Fragment args 2024 amendments

[JoviDeCroock]

This spec contains amendments to #1010, a diffed view is available at mjmahone#3.

These amendments are made from comments on the implementation PR and alterations from the new implementation

• Changing fragment-key to just be fragment-name Fragment Arguments Reference Implementation graphql-js#3835 (comment)
• Removing the substitution logic and going closer to the coerce logic we use in the general flow

In general the biggest changes are that we introduce fragmentVariableValues which will be on the groupedFieldSet, these are derived from the arguments in scope of the fragmentDefinition where this field is used.

We introduce localFragmentVariables which as we are traversing down fragment-spreads are a coerced set of variables i.e.

query($b: String!, $c: String!) {
  field3(b: $b) ## uses b from variable-values
  ...A(a: "A", b: "B")
}

fragment A($a: String!, $b: String!) on X {
  field(a: $a) ## uses $a from localVariableValues
  field2(c: $c) ## we have access to variableValues so we take c from there
  ...B(b: $b) ## uses $b from localVariableValues
}

fragment B($b: String!) on X {
  ## This fragment has access to $b and $c but not to $a
  field4(b: $b) ## uses $b from localVariableValues
}

Last but not least we introduce getArgumentValuesFromSpread which looks at the spread and fragment-definition and establishes a coerced set of localVariableValues.

[netlify]

✅ Deploy Preview for graphql-spec-draft ready!

Name	Link
🔨 Latest commit	94a0027
🔍 Latest deploy log	https://app.netlify.com/sites/graphql-spec-draft/deploys/664f72d649b9470008d527cb
😎 Deploy Preview	https://deploy-preview-1081--graphql-spec-draft.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[mjmahone]

This is clear to me :)

Very excited to see progress, and thank you for cleaning up and clarifying the PR!

[mjmahone]

nit: maybe say "for a fragment to define a variable with a name that is also defined on an operation" to disambiguate that we don't mean fragments and operations can define two of the same variable within a single definition

[JoviDeCroock]

Addressed in 48fa841, does that sound better?

[Keweiqu]

The sentence would flow better if we do something like this --

Operation-defined variables have... Fragment-defined variables have ... So a variable used within a fragment must ...

Thoughts?

[JoviDeCroock]

That sounds great, what do you think about 5a73543

[Keweiqu]

Should we specify "must have identical sets of arugments and values?"

[Keweiqu]

and directives?

[JoviDeCroock]

I actually hadn't thought about directives, that's a great point! Addressed in 48fa841

[JoviDeCroock]

@graphql/tsc is there anything I can do to move this forward?

[benjie]

It feels like section 2 should comment on the fact that shadowed variables may not be referenced implicitly by child fragment spreads. (This is probably not the right way of saying that!) E.g. in the following query, Forbidden is spread inside of F and attempts to use variable $a which is ambiguous, and hence not permitted.

query Q($a: Int!, $b: Int!) {
  ...F(a: 7)
}
fragment F($a: Int!) {
  ...Fine
  ...Forbidden
}
fragment Fine {
  b: echo(input: $b)
}
fragment Forbidden {
  a: echo(input: $a)
}

[JoviDeCroock]

@benjie from a purely technical perspective I agree, I was approaching that more from a backwards compatibility perspective. The Fragments variables are only applicable within the context of a FragmentDefinitiion so we don't carry nested values. A fragment is always looked at in isolation and not from a call-stack point of view.

i.e.

query ($a: String!) {
  operation {
    ...fields(a: 'some-value')
  }
}

fragment fields($a: String) on Type {
  field(x: $a) ## this carries 'some-value'
  ...moreFields
}

fragment moreFields on Type {
  field2(x: $a) ## this would carry the value passed into the operation
}

This also makes Fragment-Arguments easier to reason about for me atleast as it's either the Variables passed into the definition or from the operation itself.

[benjie]

I haven't got as far as section 6 yet.

[benjie]

@JoviDeCroock Ah! I thought we landed on the other side with this one, this does simplify things significantly because we can look at fragments in isolation 👍

[benjie]

Great work Jovi! Here's a few comments, but this was only a quick review so it's not exhaustive.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: JoviDeCroock / name: Jovi De Croock (a868d04, 210a814, 2e541bb, 6e91f98, c46e706, 16e8986, e027a4f, b4f3c5d, f5b3061, 44aea8b, 562edd8, 548163a, 2117038, 61d2117, 815952a, 492c556, df27219, 21e3881, 3516ca1, b60feac, 8be0561, f306736, 94a0027)
• ✅ login: mjmahone / name: Matt Mahoney (4269caf)

[benjie]

@JoviDeCroock did you rebase on a different computer? Often this happens when git is configured to use a different email address on one computer than another, so some commits don't match EasyCLA whilst others do.

[JoviDeCroock]

@benjie I did not, the commits it points at are code-suggestions I applied so maybe it's not good at the co-authored ones? Not sure 😅 can rebase to fix it

[benjie]

^ Trying to get EasyCLA to re-evaluate.