fix(GraphQL): Link xids properly if there are duplicate xids within the same add request.

[pawanrawal]

Fixes GRAPHQL-641

See https://discuss.dgraph.io/t/residual-issue-linking-grandchild-in-mutation-with-custom-ids/9613 for the issue details.

If the same nested xid was used as a leaf node within an addType mutation, it was only linked the first time that it was used and not subsequently for later objects. This change fixes that by adding the linkage mutations and only removing the mutation which was creating the object with the xid again.

---

This change is 

Docs Preview:

[abhimanyusinghgaur]

Reviewed 1 of 2 files at r1.
Reviewable status: 1 of 2 files reviewed, 2 unresolved discussions (waiting on @MichaelJCompton and @pawanrawal)

---

graphql/resolve/add_mutation_test.yaml, line 2286 at r1 (raw file):

cond: "@if(eq(len(Comment16), 0) AND eq(len(Comment14), 0))"

we are creating relpy1 only if comment1 doesn't exist. What if comment1 existed already? In that case we won't be creating reply1 and the later mutation for comment2 won't know what is reply1.

If comment1 existed already would we error out the whole mutation? I feel like it should still work.

Also, we can add some e2e tests.

---

graphql/resolve/add_mutation_test.yaml, line 2309 at r1 (raw file):

 comment1

comment2

[pawanrawal]

Reviewed 10 of 10 files at r2.
Reviewable status: all files reviewed, 5 unresolved discussions (waiting on @abhimanyusinghgaur and @MichaelJCompton)

---

graphql/e2e/common/mutation.go, line 768 at r2 (raw file):

		ExpectedNumDeletedComments int
	}{
		"2nd level nodes don't exist but third level does": {

8 cases that I could think of

---

graphql/e2e/directives/schema_response.json, line 15 at r2 (raw file):

        {
            "predicate": "Category.name",
            "type": "string"

This is a just a move from inline to the file.

---

graphql/e2e/normal/schema_response.json, line 371 at r2 (raw file):

                {
                    "name": "Category.name"
                },

Moved this to a separate file to make formatting easy, I can just use my editor's formatting now. Its harder to do that when its a string.

---

graphql/resolve/add_mutation_test.yaml, line 2286 at r1 (raw file):

Previously, abhimanyusinghgaur (Abhimanyu Singh Gaur) wrote…

cond: "@if(eq(len(Comment16), 0) AND eq(len(Comment14), 0))"

we are creating relpy1 only if comment1 doesn't exist. What if comment1 existed already? In that case we won't be creating reply1 and the later mutation for comment2 won't know what is reply1.

If comment1 existed already would we error out the whole mutation? I feel like it should still work.

Also, we can add some e2e tests.

So we do create reply1 if comment2 doesn't exist as well now. Nice catch. If comment1 existed, then we won't link it to reply1. That is the behavior for deep mutations with xids. Added e2e tests for all the 8 cases possible here.

---

graphql/resolve/add_mutation_test.yaml, line 2309 at r1 (raw file):

Previously, abhimanyusinghgaur (Abhimanyu Singh Gaur) wrote…

 comment1

comment2

Done.

[abhimanyusinghgaur]

Reviewed 10 of 10 files at r2.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @MichaelJCompton and @pawanrawal)

---

graphql/e2e/common/mutation.go, line 768 at r2 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

8 cases that I could think of

Nice set of cases, they cover all the scenarios for this 3 level test.

---

graphql/e2e/directives/schema_response.json, line 15 at r2 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

This is a just a move from inline to the file.

we will finally get rid of those formatting errors which were hard to find in the previous JSON string 😌