From e64b984ddc514a0c7d7f5db3f43590e62285c722 Mon Sep 17 00:00:00 2001
From: Roger Light <roger@atchoo.org>
Date: Thu, 10 Aug 2023 09:26:19 +0100
Subject: [PATCH] Optimise string parsing when there are no escape characters

This resulted in approximately 40% reduction in instructions used in
parse_string() when parsing a 4MB JSON file, as measured in callgrind.
---
 cJSON.c              | 124 +++++++++++++++++++++++--------------------
 tests/parse_string.c |   1 +
 2 files changed, 67 insertions(+), 58 deletions(-)

diff --git a/cJSON.c b/cJSON.c
index 54f63f4..dc1934d 100644
--- a/cJSON.c
+++ b/cJSON.c
@@ -776,6 +776,8 @@ static cJSON_bool parse_string(cJSON * const item, parse_buffer * const input_bu
     const unsigned char *input_end = buffer_at_offset(input_buffer) + 1;
     unsigned char *output_pointer = NULL;
     unsigned char *output = NULL;
+    size_t skipped_bytes = 0;
+    size_t allocation_length = 0;
 
     /* not a string */
     if (buffer_at_offset(input_buffer)[0] != '\"')
@@ -785,8 +787,6 @@ static cJSON_bool parse_string(cJSON * const item, parse_buffer * const input_bu
 
     {
         /* calculate approximate size of the output (overestimate) */
-        size_t allocation_length = 0;
-        size_t skipped_bytes = 0;
         while (((size_t)(input_end - input_buffer->content) < input_buffer->length) && (*input_end != '\"'))
         {
             /* is escape sequence */
@@ -816,65 +816,73 @@ static cJSON_bool parse_string(cJSON * const item, parse_buffer * const input_bu
         }
     }
 
-    output_pointer = output;
-    /* loop through the string literal */
-    while (input_pointer < input_end)
+    /* If there are no escaped characters, we can use memcmp */
+    if(skipped_bytes == 0)
     {
-        if (*input_pointer != '\\')
-        {
-            *output_pointer++ = *input_pointer++;
-        }
-        /* escape sequence */
-        else
-        {
-            unsigned char sequence_length = 2;
-            if ((input_end - input_pointer) < 1)
-            {
-                goto fail;
-            }
-
-            switch (input_pointer[1])
-            {
-                case 'b':
-                    *output_pointer++ = '\b';
-                    break;
-                case 'f':
-                    *output_pointer++ = '\f';
-                    break;
-                case 'n':
-                    *output_pointer++ = '\n';
-                    break;
-                case 'r':
-                    *output_pointer++ = '\r';
-                    break;
-                case 't':
-                    *output_pointer++ = '\t';
-                    break;
-                case '\"':
-                case '\\':
-                case '/':
-                    *output_pointer++ = input_pointer[1];
-                    break;
-
-                /* UTF-16 literal */
-                case 'u':
-                    sequence_length = utf16_literal_to_utf8(input_pointer, input_end, &output_pointer);
-                    if (sequence_length == 0)
-                    {
-                        /* failed to convert UTF16-literal to UTF-8 */
-                        goto fail;
-                    }
-                    break;
-
-                default:
-                    goto fail;
-            }
-            input_pointer += sequence_length;
-        }
+        memcpy(output, input_pointer, allocation_length);
+        output[allocation_length-1] = '\0';
     }
+    else
+    {
+        /* else loop through the string literal */
+        output_pointer = output;
+        while (input_pointer < input_end)
+        {
+            if (*input_pointer != '\\')
+            {
+                *output_pointer++ = *input_pointer++;
+            }
+            /* escape sequence */
+            else
+            {
+                unsigned char sequence_length = 2;
+                if ((input_end - input_pointer) < 1)
+                {
+                    goto fail;
+                }
 
-    /* zero terminate the output */
-    *output_pointer = '\0';
+                switch (input_pointer[1])
+                {
+                    case 'b':
+                        *output_pointer++ = '\b';
+                        break;
+                    case 'f':
+                        *output_pointer++ = '\f';
+                        break;
+                    case 'n':
+                        *output_pointer++ = '\n';
+                        break;
+                    case 'r':
+                        *output_pointer++ = '\r';
+                        break;
+                    case 't':
+                        *output_pointer++ = '\t';
+                        break;
+                    case '\"':
+                    case '\\':
+                    case '/':
+                        *output_pointer++ = input_pointer[1];
+                        break;
+
+                    /* UTF-16 literal */
+                    case 'u':
+                        sequence_length = utf16_literal_to_utf8(input_pointer, input_end, &output_pointer);
+                        if (sequence_length == 0)
+                        {
+                            /* failed to convert UTF16-literal to UTF-8 */
+                            goto fail;
+                        }
+                        break;
+
+                    default:
+                        goto fail;
+                }
+                input_pointer += sequence_length;
+            }
+        }
+        /* zero terminate the output */
+        *output_pointer = '\0';
+    }
 
     item->type = cJSON_String;
     item->valuestring = (char*)output;
diff --git a/tests/parse_string.c b/tests/parse_string.c
index ce1c138..2435c26 100644
--- a/tests/parse_string.c
+++ b/tests/parse_string.c
@@ -73,6 +73,7 @@ static void assert_not_parse_string(const char * const string)
 static void parse_string_should_parse_strings(void)
 {
     assert_parse_string("\"\"", "");
+    assert_parse_string("\"Simple string\"", "Simple string");
     assert_parse_string(
         "\" !\\\"#$%&'()*+,-./\\/0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\\\]^_'abcdefghijklmnopqrstuvwxyz{|}~\"",
         " !\"#$%&'()*+,-.//0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_'abcdefghijklmnopqrstuvwxyz{|}~");