[ruby/prism] Better invalid token messages

8c9bed2a4d

prism/diagnostic.c

@@ -198,8 +198,10 @@ static const pm_diagnostic_data_t diagnostic_messages[PM_DIAGNOSTIC_ID_LEN] = {
     [PM_ERR_INVALID_NUMBER_HEXADECIMAL]         = { "invalid hexadecimal number", PM_ERROR_LEVEL_FATAL },
     [PM_ERR_INVALID_NUMBER_OCTAL]               = { "invalid octal number", PM_ERROR_LEVEL_FATAL },
     [PM_ERR_INVALID_NUMBER_UNDERSCORE]          = { "invalid underscore placement in number", PM_ERROR_LEVEL_FATAL },
+    [PM_ERR_INVALID_CHARACTER]                  = { "invalid character 0x%X", PM_ERROR_LEVEL_FATAL },
+    [PM_ERR_INVALID_MULTIBYTE_CHARACTER]        = { "invalid multibyte character 0x%X", PM_ERROR_LEVEL_FATAL },
+    [PM_ERR_INVALID_PRINTABLE_CHARACTER]        = { "invalid character `%c`", PM_ERROR_LEVEL_FATAL },
     [PM_ERR_INVALID_PERCENT]                    = { "invalid `%` token", PM_ERROR_LEVEL_FATAL }, // TODO WHAT?
-    [PM_ERR_INVALID_TOKEN]                      = { "invalid token", PM_ERROR_LEVEL_FATAL }, // TODO WHAT?
     [PM_ERR_INVALID_VARIABLE_GLOBAL]            = { "invalid global variable", PM_ERROR_LEVEL_FATAL },
     [PM_ERR_IT_NOT_ALLOWED]                     = { "`it` is not allowed when an ordinary parameter is defined", PM_ERROR_LEVEL_FATAL },
     [PM_ERR_LAMBDA_OPEN]                        = { "expected a `do` keyword or a `{` to open the lambda block", PM_ERROR_LEVEL_FATAL },

prism/diagnostic.h

@@ -196,8 +196,10 @@ typedef enum {
     PM_ERR_INVALID_NUMBER_HEXADECIMAL,
     PM_ERR_INVALID_NUMBER_OCTAL,
     PM_ERR_INVALID_NUMBER_UNDERSCORE,
+    PM_ERR_INVALID_CHARACTER,
+    PM_ERR_INVALID_MULTIBYTE_CHARACTER,
+    PM_ERR_INVALID_PRINTABLE_CHARACTER,
     PM_ERR_INVALID_PERCENT,
-    PM_ERR_INVALID_TOKEN,
     PM_ERR_INVALID_VARIABLE_GLOBAL,
     PM_ERR_IT_NOT_ALLOWED,
     PM_ERR_LAMBDA_OPEN,

prism/prism.c

@@ -9590,11 +9590,21 @@ parser_lex(pm_parser_t *parser) {
                     if (*parser->current.start != '_') {
                         size_t width = char_is_identifier_start(parser, parser->current.start);
 
-                        // If this isn't the beginning of an identifier, then it's an invalid
+                        // If this isn't the beginning of an identifier, then
-                        // token as we've exhausted all of the other options. We'll skip past
+                        // it's an invalid token as we've exhausted all of the
-                        // it and return the next token.
+                        // other options. We'll skip past it and return the next
+                        // token after adding an appropriate error message.
                         if (!width) {
-                            pm_parser_err_current(parser, PM_ERR_INVALID_TOKEN);
+                            pm_diagnostic_id_t diag_id;
+                            if (*parser->current.start >= 0x80) {
+                                diag_id = PM_ERR_INVALID_MULTIBYTE_CHARACTER;
+                            } else if (char_is_ascii_printable(*parser->current.start) || (*parser->current.start == '\\')) {
+                                diag_id = PM_ERR_INVALID_PRINTABLE_CHARACTER;
+                            } else {
+                                diag_id = PM_ERR_INVALID_CHARACTER;
+                            }
+
+                            PM_PARSER_ERR_TOKEN_FORMAT(parser, parser->current, diag_id, *parser->current.start);
                             goto lex_next_token;
                         }
 

test/prism/format_errors_test.rb

@@ -16,7 +16,7 @@ module Prism
       assert_equal <<~'ERROR', Debug.format_errors('"%W"\u"', false)
         > 1 | "%W"\u"
             |     ^ expected a newline or semicolon after the statement
-            |     ^ invalid token
+            |     ^ invalid character `\`
             |        ^ expected a closing delimiter for the string literal
       ERROR
     end