Hello,
I have a workflow where I parsed the numbers from the 'ID' column to the 'Version_EDIT' column. I am trying to create a formula where I can sort by the highest version in 'Version_EDIT' by unique 'ID' group. Anything less than the highest version, the number '0' should be applied in the 'Deployment' column and in the 'Comments' column should write 'lesser version'.
In the highest version rows, leave the 'Comments' and 'Deployment' columns blank (as is).
Base file(example):
ID | VERSION | Deployment | Comments | Version_EDIT |
5698-A14 | V3 | 3 | ||
5698-A14 | V2 | 2 | ||
5698-A14 | V1 | 1 |
Expected Output(example):
ID | VERSION | Deployment | Comments | Version_EDIT |
5698-A14 | V3 | 3 | ||
5698-A14 | V2 | 0 | lesser version | 2 |
5698-A14 | V1 | 0 | lesser version | 1 |
Please see attached of my working file. Thank you!
Solved! Go to Solution.
Hey @csodesza, how does this look?
I've:
1 - Parsed the version number
2 - Sorted on this
3 - Found the max for each ID and re-appended this
4 - Done a couple of checks to see if the version # = the max for that ID
5 - If not, apply the Comments/Deployment requirements
6 - If so, left them blank
7 - Removed temporary 'Max' field used for logic
@binuacs What about in the scenario when we have the same unique ID with the same version?
For example, I'm receiving this output when implementing with your rules:
ID | Version | Deployment | Comments |
5698-A14 | V3 | ||
5698-A14 | V3 | 0 | Lesser Version |
5698-A14 | V3 | 0 | Lesser Version |
Expected Output:
ID | Version | Deployment | Comments |
5698-A14 | V3 | ||
5698-A14 | V3 | ||
5698-A14 | V3 |
Essentially they should all be null values since it can't recognize a higher version with that unique ID.
Much appreciative of your help as always!
User | Count |
---|---|
17 | |
15 | |
15 | |
8 | |
5 |