You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@arrow.apache.org by "James Campbell (JIRA)" <ji...@apache.org> on 2018/08/28 18:59:00 UTC

[jira] [Created] (ARROW-3137) pyarrow 0.10 requires newer version of numpy than specified in requirements

James Campbell created ARROW-3137:
-------------------------------------

             Summary: pyarrow 0.10 requires newer version of numpy than specified in requirements
                 Key: ARROW-3137
                 URL: https://issues.apache.org/jira/browse/ARROW-3137
             Project: Apache Arrow
          Issue Type: Bug
    Affects Versions: 0.10.0
            Reporter: James Campbell


pyarrow 0.10 appears to have a binary incompatibility with numpy versions prior to the 1.14.x series, but its requirements file claims support for numpy>=1.10.0

If an older version of numpy is used, the following RuntimeError results: {{RuntimeError: module compiled against API version 0xc but this version of numpy is 0xb}}

The following tox.ini file demonstrates the issue:

{{[tox] envlist=py27-numpy\{10,11,13,14,15}-pyarrow\{9,10} [testenv] deps = numpy10: numpy>=1.10.0,<1.11 numpy11: numpy>=1.11.0,<1.12 numpy13: numpy>=1.13.0,<1.14 numpy14: numpy>=1.14.0,<1.15 numpy15: numpy>=1.15.0,<1.16 pyarrow9: pyarrow==0.9.0 pyarrow10: pyarrow==0.10.0 pytest commands = pytest }}

Using a simple test function like the following:

{{def test_import_pyarrow(): import pyarrow }}

pyarrow 0.9 doesn't appear to have this issue. Was there a change in the setup process for pyarrow 0.10 that no longer uses Cython to build?



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)