Right now I wrote a custom parser that creates ESTree compatible tree.
The main difference here the kind => type and description => comments.
I think it might be better to do:

1. parse with graphql-js (or graphql-tools)
2. visit all nodes and convert to ESTree syntax, also include type-info from TypeInfo when schema is available
3. return ESTree nodes with complete type-info and also the original node.

Something like that:

type ESTreeNode = Omit<DocumentNode, "kind", "description"> & { type: string, description: description, rawNode: DocumentNode, typeInfo: TypeInfo  }

This way, we can force things like:

type Mutation {
   doSomething: Result
}

union Result = OkResult | ErrorResult

type OkResult {
  updated: Something
  effectedSomethingElse: SomethingElse
}

type ErrorResult {
  message: String!
  reason: String
}

To prevent:

enum Test {
  Value
  value
  ValuE
}

Source: https://github.com/ilyavolodin/graphql-eslint/blob/master/docs/rules/description-style.md

Related: ilyavolodin/graphql-eslint#3

To make it easier to write tests, based on the RuleTester from eslint.

We need to check how to parse those, since we need to make sure # is being handled correctly in the parser.

Idea: find similar things and suggest refactorings. Is it possible?

E.g. find similar types and suggest to extract them into interface, find similar sets of fields in operations and suggest to extract them into fragments, find similar input values and suggest to extract them into variables.

Or vice versa: find fragments and variables that used only in one place and suggest to inline them.

Should support nested and array use cases

consider

fragment a on A {
  id
  aField
}

# import "./a.fragment.graphql"
query AQuery {
  as { ...a }
}

Field "id" must be selected when it's available on a type. Please make sure to include it in your selection set!eslint@graphql-eslint/require-id-when-available

require-id-when-available should allow fields which contain (only?) fragment spreads. Bonus points if the fragment themselves validate as having id

Source: https://github.com/ilyavolodin/graphql-eslint/blob/master/docs/rules/input-name.md

And as idea, suggest to remove deprecated things after deprecated date.

Related: https://relay.dev/docs/en/graphql-server-specification.html#schema

Source: https://github.com/ilyavolodin/graphql-eslint/blob/master/docs/rules/prettier.md

To make sure to return a valid state

• Avoid get / set prefix

Old versions of graphql allowed # syntax for descriptions. It's no longer supported, and the preferred way should be using """

• changesets
• canary releases
• pr comments

Running graphql-eslint on our projects is slower than expected. I did some profiling and noticed that loadSchemaSync() is responsible for several seconds on execution time. I added some naive memoization to that code (just based on config.schema) and the performance improved.

Do you think this memoization is something we could support? Would it cause any issues? Should there be an option to enable the memoization?

I can make a PR if needed but I wanted to make sure it wouldn't break any assumption. If memoization can be based on config.schema it should be straightforward but I notice that the config is being spread into loadSchemaSync() options and changes every time. Is there any overlap between ParserOptions and LoadSchemaOptions? It wasn't obvious to me.

https://github.com/dotansimha/graphql-eslint/blob/15176710527dd85621ebebde3cd0d0d30b7ca584/packages/plugin/src/parser.ts#L93-L94

Thanks!

The original methods from graphql-js will only accept the "real" ast types (and not ESTree), maybe we can create a converter util that does that.

naming conventions

Source: https://github.com/ilyavolodin/graphql-eslint/blob/master/docs/rules/naming-convention.md

Hi. I was trying out this eslint plugin today and I am facing this error in all my GQL files both when linting via the CLI as well as the VSCode extension.

This error is thrown for every GQL file I use.

I tried to isolate the problem by removing all other eslint rules and just having this plugin around and still had this issue.

My graphql-config is configured correctly cause both the codegen and the VSCode GraphQL extension which uses it are working correctly. I am using multiple projects with both remote and local schema which works well with codegen and extension.

I also noticed other people reporting similar error but with apollo eslint rather:

apollographql/eslint-plugin-graphql#287
apollographql/eslint-plugin-graphql#270 (comment)

Any clues? Thanks.

My .eslintrc file:

{
  "overrides": [
    {
      "files": ["*.gql"],
      "parser": "@graphql-eslint/eslint-plugin",
      "plugins": ["@graphql-eslint"],
      "rules": {
        "avoid-operation-name-prefix": [
          "error",
          { "keywords": "get" }
        ],
        "input-name": [
          "error",
          { "checkInputType": true }
          ],
          "description-style": [
          "error",
          { "style": "block" }
          ],
          "naming-convention": [
          "error",
          {
            "FieldDefinition": "camelCase",
            "ObjectTypeDefinition": "PascalCase"
          }
          ],
          "no-anonymous-operations": ["error"],
          "require-deprecation-reason": "error",
          "no-case-insensitive-enum-values-duplicates": ["error"],
          "no-operation-name-suffix": ["error"],
          "require-id-when-available": ["error",
          {
            "fieldName": "id"
          }
          ],
          "validate-against-schema": ["error"],
          "require-description": [
            "error",
            { "on": ["ObjectTypeDefinition", "FieldDefinition"] }
          ]
      }
    }
  ]
}

My package.json

"name":"...",
"version":"...",
"scripts": {
    "lint": "eslint --ext=.gql ./packages",
}
"devDependencies": {
    "@graphql-codegen/add": "^2.0.1",
    "@graphql-codegen/cli": "^1.17.10",
    "@graphql-codegen/typed-document-node": "^1.17.9",
    "@graphql-codegen/typescript": "^1.17.10",
    "@graphql-codegen/typescript-operations": "^1.17.8",
    "@graphql-codegen/typescript-resolvers": "^1.17.10",
    "@graphql-eslint/eslint-plugin": "^0.4.1",
    "@graphql-typed-document-node/core": "^3.1.0",
    "@types/body-parser": "^1.19.0",
    "@types/express": "^4.17.8",
    "@types/node": "^14.11.8",
    "@types/qs": "6.9.5",
    "@typescript-eslint/eslint-plugin": "^4.4.1",
    "@typescript-eslint/parser": "^4.4.1",
    "concurrently": "^5.3.0",
    "eslint": "^7.11.0",
    "globby": "^11.0.1",
    "morgan": "^1.10.0",
    "nodemon": "^2.0.5",
    "plop": "^2.7.4",
    "rimraf": "^3.0.2",
    "typedoc": "^0.19.2",
    "typescript": "4.0.3",
    "utility-types": "^3.10.0"
  },
  "dependencies": {
    "graphql": "^15.4.0",
    "graphql-modules": "^1.0.0-alpha.2"
  }

CC: @dotansimha

I'm trying to add this plugin to my project, I followed the standard configuration in the documentation:

But am getting the following error in VSCode's ESLint output:

[Info  - 9:40:12 AM] ESLint server is starting
[Info  - 9:40:12 AM] ESLint server running in node v12.14.1
[Info  - 9:40:12 AM] ESLint server is running.
[Info  - 9:42:43 AM] ESLint library loaded from: /Users/peterp/x/redwoodjs/example-invoice/node_modules/eslint/lib/api.js
[Info  - 9:42:56 AM] ENOTDIR: not a directory, stat '/Users/peterp/x/redwoodjs/example-invoice/web/src/index.js/0_/Users/peterp/x/redwoodjs/example-invoice/web/src/index.js' Occurred while linting /Users/peterp/x/redwoodjs/example-invoice/web/src/index.js/0_/Users/peterp/x/redwoodjs/example-invoice/web/src/index.js:1
[Info  - 9:42:56 AM] ENOTDIR: not a directory, stat '/Users/peterp/x/redwoodjs/example-invoice/web/src/pages/InvoicePage/subcomponents/InvoiceCell/InvoiceCell.js/0_document.graphql' Occurred while linting /Users/peterp/x/redwoodjs/example-invoice/web/src/pages/InvoicePage/subcomponents/InvoiceCell/InvoiceCell.js/0_document.graphql:0

I've pushed up a branch here: https://github.com/redwoodjs/example-invoice/pull/39/files#diff-b9cfc7f2cdf78a7f4b91a753d10865a2

https://github.com/ilyavolodin/graphql-eslint/blob/master/docs/rules/require-description.md

To adjust configuration (types and so on).

See https://graphql-rules.com/rules/mutation-payload-query

This one is a bit tricky, since it requires to run multiple parsers on files, need to check that ESLint supports that.
Also, we need to make sure to mark ranges and offsets correctly within the file.

See https://github.com/cjoudrey/graphql-schema-linter#built-in-rules

It should suggest remove duplicate fields, input fields, arguments, etc.

As an option, it can suggest to use GraphQL aliases for duplicate fields instead of removing them.

Not sure how to do this yet, but we need to extend the ESTree nodes with the typeinfo in order to make it simpler to get type details while visiting the tree.

What about GitHub/CI action/app like GraphQL Inspector?

As an idea, it could be called GraphQL Doctor.

GraphQL Inspector is about detecting changes and preventing breaking changes in GraphQL schema.

GraphQL Doctor is about validating and force best design practices in GraphQL schema.

• How to use
• How to use with other parsers together
• How to write custom rules
• How to visualize in VSCode

Because of the regexes used, numbers can't be used with UPPER_CASE and snake_case formats :

https://github.com/dotansimha/graphql-eslint/blob/ee9500dadffe0bd5e5957f0b852072a439132416/packages/plugin/src/rules/naming-convention.ts#L4-L9

For instance, my_s3xy_string would fail snake_case naming convention, as would MY_S3XY_STRING for the UPPER_CASE one. We don't use snake_case, but we have a bunch of uppercase enums with numbers in them, so spotting this problem was pretty obvious in our codebase.

Thanks a lot for the effort put into this library, looking forward to seeing it grow among the community 💪

It seems like validate fails, since something in the original node is being modified (probably during estree building).
I assumes it's something with the visitor flow that causes the document to be changed.

There is conflict with the rule input-name where it will return error on connection fields, node id input etc.

Some rules might needs access to other loaded operations - maybe we can have a util on the context level for getting a fragment by name or something like that.

This might be done using ESLint as core, since we need to make sure to load the same files ESLint loaded (as defined in the configuration - exclude, extension and so on)