bug#24603: [RFC 17/18] Optimise character class matching in regexes

[Michal Nazarewicz <mina86@mina86.com>]

Use lookup tables defined in src/character.h to bundle checks together
if possible.  For example, ‘[[:lower:][:digit:]]’ would perform an
equivalence of ‘lowercasep(ch) || numericp(ch)’ check.  Now, such checks
are done all at once with at most one Unicode general category lookup.

Similarly, do at most one syntax table lookup by unrolling macros
testing character properties.

* src/regex.c (execute_charset): Use category_char_bits and call SYNTAX
at most once.

* test/src/regex-tests.el (regex-tests--letter-character-classes): New
test case for various character classes relating to letters etc.
---
 src/regex.c             | 86 +++++++++++++++++++++++++++++++------------------
 test/src/regex-tests.el | 45 ++++++++++++++++++++++++++
 2 files changed, 100 insertions(+), 31 deletions(-)

diff --git a/src/regex.c b/src/regex.c
index 02da1fb..bfd04a1 100644
--- a/src/regex.c
+++ b/src/regex.c
@@ -1789,16 +1789,26 @@ struct range_table_work_area
 /* Bits used to implement the multibyte-part of the various character classes
    such as [:alnum:] in a charset's range table.  The code currently assumes
    that only the low 16 bits are used.  */
-#define BIT_WORD	0x1
-#define BIT_LOWER	0x2
-#define BIT_PUNCT	0x4
-#define BIT_SPACE	0x8
-#define BIT_UPPER	0x10
-#define BIT_MULTIBYTE	0x20
-#define BIT_ALPHA	0x40
-#define BIT_ALNUM	0x80
-#define BIT_GRAPH	0x100
-#define BIT_PRINT	0x200
+#ifdef emacs
+#  define BIT_ALNUM	CHAR_BIT_ALNUM
+#  define BIT_ALPHA	CHAR_BIT_ALPHA
+#  define BIT_UPPER	CHAR_BIT_UPPER
+#  define BIT_LOWER	CHAR_BIT_LOWER
+#  define BIT_GRAPH	CHAR_BIT_GRAPH
+#  define BIT_PRINT	CHAR_BIT_PRINT
+#else
+#  define BIT_ALNUM	(1 << 0)
+#  define BIT_ALPHA	(1 << 1)
+#  define BIT_UPPER	(1 << 2)
+#  define BIT_LOWER	(1 << 3)
+#  define BIT_GRAPH	(1 << 4)
+#  define BIT_PRINT	(1 << 5)
+#endif
+#define BIT_WORD	(BIT_PRINT << 1)
+#define BIT_PUNCT	(BIT_PRINT << 2)
+#define BIT_SPACE	(BIT_PRINT << 3)
+#define BIT_MULTIBYTE	(BIT_PRINT << 4)
+
 \f
 
 /* Set the bit for character C in a list.  */
@@ -1988,9 +1998,6 @@ re_wctype_parse (const unsigned char **strp, unsigned limit)
            2 [:print:]
            2 [:cntrl:]
            1 [:ff:]
-
-     If you update this list, consider also updating chain of or’ed conditions
-     in execute_charset function.
    */
 
   switch (it - beg) {
@@ -4657,28 +4664,45 @@ execute_charset (const_re_char **pp, unsigned c, unsigned corig, bool unibyte)
   else if (rtp)
     {
       int class_bits = CHARSET_RANGE_TABLE_BITS (p);
+      int bits;
       re_wchar_t range_start, range_end;
 
-  /* Sort tests by the most commonly used classes with some adjustment to which
-     tests are easiest to perform.  Take a look at comment in re_wctype_parse
-     for table with frequencies of character class names. */
-
-      if ((class_bits & BIT_MULTIBYTE) ||
-	  (class_bits & BIT_ALNUM && ISALNUM (c)) ||
-	  (class_bits & BIT_ALPHA && ISALPHA (c)) ||
-	  (class_bits & BIT_SPACE && ISSPACE (c)) ||
-	  (class_bits & BIT_WORD  && ISWORD  (c)) ||
-	  ((class_bits & BIT_UPPER) &&
-	   (ISUPPER (c) || (corig != c &&
-			    c == downcase (corig) && ISLOWER (c)))) ||
-	  ((class_bits & BIT_LOWER) &&
-	   (ISLOWER (c) || (corig != c &&
-			    c == upcase (corig) && ISUPPER(c)))) ||
-	  (class_bits & BIT_PUNCT && ISPUNCT (c)) ||
-	  (class_bits & BIT_GRAPH && ISGRAPH (c)) ||
-	  (class_bits & BIT_PRINT && ISPRINT (c)))
+      if (class_bits & BIT_MULTIBYTE)
 	return !not;
 
+      /* If we are at this point, the character is not an ASCII or single byte
+	 character.  This means that whenever ISFOO macros have special case for
+	 IS_REAL_ASCII (c), we can ignore that. */
+
+      bits = class_bits & (BIT_ALNUM | BIT_ALPHA | BIT_UPPER | BIT_LOWER |
+			   BIT_GRAPH | BIT_PRINT);
+      if (bits)
+	{
+	  int char_bits = category_char_bits[char_unicode_category (c)];
+	  if (bits & char_bits)
+	    return !not;
+
+	  /* Handle case folding. */
+	  if (corig != c)
+	    {
+	      if ((bits & BIT_UPPER) && (char_bits & BIT_LOWER) &&
+		  c == downcase (corig))
+		return !not;
+	      if ((bits & BIT_LOWER) && (char_bits & BIT_UPPER) &&
+		  c == upcase (corig))
+		return !not;
+	    }
+	}
+
+      if (class_bits & (BIT_SPACE | BIT_WORD | BIT_PUNCT))
+	{
+	  enum syntaxcode s = SYNTAX (c);
+	  if (((class_bits & BIT_SPACE) && s == Swhitespace) ||
+	      ((class_bits & BIT_WORD ) && s == Sword) ||
+	      ((class_bits & BIT_PUNCT) && s != Sword))
+	    return !not;
+	}
+
       for (p = *pp; rtp < p; rtp += 2 * 3)
 	{
 	  EXTRACT_CHARACTER (range_start, rtp);
diff --git a/test/src/regex-tests.el b/test/src/regex-tests.el
index fc50344..7617823 100644
--- a/test/src/regex-tests.el
+++ b/test/src/regex-tests.el
@@ -98,6 +98,51 @@ regex--test-cc
     (eval `(ert-deftest ,name () ,doc ,(cons 'regex--test-cc test)) t)))
 
 
+(ert-deftest regex-tests--letter-character-classes ()
+  "Test character classes against various letters types."
+  (should-not
+   (cl-remove-if
+    'not
+    (let ((check-ccs (lambda (ch fold)
+                       (mapconcat
+                        (lambda (str) str)
+                        (let ((case-fold-search fold))
+                          (cl-remove-if-not
+                           (lambda (cc)
+                             (string-match-p (concat "[[:" cc ":]]")
+                                             (string ch)))
+                           '("alnum" "alpha" "upper" "lower")))
+                        " "))))
+      (mapcar
+       (lambda (entry)
+         (let ((ch (car entry)) (expected (cdr entry)))
+           (setq entry
+                 (format "%s | %s | case-fold: %s"
+                         (get-char-code-property ch 'general-category)
+                         (funcall check-ccs ch nil) (funcall check-ccs ch t)))
+           (unless (string-equal expected entry)
+             (format "\n%c        expected: %s\nU+%06X  but got: %s"
+                     ch expected ch entry))))
+       '((?A . "Lu | alnum alpha upper | case-fold: alnum alpha upper lower")
+         (?ẞ . "Lu | alnum alpha upper | case-fold: alnum alpha upper lower")
+         (?DZ . "Lu | alnum alpha upper | case-fold: alnum alpha upper lower")
+         (?a . "Ll | alnum alpha lower | case-fold: alnum alpha upper lower")
+         ;; FIXME: Should match upper when case-fold case
+         ;; (?ł . "Ll | alnum alpha lower | case-fold: alnum alpha upper lower")
+         ;; (?ß . "Ll | alnum alpha lower | case-fold: alnum alpha upper lower")
+         ;; (?fi . "Ll | alnum alpha lower | case-fold: alnum alpha upper lower")
+         ;; (?ɕ . "Ll | alnum alpha lower | case-fold: alnum alpha upper lower")
+         ;; (?dz . "Ll | alnum alpha lower | case-fold: alnum alpha upper lower")
+         (?ł . "Ll | alnum alpha lower | case-fold: alnum alpha lower")
+         (?ß . "Ll | alnum alpha lower | case-fold: alnum alpha lower")
+         (?fi . "Ll | alnum alpha lower | case-fold: alnum alpha lower")
+         (?ɕ . "Ll | alnum alpha lower | case-fold: alnum alpha lower")
+         (?dz . "Ll | alnum alpha lower | case-fold: alnum alpha lower")
+         (?Dz . "Lt | alnum alpha | case-fold: alnum alpha upper lower")
+         (?ʰ . "Lm | alnum alpha | case-fold: alnum alpha")
+         (?º . "Lo | alnum alpha | case-fold: alnum alpha")))))))
+
+
 (defmacro regex-tests-generic-line (comment-char test-file whitelist &rest body)
   "Reads a line of the test file TEST-FILE, skipping
 comments (defined by COMMENT-CHAR), and evaluates the tests in
-- 
2.8.0.rc3.226.g39d4020