Merge #6460 from ZyX-I/1476-changes

Refactor functions which find character in a string

6 files changed, 71 insertions, 142 deletions

diff --git a/src/nvim/ex_cmds.c b/src/nvim/ex_cmds.c
index f73d4c16b6..018a228843 100644
--- a/src/nvim/ex_cmds.c
+++ b/src/nvim/ex_cmds.c
@@ -5101,14 +5101,13 @@ static void helptags_one(char_u *dir, char_u *ext, char_u *tagfname,
       }
       p1 = vim_strchr(IObuff, '*');             /* find first '*' */
       while (p1 != NULL) {
-        /* Use vim_strbyte() instead of vim_strchr() so that when
-         * 'encoding' is dbcs it still works, don't find '*' in the
-         * second byte. */
-        p2 = vim_strbyte(p1 + 1, '*');          /* find second '*' */
-        if (p2 != NULL && p2 > p1 + 1) {        /* skip "*" and "**" */
-          for (s = p1 + 1; s < p2; ++s)
-            if (*s == ' ' || *s == '\t' || *s == '|')
+        p2 = (char_u *)strchr((const char *)p1 + 1, '*');  // Find second '*'.
+        if (p2 != NULL && p2 > p1 + 1) {  // Skip "*" and "**".
+          for (s = p1 + 1; s < p2; s++) {
+            if (*s == ' ' || *s == '\t' || *s == '|') {
               break;
+            }
+          }
 
           /*
            * Only accept a *tag* when it consists of valid
diff --git a/src/nvim/getchar.c b/src/nvim/getchar.c
index 6e21ee96e8..79b95272de 100644
--- a/src/nvim/getchar.c
+++ b/src/nvim/getchar.c
@@ -3596,8 +3596,8 @@ int check_abbr(int c, char_u *ptr, int col, int mincol)
       char_u *q = mp->m_keys;
       int match;
 
-      if (vim_strbyte(mp->m_keys, K_SPECIAL) != NULL) {
-        /* might have CSI escaped mp->m_keys */
+      if (strchr((const char *)mp->m_keys, K_SPECIAL) != NULL) {
+        // Might have CSI escaped mp->m_keys.
         q = vim_strsave(mp->m_keys);
         vim_unescape_csi(q);
         qlen = (int)STRLEN(q);
diff --git a/src/nvim/mbyte.c b/src/nvim/mbyte.c
index 8a68ce0915..d5907da2ed 100644
--- a/src/nvim/mbyte.c
+++ b/src/nvim/mbyte.c
@@ -359,10 +359,10 @@ int bomb_size(void)
  */
 void remove_bom(char_u *s)
 {
-  char_u *p = s;
+  char *p = (char *)s;
 
-  while ((p = vim_strbyte(p, 0xef)) != NULL) {
-    if (p[1] == 0xbb && p[2] == 0xbf) {
+  while ((p = strchr(p, 0xef)) != NULL) {
+    if ((uint8_t)p[1] == 0xbb && (uint8_t)p[2] == 0xbf) {
       STRMOVE(p, p + 3);
     } else {
       p++;
diff --git a/src/nvim/regexp.c b/src/nvim/regexp.c
index 7a00ee27bb..5448cc7131 100644
--- a/src/nvim/regexp.c
+++ b/src/nvim/regexp.c
@@ -3427,32 +3427,26 @@ static long bt_regexec_both(char_u *line,
       c = *prog->regmust;
     s = line + col;
 
-    /*
-     * This is used very often, esp. for ":global".  Use three versions of
-     * the loop to avoid overhead of conditions.
-     */
-    if (!ireg_ic
-        && !has_mbyte
-        )
-      while ((s = vim_strbyte(s, c)) != NULL) {
-        if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
-          break;                        /* Found it. */
-        ++s;
-      }
-    else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1))
+    // This is used very often, esp. for ":global".  Use two versions of
+    // the loop to avoid overhead of conditions.
+    if (!ireg_ic) {
       while ((s = vim_strchr(s, c)) != NULL) {
-        if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
-          break;                        /* Found it. */
+        if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) {
+          break;  // Found it.
+        }
         mb_ptr_adv(s);
       }
-    else
+    } else {
       while ((s = cstrchr(s, c)) != NULL) {
-        if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
-          break;                        /* Found it. */
+        if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0) {
+          break;  // Found it.
+        }
         mb_ptr_adv(s);
       }
-    if (s == NULL)              /* Not present. */
+    }
+    if (s == NULL) {  // Not present.
       goto theend;
+    }
   }
 
   regline = line;
@@ -3482,14 +3476,8 @@ static long bt_regexec_both(char_u *line,
     /* Messy cases:  unanchored match. */
     while (!got_int) {
       if (prog->regstart != NUL) {
-        /* Skip until the char we know it must start with.
-         * Used often, do some work to avoid call overhead. */
-        if (!ireg_ic
-            && !has_mbyte
-            )
-          s = vim_strbyte(regline + col, prog->regstart);
-        else
-          s = cstrchr(regline + col, prog->regstart);
+        // Skip until the char we know it must start with.
+        s = cstrchr(regline + col, prog->regstart);
         if (s == NULL) {
           retval = 0;
           break;
@@ -6299,44 +6287,37 @@ static int cstrncmp(char_u *s1, char_u *s2, int *n)
 /*
  * cstrchr: This function is used a lot for simple searches, keep it fast!
  */
-static char_u *cstrchr(char_u *s, int c)
+static inline char_u *cstrchr(const char_u *const s, const int c)
+  FUNC_ATTR_PURE FUNC_ATTR_WARN_UNUSED_RESULT FUNC_ATTR_NONNULL_ALL
+  FUNC_ATTR_ALWAYS_INLINE
 {
-  char_u      *p;
-  int cc;
-
-  if (!ireg_ic
-      || (!enc_utf8 && mb_char2len(c) > 1)
-      )
+  if (!ireg_ic) {
     return vim_strchr(s, c);
+  }
 
-  /* tolower() and toupper() can be slow, comparing twice should be a lot
-   * faster (esp. when using MS Visual C++!).
-   * For UTF-8 need to use folded case. */
+  // Use folded case for UTF-8, slow! For ASCII use libc strpbrk which is
+  // expected to be highly optimized.
   if (c > 0x80) {
-    cc = utf_fold(c);
-  } else if (mb_isupper(c)) {
-    cc = mb_tolower(c);
-  } else if (mb_islower(c)) {
-    cc = mb_toupper(c);
+    const int folded_c = utf_fold(c);
+    for (const char_u *p = s; *p != NUL; p += utfc_ptr2len(p)) {
+      if (utf_fold(utf_ptr2char(p)) == folded_c) {
+        return (char_u *)p;
+      }
+    }
+    return NULL;
+  }
+
+  int cc;
+  if (ASCII_ISUPPER(c)) {
+    cc = TOLOWER_ASC(c);
+  } else if (ASCII_ISLOWER(c)) {
+    cc = TOUPPER_ASC(c);
   } else {
     return vim_strchr(s, c);
   }
 
-  if (has_mbyte) {
-    for (p = s; *p != NUL; p += (*mb_ptr2len)(p)) {
-      if (enc_utf8 && c > 0x80) {
-        if (utf_fold(utf_ptr2char(p)) == cc)
-          return p;
-      } else if (*p == c || *p == cc)
-        return p;
-    }
-  } else
-    /* Faster version for when there are no multi-byte characters. */
-    for (p = s; *p != NUL; ++p)
-      if (*p == c || *p == cc)
-        return p;
-
-  return NULL;
+  char tofind[] = { (char)c, (char)cc, NUL };
+  return (char_u *)strpbrk((const char *)s, tofind);
 }
 
 /***************************************************************
diff --git a/src/nvim/regexp_nfa.c b/src/nvim/regexp_nfa.c
index 139772d51e..85fae9d82e 100644
--- a/src/nvim/regexp_nfa.c
+++ b/src/nvim/regexp_nfa.c
@@ -4855,17 +4855,10 @@ static int failure_chance(nfa_state_T *state, int depth)
  */
 static int skip_to_start(int c, colnr_T *colp)
 {
-  char_u *s;
-
-  /* Used often, do some work to avoid call overhead. */
-  if (!ireg_ic
-      && !has_mbyte
-      )
-    s = vim_strbyte(regline + *colp, c);
-  else
-    s = cstrchr(regline + *colp, c);
-  if (s == NULL)
+  const char_u *const s = cstrchr(regline + *colp, c);
+  if (s == NULL) {
     return FAIL;
+  }
   *colp = (int)(s - regline);
   return OK;
 }
diff --git a/src/nvim/strings.c b/src/nvim/strings.c
index 743b43c2e5..f19cf7a261 100644
--- a/src/nvim/strings.c
+++ b/src/nvim/strings.c
@@ -425,72 +425,28 @@ int vim_strnicmp(const char *s1, const char *s2, size_t len)
 }
 #endif
 
-/*
- * Version of strchr() and strrchr() that handle unsigned char strings
- * with characters from 128 to 255 correctly.  It also doesn't return a
- * pointer to the NUL at the end of the string.
- */
-char_u *vim_strchr(const char_u *string, int c)
-  FUNC_ATTR_NONNULL_ALL FUNC_ATTR_PURE
+/// strchr() version which handles multibyte strings
+///
+/// @param[in]  string  String to search in.
+/// @param[in]  c  Character to search for. Must be a valid character.
+///
+/// @return Pointer to the first byte of the found character in string or NULL
+///         if it was not found. NUL character is never found, use `strlen()`
+///         instead.
+char_u *vim_strchr(const char_u *const string, const int c)
+  FUNC_ATTR_NONNULL_ALL FUNC_ATTR_PURE FUNC_ATTR_WARN_UNUSED_RESULT
 {
-  int b;
-
-  const char_u *p = string;
-  if (enc_utf8 && c >= 0x80) {
-    while (*p != NUL) {
-      int l = (*mb_ptr2len)(p);
-
-      // Avoid matching an illegal byte here.
-      if (l > 1 && utf_ptr2char(p) == c) {
-        return (char_u *) p;
-      }
-      p += l;
-    }
+  assert(c >= 0);
+  if (c == 0) {
     return NULL;
+  } else if (c < 0x80) {
+    return (char_u *)strchr((const char *)string, c);
+  } else {
+    char u8char[MB_MAXBYTES + 1];
+    const int len = utf_char2bytes(c, (char_u *)u8char);
+    u8char[len] = NUL;
+    return (char_u *)strstr((const char *)string, u8char);
   }
-  if (enc_dbcs != 0 && c > 255) {
-    int n2 = c & 0xff;
-
-    c = ((unsigned)c >> 8) & 0xff;
-    while ((b = *p) != NUL) {
-      if (b == c && p[1] == n2)
-        return (char_u *) p;
-      p += (*mb_ptr2len)(p);
-    }
-    return NULL;
-  }
-  if (has_mbyte) {
-    while ((b = *p) != NUL) {
-      if (b == c)
-        return (char_u *) p;
-      p += (*mb_ptr2len)(p);
-    }
-    return NULL;
-  }
-  while ((b = *p) != NUL) {
-    if (b == c)
-      return (char_u *) p;
-    ++p;
-  }
-  return NULL;
-}
-
-/*
- * Version of strchr() that only works for bytes and handles unsigned char
- * strings with characters above 128 correctly. It also doesn't return a
- * pointer to the NUL at the end of the string.
- */
-char_u *vim_strbyte(const char_u *string, int c)
-  FUNC_ATTR_NONNULL_ALL FUNC_ATTR_PURE
-{
-  const char_u *p = string;
-
-  while (*p != NUL) {
-    if (*p == c)
-      return (char_u *) p;
-    ++p;
-  }
-  return NULL;
 }
 
 /*