ndarray.tofile raises ValueError on IO failure; would IOError be more appropriate? (Trac #1716)

[numpy-gitbot]

Original ticket http://projects.scipy.org/numpy/ticket/1716 on 2011-01-17 by trac user mark.dickinson, assigned to unknown.

The following exception traceback was reported to me by the user of a numpy-based application:

File "C:\EPD622\lib\site-packages\numpy\lib\io.py", line 345, in save
format.write_array(fid, arr)
File "C:\EPD622\lib\site-packages\numpy\lib\format.py", line 401, in write_array
array.tofile(fp)
ValueError: 135453 requested and 6 written

While the cause of the error was never explicitly determined, it seems likely to have been as a result of an overfull disk. The type and message of the exception are a little confusing here; would it make sense for the exception type to be something like IOError rather than ValueError?

This was with the version of numpy that ships with EPD 6.2.2, but I believe that the issue is still present in the current numpy source. The error message appears to be coming from /numpy/core/src/multiarray/convert.c, as a result of a system-level fwrite call failing to write all the expected items (i.e., encountering a system-level IO error).