Handle surrogate pair when converting from LSP encoding

Author: bullno1

ycmd/completers/language_server/language_server_protocol.py

@@ -823,6 +823,12 @@ def UTF16CodeUnitsToCodepoints( line_value, code_unit_offset ):
     # is one-past-the-end of the string in unicode codepoints
     return len( line_value ) + 1
 
+  hi_byte_at_offset = value_as_utf16_bytes[byte_offset_utf16 + 1]
+  if 0xdc <= hi_byte_at_offset <= 0xdf:
+    # If we are in the middle of a surrogate pair, then advance the offset
+    # further to skip the pair
+    code_unit_offset += 1
+
   bytes_included = value_as_utf16_bytes[ : code_unit_offset * 2 ]
   return len( bytes_included.decode( 'utf-16-le' ) )
 

ycmd/tests/language_server/language_server_protocol_test.py

@@ -204,3 +204,14 @@ def test_CodepointsToUTF16CodeUnitsAndReverse( self ):
                      equal_to( code_units ) )
         assert_that( lsp.UTF16CodeUnitsToCodepoints( line_value, code_units ),
                      equal_to( codepoints ) )
+
+  def test_CodepointsToUTF16CodeUnitsSurrogate( self ):
+    for line_value, codepoints, code_units in [
+        ( '😉', 1, 1 ),
+        ( 'f😉', 2, 2 ),
+    ]:
+      with self.subTest( line_value = line_value,
+                         codepoints = codepoints,
+                         code_units = code_units ):
+        assert_that( lsp.UTF16CodeUnitsToCodepoints( line_value, code_units ),
+                     equal_to( codepoints ) )