-
Notifications
You must be signed in to change notification settings - Fork 2k
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
Strip all Type System Directives during composition #3736
Conversation
> Note: This is not the solution, just a test that demonstrates the failure! Recently, the beginnings of [Executable directive] support were brought into federation through apollographql#3464 and its follow-up, apollographql#3536. The implementation of executable directives currently in place within federation requires that federated directives be declared identically across all implementing services, even if their implementation is different (or a no-op). This is working as intended! However, [Type system directives] need some additional pruning from the composed schema. Specifically, while type system directive declarations don't need to be declared identically across implementing services, their _definitions_ are currently (intentionally!) removed from the composed schema. That would be fine, but the _usages_ of those directives are currently being left in-tact, which results in validation errors within composed schemas since the directives, while still defined in the implementing services, have been removed. It's certainly important that the implementing services maintain those type system directives in order to act upon them within the services themselves, I don't believe those directives need to be preserved in the composed schema that runs at the gateway and is used to generate the query plan. This commit is mostly documenting a reproduction of an issue that was brought to my attention and introduces what I believe is a test case to build a solution against. [Type system directive]: https://graphql.github.io/graphql-spec/draft/#TypeSystemDirectiveLocation [Executable directive]: https://graphql.github.io/graphql-spec/draft/#ExecutableDirectiveLocation
@josemarluedke: Thank you for submitting a pull request! Before we can merge it, you'll need to sign the Apollo Contributor License Agreement here: https://contribute.apollographql.com/ |
Thanks for taking the time to put this PR together, @josemarluedke! This is pretty close to what we want, I've created a few commits and pushed them up to show how I'd expand the scope of this a bit. Feel free to take inspiration from or cherry-pick them altogether. In short, Let me know your thoughts and when you're ready for another review. I'm also happy to land these changes myself and maintain your authorship - let me know what works best for you! A changelog entry would also be appreciated: Thanks again 😄 |
@trevor-scheer Thanks for creating these commits. I have cherry-picked them here. I also added an entry to the federation changelog. Let me know if anything is missing, but I believe this is good to go. |
It's bad in this file altogether, but broad re-formatting of this file won't do anything better than stomp on history. Since this line is already changing, it's worth a fix in the same commit.
The overall height of this block seems to be a ripe opportunity to try to lock down the re-assignment of this objects' reference, even if it wasn't introduced by this PRs intended change.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you so very much for taking the initiative to pick this up where I left it and run with it. LGTM. We'll get this into a release soon.
"directives at FIELDs are executable" | ||
directive @audit(risk: Int!) on FIELD | ||
|
||
"directives at FIELD_DEFINITIONs are for the type-system" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It seems that @deprecated
defies this semantic: https://www.apollographql.com/docs/graphql-tools/schema-directives/
We think this PR may accidentally be stripping |
This commit resolves a breaking and unintentional change from #3736. While type-system directives _should_ be removed, @deprecated is a special case which can and should be left in.
This commit resolves a breaking and unintentional change from #3736. While type-system directives _should_ be removed, @deprecated is a special case which can and should be left in. Co-authored-by: Jesse Rosenberger <git@jro.cc>
Correct assessment above by @glasser in #3736 (comment) and fixed by #3792 |
…ollo-server#3736) * tests: Show composition failures w/ undeclared type-system directives. > Note: This is not the solution, just a test that demonstrates the failure! Recently, the beginnings of [Executable directive] support were brought into federation through apollographql/apollo-server#3464 and its follow-up, apollographql/apollo-server#3536. The implementation of executable directives currently in place within federation requires that federated directives be declared identically across all implementing services, even if their implementation is different (or a no-op). This is working as intended! However, [Type system directives] need some additional pruning from the composed schema. Specifically, while type system directive declarations don't need to be declared identically across implementing services, their _definitions_ are currently (intentionally!) removed from the composed schema. That would be fine, but the _usages_ of those directives are currently being left in-tact, which results in validation errors within composed schemas since the directives, while still defined in the implementing services, have been removed. It's certainly important that the implementing services maintain those type system directives in order to act upon them within the services themselves, I don't believe those directives need to be preserved in the composed schema that runs at the gateway and is used to generate the query plan. This commit is mostly documenting a reproduction of an issue that was brought to my attention and introduces what I believe is a test case to build a solution against. [Type system directive]: https://graphql.github.io/graphql-spec/draft/#TypeSystemDirectiveLocation [Executable directive]: https://graphql.github.io/graphql-spec/draft/#ExecutableDirectiveLocation * Remove directives from field definition in composition * Expand functionality to all custom type system directives * Update stripping mechanism to include all TypeSystemDirectives * Jest auto-prettification of snapshots * Rename var in composition to match implementation * Add changelog entry for PR apollographql/apollo-server#3736 * nit: Line-length. It's bad in this file altogether, but broad re-formatting of this file won't do anything better than stomp on history. Since this line is already changing, it's worth a fix in the same commit. * Change `definition` to be a `const`-ant rather than `let`. The overall height of this block seems to be a ripe opportunity to try to lock down the re-assignment of this objects' reference, even if it wasn't introduced by this PRs intended change. * Add a comment about what `null` return does on visits. * Add a comment about why we're stripping type system directives. Co-authored-by: Jesse Rosenberger <git@jro.cc> Co-authored-by: Trevor Scheer <trevor.scheer@gmail.com> Apollo-Orig-Commit-AS: apollographql/apollo-server@db71535
…l/apollo-server#3792) This commit resolves a breaking and unintentional change from apollographql/apollo-server#3736. While type-system directives _should_ be removed, @deprecated is a special case which can and should be left in. Co-authored-by: Jesse Rosenberger <git@jro.cc> Apollo-Orig-Commit-AS: apollographql/apollo-server@6005f50
@josemarluedke any ideas why we need to strip directives from service schema (introspection) itself? it's hard for service-level-engineer to work with their own schema, moreover, how to track changes to the schema? Probably we can strip all directives from SDL, but not from schema itself. Any concerns? |
Fix #3565.