 raise TypeError('Could not compare %s type with Series'

/Users/sancho/.virtualenvs/test/lib/python2.7/site-packages/pandas/core/ops.pyc in na_op(x, y)
    644                     result = lib.vec_compare(x, y.astype(np.object_), op)
    645                 else:
--> 646                     result = lib.vec_compare(x, y, op)
    647             else:
    648                 result = lib.scalar_compare(x, y, op)

pandas/lib.pyx in pandas.lib.vec_compare (pandas/lib.c:14283)()

ValueError: Buffer has wrong number of dimensions (expected 1, got 0)

svidela commented Oct 16, 2015

Hi,

In both cases I have latest numpy (1.10.1). I don't know if it's related or not but I actually got to this issue because I have a Series where values are frozensets. In that case not only fails but there is also an exception. Maybe the traceback helps you to know if this is numpy- or just pandas-related

>>> s = pd.Series([frozenset([1]), frozenset([1,2])])
>>> s
0       (1)
1    (1, 2)
dtype: object
>>> s == frozenset([1,2])
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-40-2d7a371a1a3d> in <module>()
----> 1 s == frozenset([1,2])

/Users/sancho/.virtualenvs/test/lib/python2.7/site-packages/pandas/core/ops.pyc in wrapper(self, other, axis)
    724                 other = np.asarray(other)
    725
--> 726             res = na_op(values, other)
    727             if isscalar(res):
    728                 raise TypeError('Could not compare %s type with Series'

/Users/sancho/.virtualenvs/test/lib/python2.7/site-packages/pandas/core/ops.pyc in na_op(x, y)
    644                     result = lib.vec_compare(x, y.astype(np.object_), op)
    645                 else:
--> 646                     result = lib.vec_compare(x, y, op)
    647             else:
    648                 result = lib.scalar_compare(x, y, op)

pandas/lib.pyx in pandas.lib.vec_compare (pandas/lib.c:14283)()

ValueError: Buffer has wrong number of dimensions (expected 1, got 0)
@jreback

This comment has been minimized.

Show comment
Hide comment
@jreback

jreback Oct 16, 2015

Contributor

this was a bug introduced in jreback@4fe7c68 I think. fixed in #11339

Contributor

jreback commented Oct 16, 2015

this was a bug introduced in jreback@4fe7c68 I think. fixed in #11339

@svidela

This comment has been minimized.

Show comment
Hide comment
@svidela

svidela Oct 16, 2015

Should I create another issue for the exception when comparing versus a frozenset or it's already included in the same fix?

svidela commented Oct 16, 2015

Should I create another issue for the exception when comparing versus a frozenset or it's already included in the same fix?

jreback added a commit to jreback/pandas that referenced this issue Oct 16, 2015

@jreback

This comment has been minimized.

Show comment
Hide comment
@jreback

jreback Oct 16, 2015

Contributor

fixed up.. note that using object dtype things inside a Series is very inefficient.

Contributor

jreback commented Oct 16, 2015

fixed up.. note that using object dtype things inside a Series is very inefficient.

jreback added a commit that referenced this issue Oct 16, 2015

Merge pull request #11345 from jreback/boolean
BUG: bug in comparisons vs tuples, #11339

Dr-Irv added a commit to Dr-Irv/pandas that referenced this issue Oct 24, 2015

Fix for issue #11317
This includes updates to 3 Excel files, plus a test in test_excel.py,
plus the fix in parsers.py

issue when read_html with previous fix

With read_html, the fix didn't work on Python 2.7.  Handle the string
conversion correctly

Add bug fixed to what's new

Revert "Add bug fixed to what's new"

This reverts commit 05b2344.

Revert "issue when read_html with previous fix"

This reverts commit d1bc296.

Add what's new to describe bug.  fix issue with original fix

Added text to describe the bug.
Fixed issue so that it works correctly in Python 2.7

Add round trip test

Added round trip test and fixed error in writing sheets when
merge_cells=false and columns have multi index

DEPR: deprecate pandas.io.ga, pandas-dev#11308

DEPR: deprecate engine keyword from to_csv pandas-dev#11274

remove warnings from the tests for deprecation of engine in to_csv

PERF: Checking monotonic-ness before sorting on an index pandas-dev#11080

BUG: Bug in list-like indexing with a mixed-integer Index, pandas-dev#11320

Add hex color strings test

CLN: GH11271 move _get_handle, UTF encoders to io.common

TST: tests for list skiprows in read_excel

BUG: Fix to_dict() problem when using only datetime pandas-dev#11247

Fix a bug where to_dict() does not return Timestamp when there is only
datetime dtype present.

Undo change for when columns are multiindex

There is still something wrong here in the format of the file when there
are multiindex columns, but that's for another day

Fix formatting in test_excel and remove spurious test

See title

BUG: bug in comparisons vs tuples, pandas-dev#11339

bug#10442 : fix, adding note and test

BUG pandas-dev#10442(test) : Convert datetimelike index to strings with astype(str)

BUG#10422: note added

bug#10442 : tests added

bug#10442 : note udated

BUG pandas-dev#10442(test) : Convert datetimelike index to strings with astype(str)

bug#10442: fix, adding note and test

bug#10442: fix, adding note and test

Adjust test so that merge_cells=False works correctly

Adjust the test so that if merge_cells=false, it does a proper
formatting of the columns in the single row header, and puts the row
header in the first row

Fix test for Python 2.7 and 3.5

The test is failing on Python 2.7 and 3.5, which appears to read in the
values as floats, and I cannot replicate.  So force the tests to pass by
just making the column names equal when merge_cells=False

Fix for openpyxl < 2, and for issue pandas-dev#11408

If using openpyxl < 2, and value is a string that could be a number,
force a string to be written out.  If using openpyxl >= 2.2, then fix
issue pandas-dev#11408 to do with merging cells

Use set_value_explicit instead of set_explicit_value

set_value_explicit is in openpyxl 1.6, changed in openpyxl 1.8, but
there is code in 1.8 to set set_value_explicit to set_explicit_value for
compatibility

Add line in whatsnew for issue 11408

ENH: added capability to handle Path/LocalPath objects, pandas-dev#11033

DOC: typo in whatsnew/0.17.1.txt

PERF: Release GIL on some datetime ops

BUG: Bug in DataFrame.replace with a datetime64[ns, tz] and a non-compat to_replace pandas-dev#11326

CLN: clean up internal impl of fillna/replace, xref pandas-dev#11153

PERF: fast inf checking in to_excel

PERF: Series.dropna with non-nan dtypes

fixed pathlib tests on windows

DEPR: remove some SparsePanel deprecation warnings in testing

DEPR: avoid numpy comparison to None warnings

API: indexing with a null key will raise a TypeError rather than a ValueError, pandas-dev#11356

WARN: elementwise comparisons with index names, xref pandas-dev#11162

DEPR warning in io/data.py w.r.t. order->sort_values

WARN: more elementwise comparisons to object

WARN: more uncomparables of numeric array vs object

BUG: quick fix for pandas-dev#10989

TST: add test case from Issue pandas-dev#10989

API: add _to_safe_for_reshape to allow safe insert/append with embedded CategoricalIndexes

Signed-off-by: Jeff Reback <jeff@reback.net>

BLD: conda

Revert "BLD: conda"

This reverts commit 0c8a8e1.

TST: remove invalid symbol warnings

TST: move some tests to slow

TST: fix some warnings filters

TST: import pandas_datareader, use for tests

TST: remove some deprecation warnings from imports

DEPR: fix VisibleDeprecationWarnings in sparse

TST: remove some warnings in test_nanops

ENH: Improve the error message in to_gbq when the DataFrame schema does not match pandas-dev#11359

add libgfortran to 1.8.1 build

binstar -> anaconda

remove link to issue 11328 in whatsnew

Fixes to document issue in code, small efficiency fix

Try to resolve rebase conflict in whats new
@sinhrks

This comment has been minimized.

Show comment
Hide comment
@sinhrks

sinhrks Jul 16, 2016

Member

The behavior may be changed after #13637. Pls comment if any thoughts.

Member

sinhrks commented Jul 16, 2016

The behavior may be changed after #13637. Pls comment if any thoughts.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment