"Bug in ExampleSource configuration"

imoutsatsos

I have identified what I believe to be a buggy behavior of the ExampleSource CSV parser.
When a null/missing value exists at the end of a CSV record it seems to ignore it, and as a result it throws an exception complaining about the number of columns. For example using the following CSV fragment (all rows have 6 columns):

PAB00004039,H,11,720,2,3.952,
PAB00004039,H,12,730,2,3.952,
PAB00004039,H,13,678,2,3.744,
PAB00004039,H,14,,0,,
PAB00004039,H,15,,0,,
PAB00004039,H,16,727,2,3.744,
PAB00004039,H,17,705,2,3.883,
PAB00004039,H,18,712,2.222,4,

I get the error dialog with the message:
'Cannot guess value types: Line 4 has a number of columns(5) different from the preceding lines (6). IOException caught. Line 4 has a number of columns (5) different from the preceding lines (6)'

The log indicates that no data is read passed line 3.

Are there any workarounds or other configuration options for dealing with missing values?

thanks
--ioannis--

IngoRM

Hi Ioannis,

thanks for pointing this out. This behaviour was actually already known and is already fixed in the latest CVS version of RapidMiner. Of course, this fix will also be part of the next release.

Thanks again and cheers,
Ingo

imoutsatsos

That's great. Thanks, looking forward to the new release!
--ioannis--