Merge pull request #52 from ngoldbaum/fix-free

free output strings before setting them in string to unicode cast

Author: ngoldbaum

stringdtype/stringdtype/src/casts.c

@@ -213,6 +213,7 @@ unicode_to_string(PyArrayMethod_Context *context, char *const data[],
             return -1;
         }
         ss *out_ss = (ss *)out;
+        ssfree(out_ss);
         if (ssnewemptylen(out_num_bytes, out_ss) < 0) {
             gil_error(PyExc_MemoryError, "ssnewemptylen failed");
             return -1;

stringdtype/tests/test_stringdtype.py

@@ -1,6 +1,7 @@
 import concurrent.futures
 import os
 import pickle
+import string
 import tempfile
 
 import numpy as np
@@ -81,7 +82,6 @@ def test_unicode_casts(string_list):
     np.testing.assert_array_equal(arr, expected)
 
     arr = np.array(string_list, dtype=StringDType())
-
     np.testing.assert_array_equal(
         arr.astype("U8"), np.array(string_list, dtype="U8")
     )
@@ -95,6 +95,17 @@ def test_unicode_casts(string_list):
     )
 
 
+def test_additional_unicode_cast(string_list):
+    RANDS_CHARS = np.array(
+        list(string.ascii_letters + string.digits), dtype=(np.str_, 1)
+    )
+    arr = np.random.choice(RANDS_CHARS, size=10 * 100_000, replace=True).view(
+        "U10"
+    )
+    np.testing.assert_array_equal(arr, arr.astype(StringDType()))
+    np.testing.assert_array_equal(arr, arr.astype(StringDType()).astype("U10"))
+
+
 def test_insert_scalar(string_list):
     dtype = StringDType()
     arr = np.array(string_list, dtype=dtype)