Add "encoding" attribute to string substr op, which controls how each "character" is treated:

  * BYTE: Position & length refer to bytes in the string.  (Default)
  * UTF8: The string is interpreted as UTF-8 encoded Unicode code points, and position & length are treated relative to them.

RELNOTES: Add option to get substring using Unicode characters
PiperOrigin-RevId: 215773373

2 files changed, 11 insertions, 7 deletions

diff --git a/tensorflow/core/api_def/base_api/api_def_Substr.pbtxt b/tensorflow/core/api_def/base_api/api_def_Substr.pbtxt
index 5246090ab3..fe0fcc9508 100644
--- a/tensorflow/core/api_def/base_api/api_def_Substr.pbtxt
+++ b/tensorflow/core/api_def/base_api/api_def_Substr.pbtxt
@@ -18,6 +18,16 @@ END
 Scalar defining the number of characters to include in each substring
 END
   }
+  attr {
+    name: "unit"
+    description: <<END
+The unit that is used to create the substring.  One of: `"BYTE"` (for
+defining position and length by bytes) or `"UTF8_CHAR"` (for the UTF-8
+encoded Unicode code points).  The default is `"BYTE"`. Results are undefined if
+`unit=UTF8_CHAR` and the `input` strings do not contain structurally valid
+UTF-8.
+END
+  }
   out_arg {
     name: "output"
     description: <<END
diff --git a/tensorflow/core/api_def/python_api/api_def_Substr.pbtxt b/tensorflow/core/api_def/python_api/api_def_Substr.pbtxt
index 4778d7927c..4fb9ee56e9 100644
--- a/tensorflow/core/api_def/python_api/api_def_Substr.pbtxt
+++ b/tensorflow/core/api_def/python_api/api_def_Substr.pbtxt
@@ -1,10 +1,4 @@
 op {
   graph_op_name: "Substr"
-  endpoint {
-    name: "strings.substr"
-  }
-  endpoint {
-    name: "substr"
-    deprecated: true
-  }
+  visibility: HIDDEN
 }