[datatable-help] Weird error in package with older data.table version

Simon Zehnder szehnder at uni-bonn.de
Fri Mar 14 16:01:55 CET 2014


Hi Matt,

the ‘older version’ was 1.8.10. With the newer version it bumps and stops. No change there (options(warn = 2)) ….only if I give colClasses it runs through. 

Best

Simon

On 14 Mar 2014, at 15:13, Matt Dowle <mdowle at mdowle.plus.com> wrote:

> What do you mean by older data.table version, which one?  Can you upgrade to v1.9.2?
> 
> > Therefore this column gets bumped and the program stops.
> 
> It shouldn't stop.  It should bump the column and continue. That's what happens for me.  Could be a bug then,  which is why it's confusing talking about an older version of data.table. Thinking about it, maybe those bump warning messages should be downgraded to verbose=TRUE output.    It might be stopping if you've set options(warn=2).
> 
> Matt
> 
> On 14/03/14 12:01, Simon Zehnder wrote:
>> My idea below made everything work … problem solved.
>> 
>> Best
>> 
>> Simon
>> 
>> On 13 Mar 2014, at 09:18, Simon Zehnder <szehnder at uni-bonn.de> wrote:
>> 
>>> Ok, what I found out so far is the following:
>>> 
>>> Column 9 (containing characters in the .csv-file) is read first as LGL (logical I think) because the character in the first rows of this column is just ’T’ (and ‘fread' reads T/True/TRUE as TRUE). After some lines there comes a ‘C’ and now this column cannot be anymore logical (LGL) but has to be character. Therefore this column gets bumped and the program stops.
>>> 
>>> As the ordering of columns can change in my package I need to tell ‘fread', that it should not consider LGL at all - is that possible? I would like to avoid to bother the user by asking him to provide colClasses.
>>> 
>>> My data sample is always the TRACE data but I cannot know what variables of the TRACE data a user has retrieved. So my only idea to avoid the above mentioned error in my fread would be:
>>> 
>>> 1. Read column names via ’scan’.
>>> 
>>> 2. Check what variables are in and then choose via key/value pairs the appropriate colClasses and use them in ‘fread’.
>>> 
>>> Any other suggestions?
>>> 
>>> 
>>> Best
>>> 
>>> Simon
>>> 
>>> On 12 Mar 2014, at 20:28, Simon Zehnder <szehnder at uni-bonn.de> wrote:
>>> 
>>>> I am having a weird error in a package I wrote some time ago with an older data.table version. ‘fread’ gives:
>>>> 
>>>> Internal error: attempt to bump from type 0 to type 1. Please report to datatable-help
>>>> 
>>>> The data is the same that I read in before. Any ideas?
>>>> 
>>>> 
>>>> Best
>>>> 
>>>> Simon
>>>> 
>>>> _______________________________________________
>>>> datatable-help mailing list
>>>> datatable-help at lists.r-forge.r-project.org
>>>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
>>> _______________________________________________
>>> datatable-help mailing list
>>> datatable-help at lists.r-forge.r-project.org
>>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
>> _______________________________________________
>> datatable-help mailing list
>> datatable-help at lists.r-forge.r-project.org
>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
>> 
> 



More information about the datatable-help mailing list