diff options
| author | Julian Taylor <juliantaylor108@googlemail.com> | 2013-11-09 02:20:40 -0800 |
|---|---|---|
| committer | Julian Taylor <juliantaylor108@googlemail.com> | 2013-11-09 02:20:40 -0800 |
| commit | 896863aff98a156b2fe0fca85fa9279ae8f15aec (patch) | |
| tree | 5e35be446a50083831d76d55dfa370f41713ec0d /numpy/lib/format.py | |
| parent | ff3c89fd0b0c311db316d3d574382336ad3017f7 (diff) | |
| parent | 83eb3ac0658d5e71d9860f8b1e37f285ec19c692 (diff) | |
| download | numpy-896863aff98a156b2fe0fca85fa9279ae8f15aec.tar.gz | |
Merge pull request #4027 from halfaleague/chunk-read-bug
Chunked reading fails when dtype.itemsize is larger than buffer size.
Diffstat (limited to 'numpy/lib/format.py')
| -rw-r--r-- | numpy/lib/format.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/numpy/lib/format.py b/numpy/lib/format.py index fd5496e96..c411715d8 100644 --- a/numpy/lib/format.py +++ b/numpy/lib/format.py @@ -464,7 +464,7 @@ def read_array(fp): # non-chunked case count < max_read_count, so only one read is # performed. - max_read_count = BUFFER_SIZE // dtype.itemsize + max_read_count = BUFFER_SIZE // min(BUFFER_SIZE, dtype.itemsize) array = numpy.empty(count, dtype=dtype) |