3 approaches to reshape data in Power Automate Efficiently

Earlier this week I was asked to reshape data in Power Automate. Having an array of data that needs to be reshaped just sot that an API can take the data can be tricky. So often I find clients using the Append to array action inside an apply to each. Now if your array only has a few records it probably doesn’t matter too much but what if you have 1000s of records?

The Data samples

The sample input data that I was given looks like this:

[
  {
    "firstname": "John",
    "lastname": "One",
    "phone": "+15035551234",
    "hs_language": "en-us"
  },
  {
    "firstname": "Minh",
    "lastname": "Two",
    "phone": "+15035555678",
    "hs_language": "vi"
  }
]

The required data needed to look like this:

[
    {
      "mobile": {
        "number": "+15035551234",
        "country": "US"
      },
      "firstName": "John"
    },
    {
      "mobile": {
        "number": "+15035555678",
        "country": "US"
      },
      "firstName": "Minh"
    }
]

So how do we approach this?

Approach 1 – Append to Array

The first approach, which is often used, uses the append to array action. If you have a small array this is probably ok.

However for large arrays this isn’t an efficient way of doing things, as the concurrency settings will not work inside apply to each steps when you use variables inside the apply to each.

The expressions used for phone and first name are:

items('Apply_to_each')['phone']
items('Apply_to_each')['firstname']

If you are working with fields that may be empty you could also use the JSON query syntax

items('Apply_to_each')?['phone']
items('Apply_to_each')?['firstname']

Approach 2 – Pieter’s method

The second approach, which is a bit better than the first approach is to use Pieter’s method.

Using the second approach means that we avoid any variables, which opens up the option to process the data faster when we enable to Concurrency/Parallelism in the settings of the Apply to each step.

The expressions used for phone and first name are:

items('Apply_to_each_2')['phone']
items('Apply_to_each_2')['firstname']

If you are working with fields that may be empty you could also use the JSON query syntax

items('Apply_to_each_2')?['phone']
items('Apply_to_each_2')?['firstname']

Approach 3 – Use the select action

The fastest of the three approaches is the select action. The select action is the best approach.

If you use the above Mapping mode where you specify each of the properties or the below text mode doesn’t really matter.

Putting the approaches to the test.

The expressions used for phone and first name are:

item()['phone']
item()['firstname']

If you are working with fields that may be empty you could also use the JSON query syntax

item()?['phone']
item()?['firstname']

Comparing the options

Before enabling the parallelism setting on we will find that processing 75 records takes about 13 seconds when the apply to each steps are involved. The select however takes no time at all to run.

When we enable the parallelism and set it to the maximum of 50 then we will see the real difference in the three approaches.