patch 8.2.4695: JSON encoding could be faster

Problem:    JSON encoding could be faster.
Solution:   Optimize encoding JSON strings. (closes #10086)

diff --git a/src/json.c b/src/json.c
index 5be8f7fe66ddab704dd3d013101e1b1caf5c3b23..47bf9904a312a589938a358e382d0c24b4a7a231 100644 (file)
--- a/src/json.c
+++ b/src/json.c
@@ -114,37 +114,72 @@ json_encode_lsp_msg(typval_T *val)
 }
 #endif
 
+/*
+ * Lookup table to quickly know if the given ASCII character must be escaped.
+ */
+static const char ascii_needs_escape[128] = {
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 0x0.
+    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 0x1.
+    0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x2.
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x3.
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x4.
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, // 0x5.
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x6.
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x7.
+};
+
+/*
+ * Encode the utf-8 encoded string "str" into "gap".
+ */
     static void
 write_string(garray_T *gap, char_u *str)
 {
     char_u     *res = str;
     char_u     numbuf[NUMBUFLEN];
+    char_u     *from;
+#if defined(USE_ICONV)
+    vimconv_T   conv;
+    char_u     *converted = NULL;
+#endif
+    int                c;
 
     if (res == NULL)
-       ga_concat(gap, (char_u *)"\"\"");
-    else
     {
-#if defined(USE_ICONV)
-       vimconv_T   conv;
-       char_u      *converted = NULL;
+       ga_concat(gap, (char_u *)"\"\"");
+       return;
+    }
 
-       if (!enc_utf8)
-       {
-           // Convert the text from 'encoding' to utf-8, the JSON string is
-           // always utf-8.
-           conv.vc_type = CONV_NONE;
-           convert_setup(&conv, p_enc, (char_u*)"utf-8");
-           if (conv.vc_type != CONV_NONE)
-               converted = res = string_convert(&conv, res, NULL);
-           convert_setup(&conv, NULL, NULL);
-       }
+#if defined(USE_ICONV)
+    if (!enc_utf8)
+    {
+       // Convert the text from 'encoding' to utf-8, because a JSON string is
+       // always utf-8.
+       conv.vc_type = CONV_NONE;
+       convert_setup(&conv, p_enc, (char_u*)"utf-8");
+       if (conv.vc_type != CONV_NONE)
+           converted = res = string_convert(&conv, res, NULL);
+       convert_setup(&conv, NULL, NULL);
+    }
 #endif
-       ga_append(gap, '"');
-       while (*res != NUL)
+    ga_append(gap, '"');
+    // `from` is the beginning of a sequence of bytes we can directly copy from
+    // the input string, avoiding the overhead associated to decoding/encoding
+    // them.
+    from = res;
+    while ((c = *res) != NUL)
+    {
+       // always use utf-8 encoding, ignore 'encoding'
+       if (c < 0x80)
        {
-           int c;
-           // always use utf-8 encoding, ignore 'encoding'
-           c = utf_ptr2char(res);
+           if (!ascii_needs_escape[c])
+           {
+               res += 1;
+               continue;
+           }
+
+           if (res != from)
+               ga_concat_len(gap, from, res - from);
+           from = res + 1;
 
            switch (c)
            {
@@ -164,25 +199,43 @@ write_string(garray_T *gap, char_u *str)
                    ga_append(gap, c);
                    break;
                default:
-                   if (c >= 0x20)
-                   {
-                       numbuf[utf_char2bytes(c, numbuf)] = NUL;
-                       ga_concat(gap, numbuf);
-                   }
-                   else
-                   {
-                       vim_snprintf((char *)numbuf, NUMBUFLEN,
-                                                        "\\u%04lx", (long)c);
-                       ga_concat(gap, numbuf);
-                   }
+                   vim_snprintf((char *)numbuf, NUMBUFLEN, "\\u%04lx",
+                                                                     (long)c);
+                   ga_concat(gap, numbuf);
+           }
+
+           res += 1;
+       }
+       else
+       {
+           int l = utf_ptr2len(res);
+
+           if (l > 1)
+           {
+               res += l;
+               continue;
            }
-           res += utf_ptr2len(res);
+
+           // Invalid utf-8 sequence, replace it with the Unicode replacement
+           // character U+FFFD.
+           if (res != from)
+               ga_concat_len(gap, from, res - from);
+           from = res + 1;
+
+           numbuf[utf_char2bytes(0xFFFD, numbuf)] = NUL;
+           ga_concat(gap, numbuf);
+
+           res += l;
        }
-       ga_append(gap, '"');
+    }
+
+    if (res != from)
+       ga_concat_len(gap, from, res - from);
+
+    ga_append(gap, '"');
 #if defined(USE_ICONV)
-       vim_free(converted);
+    vim_free(converted);
 #endif
-    }
 }
 
 /*

diff --git a/src/testdir/test_json.vim b/src/testdir/test_json.vim
index 0248aa9ecced194289effa1fdc4cc7f0583a25d0..3ee783703336d89e0bbd9d03ddb10c2aee0c367b 100644 (file)
--- a/src/testdir/test_json.vim
+++ b/src/testdir/test_json.vim
@@ -107,6 +107,9 @@ func Test_json_encode()
   call assert_equal('"café"', json_encode("caf\xe9"))
   let &encoding = save_encoding
 
+  " Invalid utf-8 sequences are replaced with U+FFFD (replacement character)
+  call assert_equal('"foo' . "\ufffd" . '"', json_encode("foo\xAB"))
+
   call assert_fails('echo json_encode(function("tr"))', 'E1161: Cannot json encode a func')
   call assert_fails('echo json_encode([function("tr")])', 'E1161: Cannot json encode a func')
 

diff --git a/src/version.c b/src/version.c
index 3e220f581f04924f7e29eb3655bc00e1161f785b..f4d300b39e0d0ea2161247700c95bfbda8bf5688 100644 (file)
--- a/src/version.c
+++ b/src/version.c
@@ -746,6 +746,8 @@ static char *(features[]) =
 
 static int included_patches[] =
 {   /* Add new patch number below this line */
+/**/
+    4695,
 /**/
     4694,
 /**/