Values are not sorted properly

[poonai]

What version of Dgraph are you using?

master

Have you tried reproducing the issue with the latest release?

Yes

What is the hardware spec (RAM, OS)?

Steps to reproduce the issue (command/config used to run Dgraph).

Just populate cluster with this data
̍https://github.com/dgraph-io/dgraph/blob/master/query/common_test.go#L269
and run the this query

{
		q(func:has(name), orderdesc: name, first:5) {
			 name
		 }
	 }

Expected behaviour and actual result.

actual result

{
  "data": {
    "q": [
      {
        "name": ""
      },
      {
        "name": ""
      },
      {
        "name": "Badger"
      },
      {
        "name": "name"
      },
      {
        "name": "expand"
      }
    ]
  },
  "extensions": {
    "server_latency": {
      "parsing_ns": 14640,
      "processing_ns": 375499571,
      "encoding_ns": 28760,
      "assign_timestamp_ns": 503321
    },
    "txn": {
      "start_ts": 98
    }
  }
}

Expected everything to be sorted

[akamos]

Hi!

I have a similar behaviour at Dgraph 1.1.0. I have data that contains values for name@de and name@ru predicates.
If I try to order these name string predicates with a language tag (orderasc:name@ru) I get different results from each time (its not determined behaviour). Sometimes its sorted correctly, by "ru", and sometimes not! If its not correctly sorted, it's sorted by name@de.

Check it out with following data and queries. Simply query, and the results differ from time to time.
The name and category_type predicates are indexed in schema.

Sample data:

{
  set {
    _:1 <name@de> "Hello World" .
    _:1 <name@ru> "Привет!" . 
    _:1 <category_type> "Test" .

    _:2 <name@de> "ZZZ A little test" .
    _:2 <name@ru> "Авиякомпания" .
    _:2 <category_type> "Test" .

    _:3 <name@de> "Xylophon" .
    _:3 <name@ru> "Книга" .
    _:3 <category_type> "Test" .
	}
}

My query:

{
  q(func:eq(category_type, "Test"),orderasc:name@ru)  {
    name@ru
    name@de
  }
}

[akamos]

Are there any updates to this issue? It's for me a critical point in my production system and I would continue to use Dgraph.

[arijitAD]

There are two PRs related to this issue. One is merged and other one is under review. I have linked both of them to this issue.

[akamos]

There are two PRs related to this issue. One is merged and other one is under review. I have linked both of them to this issue.

Thanks for the linked PR's!