problem after update to rm 5.2.2.

Hi,

nope, I did not have any problems. I tried both, reading a data set where the operator has been configured with an older version and using the wizard to import a fresh one. Both processes worked like a charm. Maybe someone else?

(I move this thread into the Problems & Support, probably the developers will work on this quicker then...)

Cheers,
Ingo

Hi greyhound,

can you please post a short example process that shows how to reproduce the error? All our test of the Read CSV operator work fine though..

Best,
Nils

Thank you for the express-answer. I found s.th. new about it:
the error message occurs only when I choose DATETIME
as ID in the operator READ CSV as I used to do till the update.
Now it works only if I leave it in READ CSV as "attribute"
and insert an additional SET ROLE behind. Here  I select DATETIME
as ID  -> the process runs great.
Thus its fine for me, but its a new behavior after the update.

I can reproduce the error. We will have a look at this. Thanks for notifying us!

I received the same error after the most recent update... since all of my processes require reading csv files, I'm looking to rollback to the previous build until the bug is fixed.  Despite the new bug, kudos again to the RapidMiner team for their great product!

After a couple of tests the problem appears to be caused by READ CSV data types function.  As an example, the previous code snippet worked in previous builds:

<list key="data_set_meta_data_information">
<parameter key="0" value="Instrument.true.text.batch"/>
<parameter key="1" value="jDate.true.real.id"/>
<parameter key="2" value="Att01.true.real.attribute"/>
<parameter key="3" value="Att02.true.real.attribute"/>
<parameter key="4" value="Att03.true.real.attribute"/>
<parameter key="5" value="Att04.true.real.attribute"/>
<parameter key="6" value="label.true.binominal.label"/>
</list>

However, the latest build requires changing batch or id data types to attributes before the process will execute, i.e.

<list key="data_set_meta_data_information">
<parameter key="0" value="Instrument.true.text.attribute"/>
<parameter key="1" value="jDate.true.real.attribute"/>
<parameter key="2" value="Att01.true.real.attribute"/>
<parameter key="3" value="Att02.true.real.attribute"/>
<parameter key="4" value="Att03.true.real.attribute"/>
<parameter key="5" value="Att04.true.real.attribute"/>
<parameter key="6" value="label.true.binominal.label"/>
</list>

As was previosuly mentioned, you can use the "SET ROLE" function to convert the data types later, but I previously rewrote all my processes to avoid using the SET ROLE function when I transitioned to v. 5.2.000.  Importing the data types directly, as before, I think was much more efficienct than using the "SET ROLE" function, and it created more compact code.

Since "READ CSV" worked correctly in the previous build, I'm confident the RapidMiner team will quickly recitfy this problem; therefore, I prefer not having to recode all my processes.  In the meantime, if anyone has similar processes that utilize the READ CSV function to import data and set the data types, you may want to delay updating to version 5.2.002.

Hi Darrrell,

we have fixed it in the current SVN version and will release a new version soon that will fix this bug.

Best,
Nils

Looking forward to the update... and as I stated before, thanks again for providing such a great product!