I am an experienced developer, but a new Knime user and frankly loving the product.
Currently, I am experiencing a holdup in my workflow.
I have a table of Multiple StringCell values that I process using a snippet to produce a result table with similar, but not identical values indexed by the rowID of the row in the original table which was used to generate the given row in the result table.
I would like to compare the values in the Original Table with values in the results table (I don't wish to use Joiner to combine tables), using tools similar to the Lucene query available in the Index Query. Unfortunately, the Index Query is designed for use with a single table which is indexed.
Can someone please advise as to the best means of doing this (without simply resorting to a snippet directly) ?? It seems like it should be trivially possible, but I really can't figure it out thus far.
Thanks for all of your help!
The "Reference Row Filter" node avoids joining, and returns either the identical or the non-identical items (your choice). Maybe that'd work?
Thanks, but that's nowhere near what I want to accomplish.
I need to be able to compare the VALUES in the StringCell(s), using different metrics for each column.
The Lucene syntax supported by 'Index Query' is ideal as it supports Fuzzy string matching and can also return a score column using the Lev Distance, however that Node is intended for processing a single table, not two.
Table Orig Table Result
A | B | C A1 | B1 | C1 | ...
Str | Str | Str Str | Str | Str | ...
Result Table is populated with values which are programatically generated from Table Orig with a 1: Many relationship between the parent row in Original Table and Child rows in Result Table.
I would like to compare values of the Child Rows in the result table for fuzzy matching (and possibly later, other metrics) with the value of their respective Parent Row, each can be considered a record, returning the 'Top K matches...' above whatever score.
I hope this helps a little in understanding my problem, all help is welcome!
Bump, not to be impatient, but this has to be a somewhat common use case, no?