Describe the Bug

The code generator doesn't seem to produce correct Python type annotations.

Given a simple schema in schema.graphql:

type Query {
    name: String
    other: Other
}

type Other {
    foo: String
}

Running strawberry schema-codegen schema.graphql generates the following python code:

import strawberry

@strawberry.type
class Query:
    name: str | None
    other: Other | None # <--- reference to `Other`

@strawberry.type
class Other:
    foo: str | None

schema = strawberry.Schema(query=Query)

The class Query references Other before it is defined. Running the code makes Python complain:

Traceback (most recent call last):
  File ".../schema.py", line 4, in <module>
    class Query:
  File ".../schema.py", line 6, in Query
    other: Other | None
           ^^^^^
NameError: name 'Other' is not defined. Did you mean: 'iter'?

Possible fixes are:

• Reordering the types so Other comes first and thus making the reference valid. This works as long as the types form no cycle.
• Adding from __future__ import annotations to the Python code making forward references legal. See https://mypy.readthedocs.io/en/stable/runtime_troubles.html#class-name-forward-references
• Rewrite the type annotation as typing.Optional["Other"] (Note the string quotes). Interestingly, "Other" | None is not valid, so Optional has to be used here.

System Information

• Python 3.11.6
• Operating system: Linux
• Strawberry version (if applicable): 0.215.1 (most recent)

Additional Context

Of course, for a small schema those issues can be quickly fixed, but for a large schema this isn't really feasible.