fix(Query): fix cascade with pagination

[minhaj-shakeel]

Fixes DGRAPH-1901.
This PR fixes the behavior of the @cascade directive with pagination arguments.

---

This change is 

[pawanrawal]

Reviewable status: 0 of 2 files reviewed, 5 unresolved discussions (waiting on @manishrjain, @minhaj-shakeel, and @vvbalaji-dgraph)

---

query/query.go, line 582 at r1 (raw file):

		}

		// Remove pagination arguments from the query if @cascade is mentioned since

Also check what happens if a sort is applied and no first is applied, do we apply a first of 1000 automatically? Where does that happen?

---

query/query.go, line 818 at r1 (raw file):

	// Remove pagination arguments from the query if @cascade is mentioned since
	// pagination will be applied post processing the data.
	if len(args.Cascade.Fields) > 0 {

I see this reused, can we make a function for this?

---

query/query.go, line 2170 at r1 (raw file):

			if parent == nil {
				// I'm root. We reach here if root had a function.
				newDestUidList := &pb.List{Uids: make([]uint64, 0, len(sg.DestUIDs.Uids))}

Add a comment about why this copy is being done. Say that DestUids for this level become the SrcUids for the next level. In updateUidMatrix with cascade we end up modifying the first list from the uidMatrix which ends up modifying the srcUids of the next level. So to avoid that we make a copy.

In fact, we only need to make a copy if Cascade is true for this level so add that optimization as well and do a copy only when Cascade has fields.

---

query/query.go, line 2847 at r1 (raw file):

			// first time at the root here.

			//Apply pagination at the root after @cascade.

// Apply

---

query/query0_test.go, line 493 at r1 (raw file):

	query := `
	{
		me(func: uid(0x01)) @cascade {

needs a bit more testing with three-level nodes and different scenarios.

[minhaj-shakeel]

Reviewable status: 0 of 2 files reviewed, 5 unresolved discussions (waiting on @manishrjain, @pawanrawal, and @vvbalaji-dgraph)

---

query/query.go, line 582 at r1 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

Also check what happens if a sort is applied and no first is applied, do we apply a first of 1000 automatically? Where does that happen?

Done. It still works fine.

---

query/query.go, line 818 at r1 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

I see this reused, can we make a function for this?

Done.

---

query/query.go, line 2170 at r1 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

Add a comment about why this copy is being done. Say that DestUids for this level become the SrcUids for the next level. In updateUidMatrix with cascade we end up modifying the first list from the uidMatrix which ends up modifying the srcUids of the next level. So to avoid that we make a copy.

In fact, we only need to make a copy if Cascade is true for this level so add that optimization as well and do a copy only when Cascade has fields.

Done.

---

query/query.go, line 2847 at r1 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

// Apply

Done.

---

query/query0_test.go, line 493 at r1 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

needs a bit more testing with three-level nodes and different scenarios.

Done.

[pawanrawal]

Reviewable status: 0 of 2 files reviewed, all discussions resolved (waiting on @manishrjain and @vvbalaji-dgraph)