How to merge rows in OpenRefine

Question

How to merge rows based on some ID field?

Original Table                   New Table

ID   | Field1 | Field2       ID     | Field1 | Field2
-----|------- |------         


        

Answer 1

It's not OpenRefine but I think it's a really good tool for a OpenRefine user. You could run this Miller (https://github.com/johnkerl/miller) command

mlr --csv reshape -r "Field" -o item,value \
then filter -x -S '$value==""' \
then reshape -s item,value input.csv

to have

ID,Field1,Field2
A,5,10
B,1,3
C,4,150

First I create a tidy version of the data (https://vita.had.co.nz/papers/tidy-data.pdf), and than I transform again it from long to wide format

Answer 2

For "ID" column use "add column based on this column":

filter(
  cell.cross("ProjectName", "ID").cells["Field1"].value,
  v,
  isNonBlank(v)
)[0]

This will set a value for each row identified ID.

Original Table      New Table

ID   | Field1 | Field2 | Field1_ | Field2_
-----|------- |--------|---------|--------
A        5                  5        10
A                10         5        10
B        1                  1        3
B                3          1        3
C        4                  4        150
C                150        4        150

Remove old columns.

After that, remove duplicates by using "blank down + facet by blank + remove matching rows" approach

Answer 3

In the ID column use the menu option: Edit Cells -> Blank down This should leave you with a table looking like:

ID   | Field1 | Field2 
-----|------- |--------
A        5             
                 10    
B        1             
                 3
C        4
                 150

Make sure you are in "Records" mode (this option is at the top left of the data grid). You should see the rows for each ID are grouped together.

Now use Edit Cells -> Join multi-valued cells on each of the other columns - this should leave you with a single row per record once you have done this for all columns

Answer 4

With the cross() function, you can also achieve this result, but much faster.

It's basically a VLOOKUP on itself (all the rows of the same record) — skipping rows with data — that is after filtered to remove empty cells.

filter(cross(cells.ID.value, "TEST-FillUp", "ID"), vV, vV.cells["Field1"].value != null)[0].cells.Field1.value

Answer 5

I think a simpler solution would be to use:

1° The feature "Edit Cells / Blank Down" on your ID column, in order to get something like this:

2° Then "Edit Cells / Join Multivalued cells" on the last column only (Field2), which will produce this: