Partial changes for VCF Character S->U

sgkit/io/vcf/vcf_reader.py

@@ -178,7 +178,7 @@ def _vcf_type_to_numpy(
     elif vcf_type == "Float":
         return "f4", FLOAT32_MISSING, FLOAT32_FILL
     elif vcf_type == "Character":
-        return "S1", CHAR_MISSING, CHAR_FILL
+        return "U1", CHAR_MISSING, CHAR_FILL
     elif vcf_type == "String":
         return "O", STR_MISSING, STR_FILL
     raise ValueError(
@@ -188,7 +188,7 @@ def _vcf_type_to_numpy(
 
 def _is_str_or_char(array: ArrayLike) -> bool:
     """Return True if the array is of string or character type"""
-    return array.dtype.kind in ("O", "S", "U")
+    return array.dtype.kind in ("O", "U")
 
 
 class VcfFieldHandler:

sgkit/io/vcf/vcf_writer_utils.py

@@ -11,6 +11,7 @@
     FLOAT32_MISSING_AS_INT32,
     INT_FILL,
     INT_MISSING,
+    STR_MISSING,
 )
 
 COLON = ord(":")
@@ -316,7 +317,8 @@ def vcf_values_to_byte_buf_size(a):
     elif a.dtype == np.float32:
         # values + separators
         return a.size * FLOAT32_BUF_SIZE + a.size
-    elif a.dtype.kind == "S":
+    elif a.dtype.kind == "U":
+        # NOTE! Assuming UTF-8 here?
         # values + separators
         return a.size * a.dtype.itemsize + a.size
     else:
@@ -502,8 +504,8 @@ def create_mask(arr):
         return np.all(arr == INT_MISSING, axis=axis)
     elif arr.dtype == np.float32:
         return np.all(arr.view("i4") == FLOAT32_MISSING_AS_INT32, axis=axis)
-    elif arr.dtype.kind == "S":
-        return np.all(arr == STR_MISSING_BYTE, axis=axis)
+    elif arr.dtype.kind == "U":
+        return np.all(arr == STR_MISSING, axis=axis)
     else:
         raise ValueError(f"Unsupported dtype: {arr.dtype}")