Add Support for Reading TILEDB_BLOB (#1159)

Author: nguyenv

HISTORY.md

@@ -3,6 +3,7 @@
 ## API Changes
 * Use `bool` instead of `uint8` for Boolean dtype in `dataframe_.py` [#1154](https://github.com/TileDB-Inc/TileDB-Py/pull/1154)
 * Support QueryCondition OR operator [#1146](https://github.com/TileDB-Inc/TileDB-Py/pull/1146)
+* Support `TILEDB_BLOB` dtype [#1159](https://github.com/TileDB-Inc/TileDB-Py/pull/1159)
 
 ## Bug Fixes
 * Fix error where passing a `Context` to `Group` would segfault intermittenly [#1165](https://github.com/TileDB-Inc/TileDB-Py/pull/1165)

tiledb/cc/common.cc

@@ -41,8 +41,8 @@ std::unordered_map<tiledb_datatype_t, std::string> _tdb_to_np_name_dtype = {
     {TILEDB_TIME_FS, "m8[fs]"},
     {TILEDB_TIME_AS, "m8[as]"},
 #endif
-#if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 7
-    {TILEDB_BLOB, "bytes"},
+#if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 9
+    {TILEDB_BLOB, "byte"},
 #endif
 #if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 10
     {TILEDB_BOOL, "bool"},
@@ -85,9 +85,6 @@ std::unordered_map<std::string, tiledb_datatype_t> _np_name_to_tdb_dtype = {
     {"timedelta64[fs]", TILEDB_TIME_FS},
     {"timedelta64[as]", TILEDB_TIME_AS},
 #endif
-#if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 7
-    {"bytes", TILEDB_BLOB},
-#endif
 #if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 10
     {"bool", TILEDB_BOOL},
 #endif

tiledb/core.cc

@@ -221,9 +221,9 @@ py::dtype tiledb_dtype(tiledb_datatype_t type, uint32_t cell_val_num) {
     case TILEDB_TIME_AS:
       return py::dtype("m8[as]");
 #endif
-#if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 7
+#if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 9
     case TILEDB_BLOB:
-      return py::dtype("bytes");
+      return py::dtype("byte");
 #endif
 #if TILEDB_VERSION_MAJOR >= 2 && TILEDB_VERSION_MINOR >= 10
     case TILEDB_BOOL:

tiledb/libtiledb.pyx

@@ -111,6 +111,9 @@ _tiledb_dtype_to_numpy_typeid_convert ={
     TILEDB_CHAR: np.NPY_STRING,
     TILEDB_STRING_UTF8: np.NPY_UNICODE,
 }
+IF LIBTILEDB_VERSION_MAJOR >= 2:
+    IF LIBTILEDB_VERSION_MINOR >= 9:
+        _tiledb_dtype_to_numpy_typeid_convert[TILEDB_BLOB] = np.NPY_BYTE
 IF LIBTILEDB_VERSION_MAJOR >= 2:
     IF LIBTILEDB_VERSION_MINOR >= 10:
         _tiledb_dtype_to_numpy_typeid_convert[TILEDB_BOOL] = np.NPY_BOOL
@@ -131,6 +134,9 @@ _tiledb_dtype_to_numpy_dtype_convert = {
     TILEDB_STRING_ASCII: np.bytes_,
     TILEDB_STRING_UTF8: np.dtype('U1'),
 }
+IF LIBTILEDB_VERSION_MAJOR >= 2:
+    IF LIBTILEDB_VERSION_MINOR >= 9:
+        _tiledb_dtype_to_numpy_dtype_convert[TILEDB_BLOB] = np.byte
 IF LIBTILEDB_VERSION_MAJOR >= 2:
     IF LIBTILEDB_VERSION_MINOR >= 10:
         _tiledb_dtype_to_numpy_dtype_convert[TILEDB_BOOL] = np.bool_

tiledb/tests/test_group.py

@@ -145,15 +145,14 @@ def values_equal(lhs, rhs):
         assert values_equal(grp.meta["str"], str_data)
         grp.close()
 
-        # NOTE uncomment when deleting is "synced" in core
-        # grp.open("w")
-        # del grp.meta["int"]
-        # grp.close()
-
-        # grp = tiledb.Group(grp_path, "r")
-        # assert len(grp.meta) == 2
-        # assert "int" not in grp.meta
-        # grp.close()
+        grp.open("w")
+        del grp.meta["int"]
+        grp.close()
+
+        grp = tiledb.Group(grp_path, "r")
+        assert len(grp.meta) == 2
+        assert "int" not in grp.meta
+        grp.close()
 
     def test_group_members(self, capfd):
         grp_path = self.path("test_group_members")