Showing results for 
Search instead for 
Did you mean: 
Advocate I
Advocate I

Extracting a substring from parsed JSON data in a Loop

I'm stumped on this one, folks.

I am taking JSON data from an online form, parsing it, and converting it into a .csv for import. I have built and tested a flow that works exactly as needed - almost. 

The online form gives users the option to enter information for anywhere from 0 to 10 children (family law intake). So, the JSON can contain data for 0 to 10 children. Incoming data for the fields contains either a six-or-eight-character Code, and then an explanation (it's in a drop-down on the online form). I need to export only a substring of that data (either the left 6 or 8 characters) into the .csv for import.


Problem is, I can't use Initialize Variable to do this within a loop (that will run x times, where x = 0 to 10).

Is there a way to substring just the raw parsed JSON data so I only grab the characters I need for inclusion in the variable table I have created for the data (prior to export to .csv)? I have tried to use the 'substring' expression, but it won't even let me select the variable I need from within the sub-data for "CHILDINV" - I'm stuck.

Child Data Loop.JPG

Advocate I
Advocate I

No, we want a consistent header line, even if there is no data in the field below. Makes it easier that way.

@Paulie78 's solution looks a bit less daunting tbh, but I'm stuck on understanding how he is referencing the fields.


I now see that doing it the way I was doing it, via variables, is long and cumbersome - this "compose" function makes it more of a one-shot schema creation. As much as I hate to throw away what I have done already, I like that it moves all of the expressions and fiddling with the data into one place.

I have lots to learn on this stuff, it's only the first of many I am going to have to create. Thanks so much for your patience and efforts thus far.

If you need help building it reach out via private message and will get you started off.

It does produce two rows, perhaps I copied the wrong scope. Will check in a moment. 

Super User
Super User

Absolutely, @ArchieGoodwin, here's a solution which produces all the rows, all the headers, and the full 10 kids.


I'll admit that were this me I'd space out the logic a bit more so that others won't be blinded by the long expressions on show. One of the main things I try to show folks in a work environment that Power Automate makes an office less reliable on Jon/Janey and their wizard excel formulas. As soon as you start sprinkling long 'formula' everywhere here, anyone taking over the management of it might still be a bit taken aback.


If I were to do that here, I might make the two sides of that union in the CSV separate object variables then reference them.


This solution is transferrable if you want to apply it to more than one array in the original data.


@Paulie78 thanks for the help, and for the range fiddle, too! I had originally done ten fields that long way, and thought that there *had* to be a way to get it done with less typing. 😅



This just counts the number of kids, and subtracts it by one. This saves overcomplicating the multiple if() statements in the Select action.



sub(length(variables('jsonInputVAR')?['body']?['CHILDINV']), 1)




In the 'From' field this uses a trick that Paulie taught me elsewhere, the range() function. So it iterates through 10 items, regardless.



range(0, 10)



In the 'Map' 'key' fields (left), this uses some basis logic looking at the current item() number in the range, and adds 1. This provides your child number.



add(item(), 1)



In the 'Map' 'value' fields (right), this uses a simple condition. If the current item() number is greater than the number returned in the kidsVAR variable, then it places nothing ('') in the field, otherwise it picks the child data that is correspondent to the current item() number. It looks like more than it is. 👍



if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['Id'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1NAME'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1BD'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['Gender'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1LEGAL'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1PHYS'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1VISIT'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['FIELD8'])

if(greater(item(), variables('kidsVAR')), '', variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['ItemNumber'])



Create CSV Table

This one is basically constructing an array for the Create CSV Table action to make its data from. Don't forget to place the square brackets around it in the 'From' field:



union(removeProperty(variables('jsonInputVAR')?['body'], 'CHILDINV') , json(replace(join(body('SelectKidsData'), ', '), '}, {', ',')))



The union() function is usually used to join two arrays together, but it works with objects, too, so the main thing here is ensuring that the child data is represented as an object instead of an array, after the previous Select action has done its thing.


So, in the second section of that union function this is done by:

  1. Using a join() function to join the separate children items together in a string value.
  2. Once they're in the string, they're still actually separated by curly brackets, so the replace() function removes the ones separating the children '}, {', and puts a normal comma in their place.
  3. Then finally the json() function makes that all readable by the union otherwise it would error thinking it's just a string.


The first part is getting all the non-child data by simply using the removeProperty() function to strip the CHILDINV data out of the original object.


Additional Arrays

To apply this to other fields in the original JSON just requires a few tweaks. Let's look at the potential for the DOCUMENTS array field to need it.

  1. Add a branch after the input.
  2. Initialise a docsVAR integer variable in the new branch to count the items in DOCUMENTS.
  3. For the Select copy the original and edit it to save time to SelectDocsData.
  4. Change the letters either side of the add(item(),1) expressions in the key fields to make your document headers.
  5. Edit each value to point to DOCUMENTS instead of CHILDINV, the new docsVAR for the counts, and change the referenced field names.
  6. Finally, bridge the branches with the CSV:


union(removeProperty(removeProperty(variables('jsonInputVAR')?['body'], 'CHILDINV'), 'DOCUMENTS') , json(replace(join(body('SelectKidsData'), ', '), '}, {', ',')) , json(replace(join(body('SelectDocsData'), ', '), '}, {', ',')))


If there is no set limit to the documents, then take the docsVAR and add 1 to it for the other side of the range() expression.


EDIT - I've tested this on ten children ... and they're all still alive to tell the tale!

One teeny tiny problem. I can't figure out how to put the SUBSTRING instruction into the map. I keep getting "invalid expression" errors; how do I grab only, say, the first 6 characters in C1LEGAL for example? I know it's probably a syntax issue.

Where would you squeeze in the SUBSTRING?

Also - I really only need the CHILD output from this, because I don't carry across all the fields, only ones I need. But, if I try to create a CSV table from OUTPUT, there are a bunch of extra commas in there that I don't need.


Super User
Super User

(sorry, was weirdly scared to come back in case it wasn't a good solution!)


I'm glad all is good, @ArchieGoodwin  ... apologies if I'd misunderstood. But given your responses after Paulie mentioned ensuring there were 10 kids each time, I thought the extra data was what you were after there.


Those extra commas in the example of 4 children are the empty children 5-10. (or empty fields in the base data)


If you wish to have only what's there (for the kids, at least), that's more simple, just change the SelectKidsData to be like in my OG solution with the ['body']?['CHILDINV'] source, item()['ItemNumber'] for the numbered keys, and reference the fields accordingly (item()?['Id']) on the value side. You can delete the count, too.

02 - Kids Data OG.jpg




This part is thankfully easy, I'll give you two ways you can do it and hopefully the first is the way that it can be done easiest.


Method One - Add An Extra Select Value

This assumes that you will always have "TEXT - TEXT", meaning that there is always that space-hyphen-space, there.


On the 'Key' side it will be however you like, as before. I labelled mine as C3LGLSh (where the 3 is the expression).


On the 'Value' side it will be:


if(greater(item(), variables('kidsVAR')), '', first(split(variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1LEGAL'], ' - ')))


You are splitting the "CLIENT - sdfkjhsfgldihsd fg" text on that ' - ', and then with the array that returns you take the first item. A safer bet might be to split on just the hypen, and then wrap the first() in a trim() expression.


This ensures that you are always going to get whatever text is there, even if it's longer than 6 characters.


Method Two - Substring

If you really have to use substring, then this would work with the otherwise same setup as the first method:



if(greater(item(), variables('kidsVAR')), '', substring(variables('jsonInputVAR')?['body']?['CHILDINV'][item()]['C1LEGAL'], 0, 6))



It definitely works, just tested, I just feel that the split+first (+trim!) method is going to be better data capture long term.




If you only need the CHILD output, then you can junk the union() expression and just keep the second half of it. So, non-code view thats:




In the json() you'd have:


json(replace(join(body('SelectKidsData'), ', '), '}, {', ','))




There's a few ways to go about the unimportant fields.

  1. Filter - Filter out stuff you don't want, try to use branching if you can to keep the original data available when needed.
  2. Handle Externally - Keep the extra fields, and handle the relevant output on the other end. You then have that data when you need it.
  3. Layered Property Removal - removeProperty() layers to remove the fields you don't currently need.
  4. Explicit Select Fields - Use Paulie's setup to be very exact precisely which field and where you're taking data from.

#2 would be the most accepted/normal way. For example if you were querying an API for data, you would take what you get, and use what you need.


With #3 you just layer it up, like this:


removeProperty(removeProperty(variables('jsonInputVAR')?['body'], 'CHILDINV'), 'DOCUMENTS')


That would supply an object with all the original fields except the CHILDINV, and DOCUMENTS, fields. 


Here's one more image to assist, this one has all 10 kids included, obviously:

extra bitz.jpg


Helpful resources

MPA Virtual Workshop Carousel 768x460.png

Register for a Free Workshop

Learn to digitize and optimize business processes and connect all your applications to share data in real time.

New Process Advisor Capabilities carousel.png

Read the blog for the latest news

Read the latest about new experiences and capabilities in the Power Automate product blog.

PA Survey Carousel Image.png

We want to hear from you!

If you are a small business ISV/Reseller, share your thoughts with our research team.

AI Builder AMA June 7th carousel (up on May 25th, take down June 8th) (1).png

'Ask Microsoft Anything' about AI Builder!

The AI Builder team invite you to ask questions and provide helpful answers at our next AMA.

Users online (2,022)