Save results of each "apply to each" loop into an array for use later on

[ZedYer]

In my Apply to Each, I process a table of data that, for example, has 20 rows of data, with each row having an owner. Out of the 20 rows, I could have 2 owners, with one owning 5 rows and the other owning 15 rows. 

At the end of this very large flow I am sending an adaptive card out to each of the owners so they can work on their portions. I would like to include in this card how many rows of data they have to process, but all this data is lost after this particular Apply to Each ends.

Is there a way to create a 2D array (Owner, RowCount using length()) which will append the new data with each iteration? I am fairly certain I could create a SP list to store this and just append a new row after each iteration, but that sounds too cheesy for an elegant PA flow 🙂   Currently, using a filter array based on each unique owner to divvy up their rows thanks to a solution from this post. 

Thanks!

[abm]

Hi @ZedYer 

From your data first find the owners. This will give you all owners. Then use the UNION() expression to make the owners unique collections. This way you can avoid the duplicate owners from the collection. Next iterate the unique owners collection and filter the dataset against each owner. This way you can get the total number of records for each owner.

Thanks

[ZedYer]

Hi ABM,

I already have all that figured out. It is how to store those results for use at the end of the flow, and much later than this apply to each finishes looping.

[abm]

Hi @ZedYer 

So what format you expect the data? Could you post some examples?

Thanks

[ZedYer]

I just need to store the information in an array "LoopResults".  Iteration 1--  Owner1 and Rows processed.  Iteration 2--Owner2 and rows processed.  As I mentioned, I could easily create a SP list and temporarily use that to store this data, but an array sounds faster and more "appropriate".  I need this data at the very end of the flow to include in an adaptive card.  So, I think the answer is a 2D array, both initializing and appending to it, but the people on here are so creative and genius I didn't want to lock the question specifically on an array as an answer.  Appreciate your considerations.

[abm]

Hi @ZedYer 

Thanks for your reply.  I have more questions. So once the iterations are over you need an array to say Owner 1 - 10 rows processed, Owner 2 - 12 Rows processed. Is that you after? 

Thanks

[ZedYer]

Yes, or owner3, 4, and 5. The number of owners and rows is unknown until we process the data. The issue is that I need it stored someplace, like a variable or something, so in the last step (way past this apply to each in the flow) I can send an adaptive card to each unique owner along with how many rows were processed/counted in the loop.

So, I also have an apply to each owner loop at the end to construct and send each adaptive card. I would use a select to retrieve, based on owner, the number of rows.  Honestly, I've been doing more research on this today and I think I will just use a small SP list to store this data in temporarily, and then get the data I want when I am building the adaptive card.

[Anonymous]

This is Zed, but on my work PC now.  So, I tried my own solution by adding an item for each Owner processed and how many rows were processed into a SP list.  Now, in the last step where i need to send an adaptive card to each owner and how many of their unique rows is processed.  The SP list only has TITLE (owner) and ROW_COUNT and was populated during earlier loop processing. When I send the adaptive card to owner_A, I only want to include how many rows were processed by Owner_A.  There will never be more than one row with particular owner. I first used get item, but turns out the ID is long gone from earlier processing in a prior loop.  So, switched to get items with filter query set at TITLE eq 'OWNER' to retrieve just the one row I need.  However, when I add the ROW_COUNT to the adaptive card it puts it, yet, into another apply to each. Is there a way to force the code to realize there is only ONE row to process so that it doesn't put it into another loop?  My only other course of action is to rethink the entire flow and perform both the batch processing on the entire set of input data inside a loop that only processes data for each owner. I think that will only reverse the problem though because I need to send data on the entire batch of starting data, too.

[abm]

Hi @ZedYer  @Anonymous 

I have used a dictionary object to store the values. So here in my example the owner name is hardcoded as 'Smith'. Change this value to replace your dynamic variable name of the owner from the Apply to Each Loop. Also for the demonstration purpose I have used two IF conditions. When you implement this logic you only need one IF condition. The second IF condition Yes logic move into the first IF condition Yes part. Hope it make sense.

Here is the flow I built

If condition expression I used is as follows:

string(variables('OwnerDictionary'))

Compose 2 expression is as follows:

setProperty(variables('OwnerDictionary'),'Smith',1)

Set variable is set the output of Compose 2

Next is my second IF condition

Expression I used is as follows:

For the IF condition

string(variables('OwnerDictionary'))

Compose

variables('OwnerDictionary')['Smith']

Compose 3

add(int(outputs('Compose')),1)

Compose 4

setProperty(variables('OwnerDictionary'),'Smith',outputs('Compose_3'))

Finally Set variable is set the output of Compose 4

If you have any questions please let me know.

Thanks

[ZedYer]

Hi @abm Hoping this will work as I'm learning about variables and setProperty, etc. 

Am stuck on...

Compose

variables('OwnerDictionary')['Smith']

...in your second IF demonstration. I am adding it to the YES section of the first IF statement as you advised.

I read through all this and, honestly, am not exactly sure what or how it is supposed to work. Once it is running, I should be able to figure it out though.

In this case, "Smith" should already be in the object, so I assume I need to identify it using the dynamic "owner" which is  items('Apply_to_each')['OPLM']

Thus, full expression is variables('OPLMDictionary')['items('Apply_to_each')['OPLM']']

Error: not a valid expression.  I tried a few permutations of the above with parans and such and tried searching for some solutions with no luck. Am thinking there is some escape character or something like that which is missing.

Thoughts?

[abm]

Hi @ZedYer 

Try the following

  variables('OPLMDictionary')[items('Apply_to_each')['OPLM']

[abm]

Hi @ZedYer , @Anonymous 

How you progressing with this one?

Thanks

[ZedYer]

Hi ABM, I poked around with it for two days and couldn't get it to work.  Between Compose 2 and set variable, I kept getting an error about the object not being able to be set with a string, so I messed around with wrapping it in a json function with no luck. studied up on setProperties, etc. for use in the future so learned something here.  At end of the day, I just moved all my processing inside the loop and was able to get all the other processing done at the owner level, so I have a working PA flow that does what I need.  I really appreciate your input--you put some time into it and I appreciate it.  Thanks for everything.

[abm]

Hi @ZedYer 

If needed I can look this via screenshare. Please let me know.

Thanks

[ZedYer]

Just saw private note. Thanks so much