bpo-12486: Document tokenize.generate_tokens() as public API (#6957)

* Document tokenize.generate_tokens()

* Add news file

* Add test for generate_tokens

* Document behaviour around ENCODING token

* Add generate_tokens to __all__
diff --git a/Lib/test/test_tokenize.py b/Lib/test/test_tokenize.py
index 3520a67..93e40de 100644
--- a/Lib/test/test_tokenize.py
+++ b/Lib/test/test_tokenize.py
@@ -1,8 +1,8 @@
 from test import support
 from tokenize import (tokenize, _tokenize, untokenize, NUMBER, NAME, OP,
                      STRING, ENDMARKER, ENCODING, tok_name, detect_encoding,
-                     open as tokenize_open, Untokenizer)
-from io import BytesIO
+                     open as tokenize_open, Untokenizer, generate_tokens)
+from io import BytesIO, StringIO
 import unittest
 from unittest import TestCase, mock
 from test.test_grammar import (VALID_UNDERSCORE_LITERALS,
@@ -919,6 +919,19 @@
     DEDENT     ''            (7, 0) (7, 0)
     """)
 
+class GenerateTokensTest(TokenizeTest):
+    def check_tokenize(self, s, expected):
+        # Format the tokens in s in a table format.
+        # The ENDMARKER is omitted.
+        result = []
+        f = StringIO(s)
+        for type, token, start, end, line in generate_tokens(f.readline):
+            if type == ENDMARKER:
+                break
+            type = tok_name[type]
+            result.append(f"    {type:10} {token!r:13} {start} {end}")
+        self.assertEqual(result, expected.rstrip().splitlines())
+
 
 def decistmt(s):
     result = []
diff --git a/Lib/tokenize.py b/Lib/tokenize.py
index 40e6a8b..c78d9f7 100644
--- a/Lib/tokenize.py
+++ b/Lib/tokenize.py
@@ -37,7 +37,7 @@
 blank_re = re.compile(br'^[ \t\f]*(?:[#\r\n]|$)', re.ASCII)
 
 import token
-__all__ = token.__all__ + ["tokenize", "detect_encoding",
+__all__ = token.__all__ + ["tokenize", "generate_tokens", "detect_encoding",
                            "untokenize", "TokenInfo"]
 del token
 
@@ -653,9 +653,12 @@
     yield TokenInfo(ENDMARKER, '', (lnum, 0), (lnum, 0), '')
 
 
-# An undocumented, backwards compatible, API for all the places in the standard
-# library that expect to be able to use tokenize with strings
 def generate_tokens(readline):
+    """Tokenize a source reading Python code as unicode strings.
+
+    This has the same API as tokenize(), except that it expects the *readline*
+    callable to return str objects instead of bytes.
+    """
     return _tokenize(readline, None)
 
 def main():