Make the option -s easier to use.

ChangeLog

@@ -1,3 +1,20 @@
+2002-12-07  Bruno Haible  <bruno@clisp.org>
+
+	* src/options.h (Options::get_size_multiple): Change return type to
+	float.
+	(Options::_size_multiple): Change type to float.
+	* src/options.icc (Options::get_size_multiple): Change return type to
+	float.
+	* src/options.cc (Options::long_usage): Update description of option
+	-s.
+	(Options::~Options): Update.
+	(Options::parse_options): For option -s, accept a fraction.
+	* src/search.cc (Search::prepare_asso_values): Use get_size_multiple
+	as it is.
+	* tests/test-6.exp: Update.
+	* doc/gperf.texi (Algorithmic Details): Update description of option
+	-s.
+
 2002-12-04  Bruno Haible  <bruno@clisp.org>
 
 	Improve debugging output.

doc/gperf.texi

@@ -1074,16 +1074,13 @@ table.
 Affects the size of the generated hash table.  The numeric argument for
 this option indicates ``how many times larger or smaller'' the maximum
 associated value range should be, in relationship to the number of keywords.
-If the @var{size-multiple} is negative the maximum associated value is
+It can be written as an integer, a floating-point number or a fraction.
-calculated by @emph{dividing} the number of keywords by it.  For
+For example, a value of 3 means ``allow the maximum associated value to be
-example, a value of 3 means ``allow the maximum associated value to be
+about 3 times larger than the number of input keywords''.
-about 3 times larger than the number of input keywords''.  
+Conversely, a value of 1/3 means ``allow the maximum associated value to
-
+be about 3 times smaller than the number of input keywords''.  Values
-Conversely, a value of -3 means ``allow the maximum associated value to
+smaller than 1 are useful for limiting the overall size of the generated hash
-be about 3 times smaller than the number of input keywords''.  Negative
+table, though the option @samp{-m} is better at this purpose.
-values are useful for limiting the overall size of the generated hash
-table, though this usually increases the number of duplicate hash
-values.
 
 If `generate switch' option @samp{-S} (or, equivalently, @samp{%switch}) is
 @emph{not} enabled, the maximum

src/options.cc

@@ -234,7 +234,7 @@ Options::long_usage (FILE * stream) const
            "                         in relationship to the number of keys, e.g. a value\n"
            "                         of 3 means \"allow the maximum associated value to\n"
            "                         be about 3 times larger than the number of input\n"
-           "                         keys\". Conversely, a value of -3 means \"make the\n"
+           "                         keys\". Conversely, a value of 1/3 means \"make the\n"
            "                         maximum associated value about 3 times smaller than\n"
            "                         the number of input keys\". A larger table should\n"
            "                         decrease the time required for an unsuccessful\n"
@@ -481,7 +481,7 @@ Options::~Options ()
                "\ninitializer suffix = %s"
                "\nasso_values iterations = %d"
                "\njump value = %d"
-               "\nhash table size multiplier = %d"
+               "\nhash table size multiplier = %g"
                "\ninitial associated value = %d"
                "\ndelimiters = %s"
                "\nnumber of switch statements = %d\n",
@@ -873,8 +873,44 @@ Options::parse_options (int argc, char *argv[])
           }
         case 's':               /* Range of associated values, determines size of final table. */
           {
-            if (abs (_size_multiple = atoi (/*getopt*/optarg)) > 50)
+            float numerator;
-              fprintf (stderr, "%d is excessive, did you really mean this?! (try `%s --help' for help)\n", _size_multiple, program_name);
+            float denominator = 1;
+            bool invalid = false;
+            char *endptr;
+
+            numerator = strtod (/*getopt*/optarg, &endptr);
+            if (endptr == /*getopt*/optarg)
+              invalid = true;
+            else if (*endptr != '\0')
+              {
+                if (*endptr == '/')
+                  {
+                    char *denomptr = endptr + 1;
+                    denominator = strtod (denomptr, &endptr);
+                    if (endptr == denomptr || *endptr != '\0')
+                      invalid = true;
+                  }
+                else
+                  invalid = true;
+              }
+            if (invalid)
+              {
+                fprintf (stderr, "Invalid value for option -s.\n");
+                short_usage (stderr);
+                exit (1);
+              }
+            _size_multiple = numerator / denominator;
+            /* Backward compatibility: -3 means 1/3.  */
+            if (_size_multiple < 0)
+              _size_multiple = 1 / (-_size_multiple);
+            /* Catch stupid users.  */
+            if (_size_multiple == 0)
+              _size_multiple = 1;
+            /* Warnings.  */
+            if (_size_multiple > 50)
+              fprintf (stderr, "Size multiple %g is excessive, did you really mean this?! (try '%s --help' for help)\n", _size_multiple, program_name);
+            else if (_size_multiple < 0.01f)
+              fprintf (stderr, "Size multiple %g is extremely small, did you really mean this?! (try '%s --help' for help)\n", _size_multiple, program_name);
             break;
           }
         case 'S':               /* Generate switch statement output, rather than lookup table. */

src/options.h

@@ -155,7 +155,7 @@ public:
   void                  set_total_switches (int total_switches);
 
   /* Returns the factor by which to multiply the generated table's size.  */
-  int                   get_size_multiple () const;
+  float                 get_size_multiple () const;
 
   /* Returns the generated function name.  */
   const char *          get_function_name () const;
@@ -235,7 +235,7 @@ private:
   int                   _total_switches;
 
   /* Factor by which to multiply the generated table's size.  */
-  int                   _size_multiple;
+  float                 _size_multiple;
 
   /* Names used for generated lookup function.  */
   const char *          _function_name;

src/options.icc

@@ -87,7 +87,7 @@ Options::get_total_switches () const
 }
 
 /* Returns the factor by which to multiply the generated table's size.  */
-INLINE int
+INLINE float
 Options::get_size_multiple () const
 {
   return _size_multiple;

src/search.cc

@@ -802,16 +802,11 @@ Search::get_max_keysig_size () const
 void
 Search::prepare_asso_values ()
 {
-  int size_multiple = option.get_size_multiple ();
   int non_linked_length = keyword_list_length ();
   int asso_value_max;
 
-  if (size_multiple == 0)
+  asso_value_max =
-    asso_value_max = non_linked_length;
+    static_cast<int>(non_linked_length * option.get_size_multiple());
-  else if (size_multiple > 0)
-    asso_value_max = non_linked_length * size_multiple;
-  else /* if (size_multiple < 0) */
-    asso_value_max = non_linked_length / -size_multiple;
   /* Round up to the next power of two.  This makes it easy to ensure
      an _asso_value[c] is >= 0 and < asso_value_max.  Also, the jump value
      being odd, it guarantees that Search::try_asso_value() will iterate

tests/test-6.exp

@@ -117,7 +117,7 @@ Algorithm employed by gperf:
                          in relationship to the number of keys, e.g. a value
                          of 3 means "allow the maximum associated value to
                          be about 3 times larger than the number of input
-                         keys". Conversely, a value of -3 means "make the
+                         keys". Conversely, a value of 1/3 means "make the
                          maximum associated value about 3 times smaller than
                          the number of input keys". A larger table should
                          decrease the time required for an unsuccessful