Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: add cmd to generate a JSON representation of schema #5919

Merged
merged 5 commits into from
Mar 12, 2024

Conversation

sgulseth
Copy link
Member

@sgulseth sgulseth commented Mar 6, 2024

Description

Adds a new command that can be executed within a sanity context to extract a JSON representation of the studios schema. This makes it easier for other tools(TM) to consume it without depending on browser environments etc. I tried to base most of the logic on sanity schema validate to keep consistency in the repo/package.

What to review

Correctness: Do we parse the schema correctly?
Naming things: It's hard! Can we name the command line argument or flags differently?
Note that we consider this feature beta, anything more than in the help text it should be indicated?

Testing

$ pnpm build
$ sanity schema extract # somewhere there's a sanity.config.ts

Notes for release

Beta: Extract sanity schema as JSON, to be consumed by other tools

Copy link

vercel bot commented Mar 6, 2024

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

Name Status Preview Comments Updated (UTC)
performance-studio ✅ Ready (Inspect) Visit Preview Mar 12, 2024 10:44am
test-studio ✅ Ready (Inspect) Visit Preview 💬 Add feedback Mar 12, 2024 10:44am
1 Ignored Deployment
Name Status Preview Comments Updated (UTC)
studio-workshop ⬜️ Ignored (Inspect) Visit Preview Mar 12, 2024 10:44am

Copy link
Contributor

github-actions bot commented Mar 6, 2024

No changes to documentation

Copy link
Contributor

github-actions bot commented Mar 6, 2024

Component Testing Report Updated Mar 12, 2024 10:48 AM (UTC)

File Status Duration Passed Skipped Failed
comments/CommentInput.spec.tsx ✅ Passed (Inspect) 32s 15 0 0
formBuilder/ArrayInput.spec.tsx ✅ Passed (Inspect) 6s 3 0 0
formBuilder/inputs/PortableText/Annotations.spec.tsx ✅ Passed (Inspect) 11s 3 0 0
formBuilder/inputs/PortableText/copyPaste/CopyPaste.spec.tsx ✅ Passed (Inspect) 12s 4 2 0
formBuilder/inputs/PortableText/Decorators.spec.tsx ✅ Passed (Inspect) 12s 6 0 0
formBuilder/inputs/PortableText/FocusTracking.spec.tsx ✅ Passed (Inspect) 31s 15 0 0
formBuilder/inputs/PortableText/Input.spec.tsx ✅ Passed (Inspect) 1m 0s 15 0 0
formBuilder/inputs/PortableText/ObjectBlock.spec.tsx ✅ Passed (Inspect) 1m 1s 18 0 0
formBuilder/inputs/PortableText/RangeDecoration.spec.tsx ✅ Passed (Inspect) 11s 6 0 0
formBuilder/inputs/PortableText/Styles.spec.tsx ✅ Passed (Inspect) 13s 6 0 0
formBuilder/inputs/PortableText/Toolbar.spec.tsx ✅ Passed (Inspect) 20s 9 0 0

Copy link

socket-security bot commented Mar 6, 2024

New dependencies detected. Learn more about Socket for GitHub ↗︎

Package New capabilities Transitives Size Publisher

View full report↗︎

Copy link
Member

@rexxars rexxars left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll leave the bulk of the review up to the tagged reviewers, but I have some thoughts/questions:

  • A bit reluctant to tie @sanity/schema to groq-js
    • A schema extract command makes total sense to me, but the resulting format should preferably be generic enough that it can be used to generate all kinds of things; GraphQL types, TypeScript types, tailored API clients etc.
    • Generally I would favor using a more established output type, say JSONSchema, as long as it is able to represent our schema completely.
    • Perhaps we could introduce a format parameter/flag? We don't have to implement any other formats right now, but I think requiring the flag for now makes it more future-compatible if we want to default to (say) JSONSchema in the future.
  • Using the compiled schema (Schema.compile()/createSchema()) would be safer
    • We shouldn't have to special-case image/file fields/types etc - and defining their properties are already done for you with this approach. It feels brittle/unnecessary to have to maintain two code paths if we add new fields.
    • The generated type output may be missing types such as image and file asset documents, geopoints, slugs and other non-user-defined types
    • Block types shouldn't have to be special-cased - they (should) just be an object type like any other. Given you can create named block types, this is important - as the name property may not be block
    • Without having looked deeply at the code, I am curious if the current approach handles aliased types, or if that is left to whatever acts on these types to figure out

packages/@sanity/schema/package.json Show resolved Hide resolved
Options
--workspace <name> The name of the workspace to generate a schema for
--path Optional path to specify destination of the schema file
--enforce-required-fields Makes the schema generated respect required fields in the schema. This can be used when generating types if you are only querying for published documents. Defaults to false.
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  • This seems broader than just for querying - I would phrase this to be "treat fields marked as required as non-optional", as what it impacts is the resulting optional attributes
  • "querying for published documents" doesn't really encompass this - both drafts and published documents can be incomplete/not follow schema (documents that were published prior to a field being introduced, for instance)

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems broader than just for querying - I would phrase this to be "treat fields marked as required as non-optional", as what it impacts is the resulting optional attributes

True, and agree on your phrasing 👍

This seems broader than just for querying - I would phrase this to be "treat fields marked as required as non-optional", as what it impacts is the resulting optional attributes
"querying for published documents" doesn't really encompass this - both drafts and published documents can be incomplete/not follow schema (documents that were published prior to a field being introduced, for instance).

Yeah, we do however consider the schema the source of truth so if anything diverges from your schema we don't give any guarantees with the types at least 🤔

packages/sanity/src/_internal/cli/threads/extractSchema.ts Outdated Show resolved Hide resolved
packages/sanity/src/_internal/cli/threads/extractSchema.ts Outdated Show resolved Hide resolved
Copy link

socket-security bot commented Mar 8, 2024

👍 Dependency issues cleared. Learn more about Socket for GitHub ↗︎

This PR previously contained dependency changes with security issues that have been resolved, removed, or ignored.

View full report↗︎

@sgulseth
Copy link
Member Author

sgulseth commented Mar 8, 2024

A bit reluctant to tie @sanity/schema to groq-js.

Where would you prefer these types to live?

I don't see any reasons this couldn't be used with our graphql generator as well, would potentially need to extend ObjectAttribute to include more data, but that's doable.

We did talk about using jsonschema, but the ecosystem is a bit fragmented with different specification and support, so we opted for our own implementation here.

cc. @judofyr for additional feedback.

Perhaps we could introduce a format parameter/flag

We could. Would then this be a required flag to set? ie sanity schema extract --format=typenode Could it default to typenode at least while being beta? To simplify DX and feedback for now.

Using the compiled schema (Schema.compile()/createSchema()) would be safer

Ah, nice. I'll look into this!

Without having looked deeply at the code, I am curious if the current approach handles aliased types, or if that is left to whatever acts on these types to figure out

It does, they are referred to as InlineTypeNode since the value are inlined in Content Lake

@bjoerge
Copy link
Member

bjoerge commented Mar 8, 2024

A bit reluctant to tie @sanity/schema to groq-js.

Where would you prefer these types to live?

Any reason extractSchema couldn't live with the cli command for now?

@sgulseth
Copy link
Member Author

sgulseth commented Mar 8, 2024

@bjoerge then we either need to have the types in two places, guess that works with how ts infers types. Or groq-js would need to depend on the cli, which isn't ideal either

@rexxars
Copy link
Member

rexxars commented Mar 8, 2024

Where would you prefer these types to live?

Not sure, honestly - probably fine for now.

We did talk about using jsonschema, but the ecosystem is a bit fragmented with different specification and support, so we opted for our own implementation here.

Would be interested in learning more about this, but I don't need a full write-up here - if there's any TLDR or links I can use to understand more that'd be great.

Perhaps we could introduce a format parameter/flag

We could. Would then this be a required flag to set? ie sanity schema extract --format=typenode Could it default to typenode at least while being beta? To simplify DX and feedback for now.

Ideally I would default to something that is a defined standard, such as JSONSchema, and make other formats a flag. I understand that it's an improved DX when we only have a single format, but it would also be a worse DX when we change the default (breaking change).

@sgulseth
Copy link
Member Author

I have adressed the feedback and we are now processing the compiled types. OK to move the format discussion to slack?

Copy link
Member

@rexxars rexxars left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Overall this looks good to me, but would love if @bjoerge who knows the schema better could have a look.

One curiosity; seems to me like Portable Text block and span types do not have a _type property? Span should always be span (the type cannot be changed in any way), and the block type should either be block or the name given by the user.

Copy link
Member

@bjoerge bjoerge left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Great work! Left a few comments for your consideration, only nits/nice-to-haves, so good to merge from my side!

packages/@sanity/schema/package.json Outdated Show resolved Hide resolved
packages/@sanity/schema/src/sanity/extractSchema.ts Outdated Show resolved Hide resolved
packages/@sanity/schema/src/sanity/extractSchema.ts Outdated Show resolved Hide resolved
@sgulseth sgulseth added this pull request to the merge queue Mar 12, 2024
Merged via the queue into next with commit c1e4f2a Mar 12, 2024
37 checks passed
@sgulseth sgulseth deleted the feat/schema-generation branch March 12, 2024 10:52
juice49 pushed a commit that referenced this pull request Mar 13, 2024
* feat: add cmd to generate a JSON representation of schema

* feat: refactor to use compiled schema

* fix: append _type to objects

* chore(schema): clean up and describe execution

* chore(schema): lock groq-js to minor canary
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants