Question: Database sandbox for tests

[martinthenth]

One problem I've had while working with Dgraph, is that Dgraph does not have a test sandbox like Ecto has for Postgres. Since Dgraph does not offer multiple databases in a single instance, I'm wondering how other developers are solving this problem.

To illustrate, when running tests: The test may create new data and fetch old data:

1. First run: Create node with external_id: "1", returns uid: 0x1, external_id: "1"
2. First run: Fetch node with external_id: "1", returns uid: 0x1, external_id: "1"
3. Second run: Create node with external_id: "1", returns uid: 0x2, external_id: "1"
4. Second run: Fetch node with external_id: "1", returns nodes with uid: 0x1, external_id: "1" and uid: 0x2, external_id: "1"

A workaround I that use, is to add first: 1 and sometimes orderasc: created_on to queries. This works, and makes tests pass using the new data; but adding code simply to make tests work is obviously bad practice.

How are others solving this?

[m0rt3nlund]

This is something I am also wondering about.. We run separate servers just for testing due to this...