Implement DropData operation.

[martinmr]

This new operation acts as a DropAll except the schema and types are
left intact. Useful when the schema and types will be resued and/or the
schema is large enough that there will be a delay recreating it.

This feature is motivated by
#3236.

---

This change is 

[martinmr]

I copied the pb.go file from dgo in the meantime. I will revert those changes and properly version dgo once dgraph-io/dgo#53 is merged.

[srfrog]

IMO this should be an enum and deprecate drop_all. Have you considered that?
e.g.,

enum DropOp {
  NONE = 0;
  ALL = 1;
  SCHEMA = 2;
  DATA = 3;
}
DropOp drop_op = 8;

Reviewable status: 0 of 12 files reviewed, all discussions resolved

[martinmr]

I don't want to deprecate existing features because of this, which will probably be mostly used during tests.

Reviewable status: 0 of 12 files reviewed, all discussions resolved

[manishrjain]

Same comment as @srfrog, which I arrived independently below. It does seem like an enum would make sense.

Reviewed 12 of 12 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @martinmr)

---

gql/mutation.go, line 39 at r1 (raw file):

	Del      []*api.NQuad
	DropAll  bool
	DropData bool

Does it make sense to make Drop an enum. So, it can tell what to drop? I can think of other scenarios in the future, where you'd want to drop only the types or only the schema or other things.

[martinmr]

Reviewable status: 4 of 26 files reviewed, 1 unresolved discussion (waiting on @manishrjain)

---

gql/mutation.go, line 39 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Does it make sense to make Drop an enum. So, it can tell what to drop? I can think of other scenarios in the future, where you'd want to drop only the types or only the schema or other things.

Done. I used oneof instead of an enum. It has the advantage that it will force only one of the values inside to be set and it let's us keep the current HTTP interface without having to make any changes.

[martinmr]

Using enums now but kept drop_all and drop_attr for now for backwards compatibility.

Reviewable status: 5 of 12 files reviewed, 1 unresolved discussion (waiting on @manishrjain)

Addressed comments

[manishrjain]

Got a few comments. Address before merging.

Reviewed 8 of 8 files at r2.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @manishrjain and @martinmr)

---

edgraph/server.go, line 319 at r2 (raw file):

	// if it lies on some other machine. Let's get it for safety.
	m := &pb.Mutations{StartTs: State.getTimestamp(false)}
	if op.DropAll || op.DropOp == api.Operation_ALL {

Given how often this is being used, might make sense to have a helper func isDropAll(op).

---

worker/mutation.go, line 479 at r2 (raw file):

	if src.DropOp == pb.Mutations_ALL {
		for _, gid := range groups().KnownGroups() {

There's no need to copy the loop.

if src.DropOp == all || src.DropOp == data {
 ...
 mu.DropOp = src.DropOp
}

[martinmr]

Reviewable status: 7 of 11 files reviewed, 3 unresolved discussions (waiting on @manishrjain)

---

edgraph/server.go, line 319 at r2 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Given how often this is being used, might make sense to have a helper func isDropAll(op).

Done.

---

worker/mutation.go, line 479 at r2 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

There's no need to copy the loop.

if src.DropOp == all || src.DropOp == data {
 ...
 mu.DropOp = src.DropOp
}

Done.