What is the valid regex or the valid characters for the name of a type in a model?

[bubianchi-criteo]

In a model in DSL format or Json format, it looks like defining types with names that contain '/' or '.' characters are supported. For example:
type test/resource
type test.user
It works correctly in a openfga server.
Nevertheless, in its playground or in the VSCode extension, the syntax highlighting and parser report them as errors.
I didn't find in the documentation about a regex or valid characters for type names (as well as for ids).
Is there some doc specifying what are valid type names?

I find pretty intuitive to have characters like '/' , '.'` that would help to represent some "name scope" (like C# namespaces do) and to organize better the model definition, and it would be worth to make them "officially" supported in the doc

[rhamzeh]

The correct regex is the one used by the API:

• Type: https://github.com/openfga/api/blob/main/openfga/v1/openfga.proto#L19 | "^[^:#@\\s]{1,254}$"
• Relation: https://github.com/openfga/api/blob/main/openfga/v1/openfga.proto#L68 | "^[^:#@\\s]{1,50}$"
• Condition: https://github.com/openfga/api/blob/main/openfga/v1/openfga.proto#L39C16-L39C33 | "^[^\\s]{2,256}$"
• Object: https://github.com/openfga/api/blob/main/openfga/v1/openfga.proto#L80 | "^[^\\s]{2,256}$"

There's additionally these set of regexes defined in the API:
https://github.com/openfga/openfga/blob/main/pkg/tuple/tuple.go#L35-L38

	userIDRegex   = regexp.MustCompile(`^[^:#\s]+$`)
	objectRegex   = regexp.MustCompile(`^[^:#\s]+:[^#:\s]+$`)
	userSetRegex  = regexp.MustCompile(`^[^:#\s]+:[^#\s]+#[^:#\s]+$`)
	relationRegex = regexp.MustCompile(`^[^:#@\s]+$`)

Validation has to pass both sets of regexes in order to be valid according to the API.

---

On the VS Code/Playground side we use a different set of less permissive regexes:
https://github.com/openfga/vscode-ext/blob/07e65ccd3a0d533bcbbb2b4495899d73d2656e3b/server/src/openfga-yaml-schema.ts#L126-L154

We have yet more regexes in language:
https://github.com/openfga/language/blob/35310a2e97386126a01db1c4360e013d984f98f2/pkg/js/validator/validate-dsl.ts#L7-L10

We do plan on standardizing these, this will probably happen sometime within the next couple of months as part of a larger work to move validation ownership to the language repo.

Relevant issue: openfga/language#226

[bubianchi-criteo]

Thank you for clarifying

[bubianchi-criteo]

Note tha the lexer of the DSL grammar should also probably updated since it defines as
IDENTIFIER: (LETTER | '_') (LETTER | DIGIT | '_' | MINUS)*;
See https://github.com/openfga/language/blob/35310a2e97386126a01db1c4360e013d984f98f2/OpenFGALexer.g4#L132

[rhamzeh]

That is a different identifier - that is for the names of the types/relations, not the IDs. Yes, we will update it, though it should be more strict than the IDs.

E.g. not sure we want to allow organ!zat!ion as a type/relation name

[bubianchi-criteo]

I would like to propose a pull request with this kind of "extended identifier" lexeme (additionally to the original CEL grammar )
EXTENDED_IDENTIFIER: (LETTER | '_')((LETTER | DIGIT | '_' | MINUS| SLASH | DOT)*(LETTER | DIGIT | '_'))*;
that will be used in the parser rules for the names of types and relations