KNIME Analytics Platform 3.2 released

Hi everyone,

we just released KNIME Analytics Platform 3.2 and as always we appreciate any feedback you have. You can always report problems in our forum, just open a new thread for it and we will make sure to have it resolved asap.

Cheers and happy KNIMEing,



Hi Iris-

That is great news - thank you.   A quick question - tried update by using the "update KNIME" option in the File tab but that didn't work.  Is 3.2 a separate download and then I need to import all the existing workflows?

thank you, Tom


Like trichebacher, I cannot update to 3.2 through the 3.1.2 application.

Also, after downloading the 3.2 installer, I find that I am unable to install it because the C:\Program Files\KNIME directory already exists.

Do I really have to uninstall 3.1.2 just to install 3.2?

Hi iris,

is there a step by step tutorial to install tableau SDK? because I had trouble with the existing instructions.

thanks before.

"Cannot update" is not enough information to diagnose this problem. Can you be a bit more specific about what your are doing? Are there any error message? Did you get the dialog that asks you whether the new update sites should be entered? Are the new update sites entered automatically after you agreed?

The best description is on the Tableau page, which is also linked from the node description (

For Linux and Mac (but not for Windows!) the SDK is part of the KNIME update site. We are working with the staff at Tableau to also get the SDK to be part of the KNIME update site. There are some glitches in the SDK that Tableau needs to fix (but they are aware of it).

Where did you get stuck?

I'm using windows server 2008 R2

and I've installed:

- Knime 3.2.0

- Visual C++ Redistributable Packages for Visual Studio 2013

I've downloaded "Tableau-SDK-C-Java-64Bit-9-3-3" and extract it to "<knime folder install>\Tableau-SDK-C-Java-64Bit-9-3-3\"

I was confused at "Setting environment variable" for "TAB_SDK_LOGDIRTAB_SDK_LOGDIR" and "TAB_SDK_TMPDIR (


Btw, what should I install tableau on the same server with knime? because the server that I use for this knime, not installed desktop tableau.


*sorry for my bad english



Great News!

Some comments about my own experience:

 - I had to create a new folder "Knime3.2" for the new version because of the message "KNIME directory already exists".

- Took me sometime to learn and understand how to set up the Tableau Writer but now it is working properly (I'm not using Tableau Server).

- I couldn't find the "Workflow Coach" but now its working: Go to VIEW>OTHER>WORKFLOW COACH and follow the settings.




I can no longer read Longs from SQLite, no matter what column definition or data is involved.


KNIME 3.2 now guesses the datatype of a Database Reader behind SQLite Connections using the table definition, and not the first row, which is an improvement. However, Longs (and any non-real numeric types) are now always parsed to an Integer and thus overflow if they are too large (i.e. Java-Style Timestamps with Milliseconds).

[Expected Behaviour]

I could not find a way to enforce KNIME to read numeric columns as Long instead of Integer.

[Example Workflow]

New Table with long values --> Database Writer (using any type name in SQL Datatypes and having "append" disabled)

SQLite Connector --> Database Reader

Hi hitorijanai,

I was confused at "Setting environment variable" for "TAB_SDK_LOGDIRTAB_SDK_LOGDIR" and "TAB_SDK_TMPDIR (

Those are optional. Key is the PATH variable. Have you found where to set it up? It needs to point to "<knime folder install>\Tableau-SDK-C-Java-64Bit-9-3-3\bin"

I just realized this is badly documented (in both the Tableau web site and KNIME ;-) ): It's point (4) on this page:

Completely confused now?


KNIME doesn't do any auto-guessing for types. We take the SQL type returned by the driver and map it to an KNIME type. If the JDBC driver doesn't tell us that the column is a BIGINT we treat it as an integer (like for any database). It's absolutely possible that the new driver fixed one problem (looking only at the first row for determining types) but has introduced a new one.

I recommend using the new H2 node instead of SQLite as it works much more reliably.

I just checked and SQLite also wasn't able to properly read long values in KNIME 3.1. The driver simply reports the wrong type and there data is truncated when we read it. This isn't a new problem and also one that we cannot fix. I really strongly recommend switching to H2. SQLite simply sucks.

In Knime 3.2.0, creating a Group By with a "unique concatenate" column will convert a single missing value input line to a string "?". Is this a desirable behavior?

@windowsfreak: this is not 3.2 dependent. If you use the Unique Concatenate and activate the missing checkbox, missing will be represented as '?' in your String Output. If you want to skip missings, deactivate the checkbox

Oh and you can just start a new thread if you have a new question, so we see it faster.

Best, iris

Thanks :))