[GitHub] [arrow] kiszk commented on a change in pull request #10317: ARROW-12713 [C++] String reverse kernel

[GitBox <gi...@apache.org>]

kiszk commented on a change in pull request #10317:
URL: https://github.com/apache/arrow/pull/10317#discussion_r633107773



##########
File path: cpp/src/arrow/compute/kernels/scalar_string.cc
##########
@@ -266,6 +266,54 @@ void EnsureLookupTablesFilled() {}
 
 #endif  // ARROW_WITH_UTF8PROC
 
+template <typename Type>
+struct AsciiReverse : StringTransform<Type, AsciiReverse<Type>> {
+  using Base = StringTransform<Type, AsciiReverse<Type>>;
+  using offset_type = typename Base::offset_type;
+
+  bool Transform(const uint8_t* input, offset_type input_string_ncodeunits,
+                 uint8_t* output, offset_type* output_written) {
+    uint8_t utf8_char_found = 0;
+    for (offset_type i = 0; i < input_string_ncodeunits; i++) {
+      // if a utf8 char is found, report to utf8_char_found
+      utf8_char_found |= input[i] & 0x80;
+      output[input_string_ncodeunits - i - 1] = input[i];
+    }
+    *output_written = input_string_ncodeunits;
+    return utf8_char_found == 0;
+  }
+};
+
+/*
+ * size of a valid UTF8 can be determined by looking at leading 4 bits of BYTE1
+ * UTF8_BYTE_SIZE_LUT[0..7] --> pure ascii chars --> 1B length
+ * UTF8_BYTE_SIZE_LUT[8..11] --> internal bytes --> 1B length
+ * UTF8_BYTE_SIZE_LUT[12,13] --> 2B long UTF8 chars
+ * UTF8_BYTE_SIZE_LUT[14] --> 3B long UTF8 chars
+ * UTF8_BYTE_SIZE_LUT[15] --> 4B long UTF8 chars
+ */
+const std::array<uint8_t, 16> UTF8_BYTE_SIZE_LUT{1, 1, 1, 1, 1, 1, 1, 1,
+                                                 1, 1, 1, 1, 2, 2, 3, 4};
+
+template <typename Type>
+struct Utf8Reverse : StringTransform<Type, Utf8Reverse<Type>> {
+  using Base = StringTransform<Type, Utf8Reverse<Type>>;
+  using offset_type = typename Base::offset_type;
+
+  bool Transform(const uint8_t* input, offset_type input_string_ncodeunits,
+                 uint8_t* output, offset_type* output_written) {
+    offset_type i = 0;
+    while (i < input_string_ncodeunits) {
+      uint8_t offset = UTF8_BYTE_SIZE_LUT[input[i] >> 4];
+      offset_type stride = std::min(i + offset, input_string_ncodeunits);
+      std::copy(input + i, input + stride, output + input_string_ncodeunits - stride);
+      i += offset;
+    }
+    *output_written = input_string_ncodeunits;
+    return true;

Review comment:
       just curiosity: If we have all pure ascii chars, do we return true?




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org