Wrap delta filter

x-ref TileDB-Inc/TileDB#3969

Author: ihnorton

tiledb/__init__.py

@@ -53,6 +53,7 @@
     Bzip2Filter,
     ChecksumMD5Filter,
     ChecksumSHA256Filter,
+    DeltaFilter,
     DictionaryFilter,
     DoubleDeltaFilter,
     Filter,

tiledb/cc/enum.cc

@@ -87,6 +87,7 @@ void init_enums(py::module &m) {
       .value("LZ4", TILEDB_FILTER_LZ4)
       .value("RLE", TILEDB_FILTER_RLE)
       .value("BZIP2", TILEDB_FILTER_BZIP2)
+      .value("DELTA", TILEDB_FILTER_DELTA)
       .value("DOUBLE_DELTA", TILEDB_FILTER_DOUBLE_DELTA)
       .value("BIT_WIDTH_REDUCTION", TILEDB_FILTER_BIT_WIDTH_REDUCTION)
       .value("BITSHUFFLE", TILEDB_FILTER_BITSHUFFLE)

tiledb/filter.py

@@ -249,6 +249,35 @@ def _attrs_(self):
         return {}
 
 
+class DeltaFilter(CompressionFilter):
+    """
+    Filter that compresses using run-length encoding (RLE).
+
+    :param level: -1 (default) sets the compressor level to the default level as specified in TileDB core. Otherwise, sets the compressor level to the given value.
+    :type level: int
+
+    **Example:**
+
+    >>> import tiledb, numpy as np, tempfile
+    >>> with tempfile.TemporaryDirectory() as tmp:
+    ...     dom = tiledb.Domain(tiledb.Dim(domain=(0, 9), tile=2, dtype=np.uint64))
+    ...     a1 = tiledb.Attr(name="a1", dtype=np.int64,
+    ...                      filters=tiledb.FilterList([tiledb.RleFilter()]))
+    ...     schema = tiledb.ArraySchema(domain=dom, attrs=(a1,))
+    ...     tiledb.DenseArray.create(tmp + "/array", schema)
+
+    """
+
+    def __init__(self, level: int = -1, ctx: Optional[Ctx] = None):
+        if not isinstance(level, int):
+            raise ValueError("`level` argument must be a int")
+
+        super().__init__(lt.FilterType.DELTA, level, ctx)
+
+    def _attrs_(self):
+        return {}
+
+
 class DoubleDeltaFilter(CompressionFilter):
     """
     Filter that performs double-delta encoding.
@@ -737,6 +766,7 @@ class FilterList(CtxMixin, lt.FilterList):
         lt.FilterType.LZ4: LZ4Filter,
         lt.FilterType.BZIP2: Bzip2Filter,
         lt.FilterType.RLE: RleFilter,
+        lt.FilterType.DELTA: DeltaFilter,
         lt.FilterType.DOUBLE_DELTA: DoubleDeltaFilter,
         lt.FilterType.BIT_WIDTH_REDUCTION: BitWidthReductionFilter,
         lt.FilterType.BITSHUFFLE: BitShuffleFilter,

tiledb/tests/test_filters.py

@@ -15,6 +15,7 @@
     tiledb.LZ4Filter,
     tiledb.RleFilter,
     tiledb.Bzip2Filter,
+    tiledb.DeltaFilter,
     tiledb.DoubleDeltaFilter,
     tiledb.DictionaryFilter,
     tiledb.BitWidthReductionFilter,
@@ -157,3 +158,23 @@ def test_float_scaling_filter(self, factor, offset, bytewidth):
 
             # TODO compute the correct tolerance here
             assert_allclose(data, A[:][""], rtol=1, atol=1)
+
+    def test_delta_filter(self):
+        path = self.path("test_delta_filter")
+
+        dom = tiledb.Domain(tiledb.Dim(name="row", domain=(0, 9), dtype=np.uint64))
+
+        filter = tiledb.DeltaFilter()
+
+        attr = tiledb.Attr(dtype=np.int64, filters=tiledb.FilterList([filter]))
+        schema = tiledb.ArraySchema(domain=dom, attrs=[attr], sparse=False)
+        tiledb.Array.create(path, schema)
+
+        data = np.random.randint(0, 10_000_000, size=10)
+
+        with tiledb.open(path, "w") as A:
+            A[:] = data
+
+        with tiledb.open(path) as A:
+            res = A[:]
+            assert_array_equal(res, data)