allow ignore_changes to reference any map key

[jbardin]

This change will allow specifying a map["key"] path in ignore_changes that may not exist in the config or the prior value. This allows users to prevent a single map element from changing to or from being unset.

This didn't work previously because the transformation always started off with the configuration, and would never encounter a key if it was not present in the configuration. We also couldn't retain the absence of a value, since the lookup wouldn't find the key in the prior value.

The following configuration will now prevent the "unwanted" value from showing up in the plan if it is not originally present in the prior state.

resource "test_resource" "a" {
  map = {
    wanted = "value"
    unwanted = "added after initial apply"
  }
  lifecycle = {
    ignore_changes = [map["unwanted"]]
  }
}

The following example will now prevent the "wanted" value from being removed if it exists the the prior state, but is removed from the configuration

resource "test_resource" "a" {
  map = {
    // wanted = "removed after initial apply"
  }
  lifecycle = {
    ignore_changes = [map["wanted"]]
  }
}

While these examples use map literals for clarity, this feature would normally be used when the map is derived from another source, and not manipulated directly in the resource configuration. This also allows users in some cases to work around misbehaving legacy providers that have altered the state in conflict with the configuration.

[codecov]

Codecov Report

Merging #26421 into master will increase coverage by 0.01%.
The diff coverage is 86.79%.

Impacted Files	Coverage Δ
terraform/eval_diff.go	70.45% <86.79%> (+1.70%)	⬆️
dag/marshal.go	53.42% <0.00%> (-1.37%)	⬇️
terraform/evaluate.go	50.11% <0.00%> (-0.46%)	⬇️

[apparentlymart]

I unfortunately don't have enough time left today for a deep review of this, but just wanted to quickly note that this changes the documented behavior of ignore_changes and so we'll need to do some documentation updates along with this behavior change.

[jbardin]

Thank you, I forgot we documented that caveat of the implementation. I'll add the doc changes to his PR too.

[mildwonkey]

👍 the real-world use case for this makes lots of sense to me!

[jbardin]

Just rebased in some comment changes which I thought I would help with review.

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.