diff options
author | freeman <the.freeman.lab@gmail.com> | 2014-10-22 09:33:12 -0700 |
---|---|---|
committer | Xiangrui Meng <meng@databricks.com> | 2014-10-22 09:33:12 -0700 |
commit | 97cf19f64e924569892e0a0417de19329855f4af (patch) | |
tree | 3252c2f54d296d8e79f97b2094a81cb5cf9cf0be /yarn | |
parent | f05e09b4c95d799bdda3c3ff7fb76a4cd656415d (diff) | |
download | spark-97cf19f64e924569892e0a0417de19329855f4af.tar.gz spark-97cf19f64e924569892e0a0417de19329855f4af.tar.bz2 spark-97cf19f64e924569892e0a0417de19329855f4af.zip |
Fix for sampling error in NumPy v1.9 [SPARK-3995][PYSPARK]
Change maximum value for default seed during RDD sampling so that it is strictly less than 2 ** 32. This prevents a bug in the most recent version of NumPy, which cannot accept random seeds above this bound.
Adds an extra test that uses the default seed (instead of setting it manually, as in the docstrings).
mengxr
Author: freeman <the.freeman.lab@gmail.com>
Closes #2889 from freeman-lab/pyspark-sampling and squashes the following commits:
dc385ef [freeman] Change maximum value for default seed
Diffstat (limited to 'yarn')
0 files changed, 0 insertions, 0 deletions