Create a filter implementation for K8s runtime.Objects

[michaelmdresser]

What does this PR change?

• Introduces a new filter parser and MatchCompiler implementation targeting the runtime.Object type in the K8s library
• Moves filter field strings to a common package to avoid duplication and drift in future filter implementations

Does this PR relate to any other PRs?

• Not in Opencost (will be used by a downstream dependency to use Opencost-style filter strings for a new purpose)

How will this PR impact users?

• N/A

Does this PR address any GitHub or Zendesk issues?

• N/A

How was this PR tested?

• New unit tests

Does this PR require changes to documentation?

• No

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
opencost	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Mar 13, 2024 7:50pm

[mbolt35]

One very explicit thing you will need is to register the fields in the ops package either by adding it directly here:

opencost/core/pkg/filter/ops/ops.go

Lines 28 to 33 in 8a4002a

	var defaultFieldByType = map[string]any{
	typeutil.TypeOf[allocation.AllocationField](): allocation.DefaultFieldByName,
	typeutil.TypeOf[asset.AssetField](): asset.DefaultFieldByName,
	typeutil.TypeOf[cloudcost.CloudCostField](): cloudcost.DefaultFieldByName,
	// typeutil.TypeOf[containerstats.ContainerStatsField](): containerstats.DefaultFieldByName,
	}

(this is likely the best choice -- just replace the // typeutil.TypeOf[containerstats.ContainerStatsField](): containerstats.DefaultFieldByName, line - heheheh)

If the fields are added outside of the opencost repo, you can use this in an init:

opencost/core/pkg/filter/ops/ops.go

Line 36 in 8a4002a

func RegisterDefaultFieldLookup[T ~string](lookup func(T) *ast.Field) {

[michaelmdresser]

TIL! Done, that point about stuff outside of the Opencost repo is neat as heck

[mbolt35]

Did go force you to add the type cast when you referred to the fieldstrings constants?

ie:

const FieldNamespace K8sObjectField = "namespace"

is fine, but

const FieldNamespace K8sObjectField = fieldstrings.FieldNamespace

barks at you that they're not the same type?

Bah

[michaelmdresser]

Yeah, it barks; I'm glad it does, otherwise these enum-like things would be pretty useless.

[mbolt35]

Just wanted to point out (if you haven't noted field attributes yet) that ast.FieldAttributeNilable (or any attribute that might be needed) on a field is simply metadata that can be accessed in the compiler or code generator to help alleviate any special cases that may exist.

[mbolt35]

Attributes are just bitflags (sorry, I think I'm showing my age), so they're not super expansive, but likely more space than we'll ever need. I think Nilable starts at the 5th bit, and we're using an int, so it's expandable to another 27 custom attributes 🤷

[michaelmdresser]

Oh very neat, I just copied whatever Allocation was using for consistency; I think it also makes sense here.

[mbolt35]

LGTM

[sonarcloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
No data about Coverage
No data about Duplication

See analysis details on SonarCloud