Bug with 'Document Vector': Execute failed: Error while writing to buffer ...

Hi,

I'm using KNIME for some time - and I really love it.But now I encountered a problem with the 'Document Vector' node in workflows, which performed well in the past.

I'm using version 2.11.0 on Windows 7 and created a demo WF consisting of 4 nodes: 'Table Creator -> Strings To Document -> 'Keygraph keyword extractor' -> 'Document Vector'. In 'Table Craetor' only one string ('AAA BBB CCC') was placed in 'column1'. All settings/configurations of the nodes were left unchanged (defaults).

Running this sample (after creation or restart of KNIME) works fine, but resetting 'Document Vecor' (or another node) and executing again results in : ERROR     Document vector                    Execute failed: Error while writing to buffer, failed to write to file "knime_container_20150111_6183533940375088623.bin.gz": Failed copying file stores to local handler.

Looking into the KNIME.LOG I find:  ..... Caused by: java.io.FileNotFoundException: C:\Users\erich\AppData\Local\Temp\knime_Workflow Manage91561\knime_fs-Document_vector_2_2-91564\000\000\7873b690-2b22-436b-a66b-f1f9f4b3297f (Das System kann den angegebenen Pfad nicht finden) .....

I can re-execute the sample either by (i) restarting knime or by (ii) creating a new 'Document Vector' node.

So it seems to my, that the problem is related to a multiple execution of a 'Document Vector' node instance - BUT ONLY in version 2.11.0 of KNIME. The WF above works fine with 2.10.4!

Any idea what is causing this problem?

Erich

 

 

 

Hi Erich,

thank you for reporting this problem. I have experienced the same problem with 2.11 and we are on it to find and fix the problem.

As workaround you can create document vectors by using the Pivoting node on a bow. First convert terms to strings. Then use the Pivoting node with documents as groups, terms as pivots and e.g. a number / constant as value. Use the missing value node right after to replace missing values with 0's.

Cheers, Kilian

Hi Kilian,

thank you for your quick answer and the proposed work-around.

Best

Erich

Hi Erich,

we found the problem and fixed it for version 2.11.1 which will be released end of this week or beginning of next week.

Cheers, Kilian