Introspection: MongoDB: Warnings mention Native Types instead of PSL types (e.g. Document instead of Json)

[Jolg42]

Now that we use Json for properties with different types, we might want to tweak the language used in the comment and CLI warning:

Example 1, (since Json is picked as default with engine PR)

      // *** WARNING ***
      // 
      // The following fields had data stored in multiple types. Either use Json or normalize data to the wanted type.
      // - Model "users", field: "numberOrString1", chosen data type: "Document"
      // - Type "UsersHobbies", field: "numberOrString2", chosen data type: "Document"
      // - Type "UsersHobbiesObjects", field: "numberOrString3", chosen data type: "Document"

Here for the field field: "numberOrString3" the chosen data type is Document but the type in the schema.prisma file will be Json

See db pull test

prisma/packages/migrate/src/__tests__/DbPull.test.ts

Lines 752 to 807 in 755d81b

	test('introspection --print --composite-type-depth=-1 (no existing models)', async () => {
	ctx.fixture('schema-only-mongodb')
	const introspect = new DbPull()
	const result = introspect.parse(['--schema=./prisma/no-model.prisma', '--print', '--composite-type-depth=-1'])
	await expect(result).resolves.toMatchInlineSnapshot(``)
	expect(ctx.mocked['console.log'].mock.calls.join('\n')).toMatchInlineSnapshot(`
	generator client {
	provider = "prisma-client-js"
	previewFeatures = ["mongoDb"]
	}
	datasource my_db {
	provider = "mongodb"
	url = env("TEST_MONGO_URI_MIGRATE")
	}
	type UsersHobbies {
	name String
	/// Multiple data types found: String: 50%, Int32: 50% out of 2 sampled entries
	numberOrString2 Int?
	objects UsersHobbiesObjects[]
	tags String[]
	}
	type UsersHobbiesObjects {
	name String
	/// Multiple data types found: String: 50%, Int32: 50% out of 2 sampled entries
	numberOrString3 Int
	tags String[]
	}
	model users {
	id String @id @default(auto()) @map("_id") @my_db.ObjectId
	admin Boolean
	email String
	hobbies UsersHobbies[]
	name String
	/// Multiple data types found: String: 50%, Int32: 50% out of 2 sampled entries
	numberOrString1 Int
	}
	// introspectionSchemaVersion: NonPrisma,
	`)
	expect(ctx.mocked['console.info'].mock.calls.join('\n')).toMatchInlineSnapshot(``)
	expect(ctx.mocked['console.error'].mock.calls.join('\n')).toMatchInlineSnapshot(`
	// *** WARNING ***
	//
	// The following fields had data stored in multiple types. The most common type was chosen. If loading data with a type that does not match the one in the data model, the client will crash. Please see the issue: https://github.com/prisma/prisma/issues/9654
	// - Model "users", field: "numberOrString1", chosen data type: "Int32"
	// - Type "UsersHobbies", field: "numberOrString2", chosen data type: "Int32"
	// - Type "UsersHobbiesObjects", field: "numberOrString3", chosen data type: "Int32"
	//
	`)
	})

Example 2 (Before the default changed to Json)

// *** WARNING ***
// 
// The following fields had data stored in multiple types. The most common type was chosen. If loading data with a type that does not match the one in the data model, the client will crash. Please see the issue: https://github.com/prisma/prisma/issues/9654
// - Model "shipwrecks", field: "depth", chosen data type: "Int32"
// 

Here for the field field: "depth" the chosen data type is Int32 but the type in the schema.prisma file will be Int

See https://github.com/prisma/introspection-ci/blob/4c4b4f7aeaa5d0eea96d4a3d00630c8a52b3e3f9/introspection-analysis/output/mongodb_public/sample_geospatial_stderr.log

Internal thread https://prisma-company.slack.com/archives/C4GCG53BP/p1647246722174479

MongoDB data

[
  {
    "name": "Jane Smith",
    "email": "1@prisma.io",
    "admin": true,
    "numberOrString1": 1234,
    "hobbies": [
      {
        "name": "Swimming",
        "tags": ["tag"],
        "numberOrString2": 1234,
        "objects": [
          {
            "name": "Object 1",
            "tags": ["tag"],
            "numberOrString3": 1234
          }
        ]
      }
    ]
  },
  {
    "name": "John Smith",
    "email": "2@prisma.io",
    "admin": false,
    "numberOrString1": "567",
    "hobbies": [
      {
        "name": "Running",
        "tags": ["tag"],
        "numberOrString2": "567",
        "objects": [
          {
            "name": "Object 1",
            "tags": ["tag"],
            "numberOrString3": "567"
          }
        ]
      },
      {
        "name": "Coding",
        "tags": ["tag"]
      }
    ]
  }
]

Note: introspection-ci results when default changed to Json https://github.com/prisma/introspection-ci/commit/b362f3e9ab0361a9b9efa96c382359303273cc4d

[janpio]

We can probably remove the "chosen data type" information from the CLI warning completely. The introduction of the list can mention that already and can be improved to be more readable. (We can also create documentation how to proceed from there to fix the data and be able to change the schema to one of the existing data types.)

Could look like this:

      // *** WARNING ***
      // 
      // The following fields had data stored in multiple types and were written as `Json` to the schema, and a comment added with type information. 
      // Go to https://pris.ly/d/mongodb-field-multiple-types to learn how to proceed from here.
      // - Model "users", field: "numberOrString1"
      // - Type "UsersHobbies", field: "numberOrString2"
      // - Type "UsersHobbiesObjects", field: "numberOrString3"

[Jolg42]

That would be very easy to change

[janpio]

I think that would fix the concrete problem in the CLI.
It would still leave the MongoDB types in the comment we render to the schema, but we should track that separately in a new issue (and also work on it, but independently)

[janpio]

Ok, seems we went with changing the actual returned data type string instead:

// The following fields had data stored in multiple types. Either use Json or normalize data to the wanted type.
- // - Type "MoviesImdb", field: "rating", chosen data type: "Document"
- // - Type "MoviesImdb", field: "votes", chosen data type: "Document"
+ // - Type "MoviesImdb", field: "rating", chosen data type: "Json"
+ // - Type "MoviesImdb", field: "votes", chosen data type: "Json"