Order of records changed after using Data Cleansing tool

[WuQi]

Hi Alteryx community,

I found that the order of the records changed randomly after cleaning all NULL rows with the Data Cleansing tool. Please see the attached picture for the details. I know the order should remain the same before and after the tool. Is there any setting issue here that causes the problem? Thanks very much!

[BS_THE_ANALYST]

I'm not sure it's "changed randomly". 

If you wanted to identify if a row is null, I assume you'd need to transpose the data, and leverage the summarize tool and do a (group by row: count Null / count rows). If this is the case, when it's cross-tabbed after (if it is), the row order will likely be lost.

You could check this by stepping into the data-cleanse tool by right-clicking and "opening the macro".

Check what's going on under the hood.

All the best,

BS

[BS_THE_ANALYST]

@WuQi  I'm attaching a workflow. I think this logic should work for removing null rows. Tweak it accordingly.

All the best,

BS

[binuacs · Accepted solution]

@WuQi  Turn off the AMP Engine settings. Go to options-user settings - un-check the AMP engine option. Restart the Alteryx designer then run the workflow

[Qiu]

@WuQi 
It works fine on my end.

Maybe you are using the old version?

[WuQi]

Hi @BS_THE_ANALYST , thank you for the suggestion. Really appreciate it! 

[WuQi]

Hi @binuacs , thanks very much for the answer. I have tried and it solved the problem. Can I ask why AMP engine specifically affects the result?

[flying008]

Hi, @WuQi 

AMP is the all .

[Qiu]

@flying008 
Nice one!

[BS_THE_ANALYST · Accepted solution]

@WuQi there would have to be a specific reason for me to turn AMP off, in my opinion. 

It's effectively speeding up the processing of your workflow as it's allows components of the workflow to parallel process. If components are working in parallel, this can cause the records to lose order upon completion. I guess you can imagine that you have your CPU which has X amount of cores on it. Each core will act like a worker and spread the processing across them to get the task done asap. You can't exactly guarantee the order of completion (as it's not sequential) therefore when it gets put back together it may not be the same. (you'll probably get a better explanation out of AI 😂)

I think turning AMP off in this situation is a quick fix. Surely putting a few tools on the canvas to recreate the logic makes more sense and then you can continue to leverage AMP. 

All the best,

BS