Ensuring outputs between two workflows exactly match.

[dbmurray]

Hi all

I’m currently engaged in the task of optimising some older but very complex Alteryx workflow, and I’m trying to create a standard methodology for ensuring that my newer, optimised work matches exactly the output of the original workflow.

I’m unsure if there is a standard methodology recommended by Alteryx experts for such a task.

I’d previously simply outputted a CSV file of all the data from my new optimised workflows and simply joined each column to the data in the older workflows and checking how many matches they are.

If all matched, then I assume that the workflows essentially provide the same output.

However, my intuition is that this approach isn’t robust enough and I thought I’d ask here.

In other platforms (e.g. R) I have often tested the similarities between data frames using anti-joins.

Many thanks in advance for your assistance.

D

[FrederikE · Accepted solution]

Hey @dbmurray,

There is no reason to think that your approach wouldn't work - matching on all columns and checking if L-/R-Outputs exist is totally valid and would be my go-to solution as well. For "robustness" you could pivot your data first on both sides and do the join (On Names and Values), this way you can include additional columns automatically. 

[ArnaldoSandoval · Accepted solution]

Hi @dbmurray 

You may also take a look at the CRew Test group of macros, one of them "Expect Equal" compares two data streams. Your test-verification workflow may look like the one in the screenshot.

The CReW Test Category Group includes the macros shown below:

Hope this helps,

Arnaldo

[dbmurray]

Hi @FrederikE and @ArnaldoSandoval - if I could accept both your solutions as 'right' I would.

Firstly, I love the idea of the pivot to test names and values - it allows quite a bit of flexibility and I'll likely employ this approach.

I also will try the CReW macro as well - I like the idea of some error checking redundancy, because it gives me confidence that a solution has worked.

Thanks folks, this has been very helpful.