Implement deterministic dask tokenization. (#320)

* Test deterministic dask tokenization.

* Implement __dask_tokenize__.

* Add missing test dependency.

* Apply black.

requirements/tests.txt

@@ -1,3 +1,4 @@
+dask[array]
 pytest>=3.5
 pytest-black
-pytest-cov
+pytest-cov

sparse/_coo/core.py

@@ -284,6 +284,14 @@ def __setstate__(self, state):
         self.coords, self.data, self.shape, self.fill_value = state
         self._cache = None
 
+    def __dask_tokenize__(self):
+        "Produce a deterministic, content-based hash for dask."
+        from dask.base import normalize_token
+
+        return normalize_token(
+            (type(self), self.coords, self.data, self.shape, self.fill_value)
+        )
+
     def copy(self, deep=True):
         """Return a copy of the array.
 

sparse/tests/test_dask_interop.py

@@ -0,0 +1,11 @@
+from dask.base import tokenize
+import sparse
+
+
+def test_deterministic_token():
+    a = sparse.COO(data=[1, 2, 3], coords=[10, 20, 30], shape=(40,))
+    b = sparse.COO(data=[1, 2, 3], coords=[10, 20, 30], shape=(40,))
+    assert tokenize(a) == tokenize(b)
+    # One of these things is not like the other....
+    c = sparse.COO(data=[1, 2, 4], coords=[10, 20, 30], shape=(40,))
+    assert tokenize(a) != tokenize(c)