Index out of bounds when nonempty domain is empty string (#1182)

* Add Test to Ensure Empty String's NED is `(None, None)`

Author: robertbindar

HISTORY.md

@@ -3,6 +3,9 @@
 ## Improvements
 * `setup.py` retrieves core version by using `ctypes` to call `tiledb_version` rather than parsing `tiledb_version.h` [#1191](https://github.com/TileDB-Inc/TileDB-Py/pull/1191)
 
+## Bug Fixes
+* Set nonempty domain to `(None, None)` for empty string [#1182](https://github.com/TileDB-Inc/TileDB-Py/pull/1182)
+
 ## API Changes
 * Support `QueryCondition` for dense arrays [#1198](https://github.com/TileDB-Inc/TileDB-Py/pull/1198)
 * Querying dense array with `[:]` returns shape that matches nonempty domain, consistent with `.df[:]` and `.multi_index[:]` [#1199](https://github.com/TileDB-Inc/TileDB-Py/pull/1199)

tiledb/libtiledb.pyx

@@ -3837,7 +3837,7 @@ cdef class Array(object):
                     continue
 
                 buf_dtype = 'S'
-                start_buf = np.empty(end_size, 'S' + str(start_size))
+                start_buf = np.empty(start_size, 'S' + str(start_size))
                 end_buf = np.empty(end_size, 'S' + str(end_size))
                 start_buf_ptr = np.PyArray_DATA(start_buf)
                 end_buf_ptr = np.PyArray_DATA(end_buf)
@@ -3854,7 +3854,11 @@ cdef class Array(object):
                     _raise_ctx_err(ctx_ptr, rc)
                 if is_empty:
                     return None
-                results.append((start_buf.item(0), end_buf.item(0)))
+
+                if start_size > 0 and end_size > 0:
+                    results.append((start_buf.item(0), end_buf.item(0)))
+                else:
+                    results.append((None, None))
             else:
                 rc = tiledb_array_get_non_empty_domain_from_index(
                         ctx_ptr, array_ptr, dim_idx, start_buf_ptr, &is_empty

tiledb/tests/test_libtiledb.py

@@ -5,6 +5,7 @@
 import pickle
 import random
 import re
+from tkinter import NONE
 import urllib
 import subprocess
 import sys
@@ -1022,6 +1023,21 @@ def test_nonempty_domain_scalar(self):
             assert isinstance(ned[0][0], int)
             assert isinstance(ned[0][1], int)
 
+    def test_nonempty_domain_empty_string(self):
+        uri = self.path("test_nonempty_domain_empty_string")
+        dims = [tiledb.Dim(dtype="ascii")]
+        schema = tiledb.ArraySchema(tiledb.Domain(dims), sparse=True)
+        tiledb.Array.create(uri, schema)
+
+        with tiledb.open(uri, "r") as A:
+            assert_array_equal(A.nonempty_domain(), ((None, None),))
+
+        with tiledb.open(uri, "w") as A:
+            A[""] = None
+
+        with tiledb.open(uri, "r") as A:
+            assert_array_equal(A.nonempty_domain(), ((b"", b""),))
+
     def test_create_array_overwrite(self):
         uri = self.path("test_create_array_overwrite")
         dims = tiledb.Dim(domain=(0, 10), dtype=np.int64)