A unified metadata approach

Dear KNIMErs,

The "Extract table spec" and "Extract context properties" are excellent starting points for more stringent metadata handling, as is the forthcoming workflow search functionality demoed at the 2014 UGM.

However, I'm wondering whether we will see even more metadata exploitation mechanishms in the future? Parsing node names, manipulating the (otherwise well-hidden) node descriptions, or dealing with the ever-popular node labels would be really great!

Especially the latter often contain important contextual information, such as the transformation applied in the branch, etc. To give a very tangilbe example, I have transformed data in four different ways, generating four (very different) plots from it. Ideally, I'd like to parse each plot's node label for the transformation descriptor it contains ("sqrt", "cosh", etc.) and add these as constant "label cells columns" before concatenating the plots into one plot/transformation table.

Driving node labels from data would of course also be interesting fro other applications - at-a-glance outcome monitoring without using the node monitor views!

Just my usual .02€... :-)

Cheers
E

*cough* Does this make sense? :)

TY,
E

*cough* Does this make sense? :)

TY,
E

*cough* Does this make sense? :)

TY,
E

Hello, 

Indeed this does make some sense, but will likely not make it onto this year's roadmap.  There may currently ways to do most of what you are looking for, but of course it is a bit "advanced" :)

What about exporting node settings as flow variables and then programatically dealing their outputs to drive workflow execution? Also, have you tried the new constant value column node?  

As for parsing node comments, this can already be done using XPATH along side the list files node.  A bit of pain to set it up the first time, but then you can turn it into a metanode and pull out the data fairly quickly fo reach of your workflows.  

I hope this helps a bit.

Regards,

Aaron

 

 

 

 

Aaron,

Thanks, good pointers in fact... slightly tricky, but you have made me curious to experiment! :-) I may come back with results to publish here or elsewhere -- watch this space.

Cheers
E

Aaron,

Just to clarify, some of what I'd like to to can be done with some "advanced magic", such as extracting existing node labels for re-use. It does require programatically saving the workflow, though, which limits its utility for big flows and critical runtimes.

However, even more so I'd like to be able to set/update the node label text by feeding it with variable contents - kind of an in-workflow node monitor for selected values. Not on the roadmap yet, but feasible in principle, isn't it? :)

Thanks
E