feat: Add support for (custom) types #498

gfrey · 2025-04-09T07:28:58Z

This is the first implementation that was discussed in the slack channel. The other idea of having the types field in the spec part of the schema, seems to be more complicated to implement.

Fixes #144.

This fragment is repeated in the downstream functions. I guess it is a left-over from refactorings.

This adds the `Types` field to the ResourceGraphDefinition CRD. It contains a map of names to type definitions, where the definitions follow the simple schema. The defined types are then loaded into the transformer as predefined types. Appearances in the spec are then properly resolved. An example for usage would be: ``` yaml apiVersion: kro.run/v1alpha1 kind: ResourceGraphDefinition metadata: name: service-entry-dns spec: schema: apiVersion: v1alpha1 kind: ServiceEntryDNS spec: ports: "[]port | required=true" types: port: name: string protocol: string number: integer | required ``` Known issues: Currently there is no support for references of custom types within the definition of a custom type! Fixes kro-run#144

If a simple type was used with the required marker: ``` yaml types: myType: string | required=true description="my description" spec: myValue: myType ``` the required marker was lost, while the description was set properly. This happened as in the applyMarkers method of the transformer the required information is set to the parent. While loading the predefined types this parent is just a shell object, that is later read to fill the list of predefined types and then discarded. Now the information on required types is captured in the `predefinedType` type and then later used to set the parent object required fields when handling predefined types.

website/docs/docs/concepts/10-simple-schema.md

a-hilaly

Thanks @gfrey! left two tiny comments

a-hilaly · 2025-04-16T05:30:26Z

pkg/simpleschema/transform.go

+// A predefined type is a type that is predefined in the schema.
+// It is used to resolve references in the schema, while capturing the fact
+// whether the type has the required marker set (this information would
+// otherwise be lost in the parsing process).
+type predefinedType struct {


Good call! initially i wasn't sure about this, but i just remembered that required fields always live parent-type-schema-definition.

I just noticed that adding this broke the test for the predefined types. So, I fixed that.

a-hilaly · 2025-04-16T05:42:53Z

pkg/simpleschema/transform_test.go

-	if err != nil {
-		t.Fatalf("Failed to load pre-defined types: %v", err)
-	}
+func floatPtr(f float64) *float64 {


looks like this function isn't used, shall we remove it?

Absolutely!

a-hilaly · 2025-04-21T22:36:05Z

flake.lock

can you please exclude nix files from this PR?

Yes, I have force pushed the commits that contained them. Sorry for that!

I would love to see the nix files setup for kro. Would you share it in a gist please ?

@barney-s https://gist.github.com/gfrey/2101b155e1d1d9f56517bfd00fa06dcb is what I'm using.

This fragment is repeated in the downstream functions. I guess it is a left-over from refactorings.

This is reported when `gofmt` is run with the `-s` option.

a-hilaly

Neat, thank you @gfrey !!

barney-s · 2025-05-08T23:34:39Z

Have a top level Q.
Since this is schema level changes how is spec and status different from custom types ?

Instead of this:

schema:
  types:
    Person:
      name: string
      age: integer
  spec:
    people: '[]Person | required=true`

How does this look like ?

schema:
   Person:
     name: string
     age: integer
  spec:
    people: '[]Person | required=true`

??

barney-s · 2025-05-13T00:02:28Z

Another alternative is:

schema:
  spec:
    people:
    - _properties: "array | required=true"
      name: string
      age: integer

for maps:

schema:
  spec:
    people: `[]string | required=true`
    peeps:
    - _properties: "array | required=true"
      name: string
      age: integer

* Add support for (custom) types This adds the `Types` field to the ResourceGraphDefinition CRD. It contains a map of names to type definitions, where the definitions follow the simple schema. The defined types are then loaded into the transformer as predefined types. Appearances in the spec are then properly resolved. An example for usage would be: ``` yaml apiVersion: kro.run/v1alpha1 kind: ResourceGraphDefinition metadata: name: service-entry-dns spec: schema: apiVersion: v1alpha1 kind: ServiceEntryDNS spec: ports: "[]port | required=true" types: port: name: string protocol: string number: integer | required ``` Known issues: Currently there is no support for references of custom types within the definition of a custom type! Fixes kro-run#144 * Add documentation for (custom) types * Fix (custom) simple types with required marker If a simple type was used with the required marker: ``` yaml types: myType: string | required=true description="my description" spec: myValue: myType ``` the required marker was lost, while the description was set properly. This happened as in the applyMarkers method of the transformer the required information is set to the parent. While loading the predefined types this parent is just a shell object, that is later read to fill the list of predefined types and then discarded. Now the information on required types is captured in the `predefinedType` type and then later used to set the parent object required fields when handling predefined types. * Fix typo in documentation * Remove unused code * Fix and improve test for predefined types * Simplify code

Gereon Frey added 4 commits April 8, 2025 14:12

Remove unused code

be3d278

This fragment is repeated in the downstream functions. I guess it is a left-over from refactorings.

Add documentation for (custom) types

9b1331e

csantanapr reviewed Apr 10, 2025

View reviewed changes

website/docs/docs/concepts/10-simple-schema.md Outdated Show resolved Hide resolved

Fix typo in documentation

d096b06

a-hilaly reviewed Apr 16, 2025

View reviewed changes

a-hilaly reviewed Apr 21, 2025

View reviewed changes

Gereon Frey added 3 commits April 22, 2025 09:16

Remove unused code

c8af39f

This fragment is repeated in the downstream functions. I guess it is a left-over from refactorings.

Fix and improve test for predefined types

b5d4707

Simplify code

c824534

This is reported when `gofmt` is run with the `-s` option.

gfrey force-pushed the custom_types branch from fb84337 to c824534 Compare April 22, 2025 07:17

gfrey requested a review from a-hilaly April 29, 2025 07:27

a-hilaly approved these changes May 8, 2025

View reviewed changes

barney-s merged commit 095e3c0 into kro-run:main May 13, 2025
5 checks passed

gfrey deleted the custom_types branch May 23, 2025 12:53

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: Add support for (custom) types #498

feat: Add support for (custom) types #498

Uh oh!

gfrey commented Apr 9, 2025

Uh oh!

Uh oh!

a-hilaly left a comment

Uh oh!

a-hilaly Apr 16, 2025

Uh oh!

gfrey Apr 16, 2025

Uh oh!

a-hilaly Apr 16, 2025

Uh oh!

gfrey Apr 16, 2025

Uh oh!

a-hilaly Apr 21, 2025

Uh oh!

gfrey Apr 22, 2025

Uh oh!

barney-s Apr 30, 2025

Uh oh!

gfrey Apr 30, 2025

Uh oh!

a-hilaly left a comment

Uh oh!

barney-s commented May 8, 2025

Uh oh!

barney-s commented May 13, 2025 •

edited by a-hilaly

Loading

Uh oh!

Uh oh!

Uh oh!

feat: Add support for (custom) types #498

feat: Add support for (custom) types #498

Uh oh!

Conversation

gfrey commented Apr 9, 2025

Uh oh!

Uh oh!

a-hilaly left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

a-hilaly left a comment

Choose a reason for hiding this comment

Uh oh!

barney-s commented May 8, 2025

Uh oh!

barney-s commented May 13, 2025 • edited by a-hilaly Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

barney-s commented May 13, 2025 •

edited by a-hilaly

Loading