Re: mention of "use sort" botched in perlfunc? + PATCH

From: "John P. Linderman" <jpl@research.att.com>
Message-id: <200207211613.MAA69505@raptor.research.att.com>

p4raw-id: //depot/perl@17685

diff --git a/lib/sort.pm b/lib/sort.pm
index 3db4777..e785003 100644 (file)
--- a/lib/sort.pm
+++ b/lib/sort.pm
@@ -1,6 +1,6 @@
 package sort;
 
-our $VERSION = '1.01';
+our $VERSION = '1.02';
 
 # Currently the hints for pp_sort are stored in the global variable
 # $sort::hints. An improvement would be to store them in $^H{SORT} and have
@@ -33,6 +33,30 @@ sub import {
            $hints |=  $sort::mergesort_bit;
        } elsif ($_ eq 'stable') {
            $hints |=  $sort::stable_bit;
+       } elsif ($_ eq 'defaults') {
+           $hints =   0;
+       } else {
+           require Carp;
+           Carp::croak("sort: unknown subpragma '$_'");
+       }
+    }
+}
+
+sub unimport {
+    shift;
+    if (@_ == 0) {
+       require Carp;
+       Carp::croak("sort pragma requires arguments");
+    }
+    local $_;
+    no warnings 'uninitialized';       # bitops would warn
+    while ($_ = shift(@_)) {
+       if (/^_q(?:uick)?sort$/) {
+           $hints &= ~$sort::sort_bits;
+       } elsif ($_ eq '_mergesort') {
+           $hints &= ~$sort::sort_bits;
+       } elsif ($_ eq 'stable') {
+           $hints &= ~$sort::stable_bit;
        } else {
            require Carp;
            Carp::croak("sort: unknown subpragma '$_'");
@@ -63,6 +87,8 @@ sort - perl pragma to control sort() behaviour
     use sort 'stable';         # guarantee stability
     use sort '_quicksort';     # use a quicksort algorithm
     use sort '_mergesort';     # use a mergesort algorithm
+    use sort 'defaults';       # revert to default behavior
+    no  sort 'stable';         # stability not important
 
     use sort '_qsort';         # alias for quicksort
 
@@ -70,11 +96,11 @@ sort - perl pragma to control sort() behaviour
 
 =head1 DESCRIPTION
 
-With the sort pragma you can control the behaviour of the builtin
-sort() function.
+With the C<sort> pragma you can control the behaviour of the builtin
+C<sort()> function.
 
 In Perl versions 5.6 and earlier the quicksort algorithm was used to
-implement sort(), but in Perl 5.8 a mergesort algorithm was also made
+implement C<sort()>, but in Perl 5.8 a mergesort algorithm was also made
 available, mainly to guarantee worst case O(N log N) behaviour:
 the worst case of quicksort is O(N**2).  In Perl 5.8 and later,
 quicksort defends against quadratic behaviour by shuffling large
@@ -98,16 +124,78 @@ The best algorithm depends on many things.  On average, mergesort
 does fewer comparisons than quicksort, so it may be better when
 complicated comparison routines are used.  Mergesort also takes
 advantage of pre-existing order, so it would be favored for using
-sort to merge several sorted arrays.  On the other hand, quicksort
-is often faster for small arrays, and on platforms with small memory
-caches that are much faster than main memory.  You can force the
+C<sort()> to merge several sorted arrays.  On the other hand, quicksort
+is often faster for small arrays, and on arrays of a few distinct
+values, repeated many times.  You can force the
 choice of algorithm with this pragma, but this feels heavy-handed,
 so the subpragmas beginning with a C<_> may not persist beyond Perl 5.8.
+The default algorithm is mergesort, which will be stable even if
+you do not explicitly demand it.
+But the stability of the default sort is a side-effect that could
+change in later versions.  If stability is important, be sure to
+say so with a
+
+  use sort 'stable';
+
+The C<no sort> pragma doesn't
+I<forbid> what follows, it just leaves the choice open.  Thus, after
+
+  no sort qw(_mergesort stable);
+
+a mergesort, which happens to be stable, will be employed anyway.
+Note that
+
+  no sort "_quicksort";
+  no sort "_mergesort";
+
+have exactly the same effect, leaving the choice of sort algorithm open.
 
 =head1 CAVEATS
 
-This pragma is not lexically scoped : its effect is global to the program
-it appears in.  This may change in future versions.
+This pragma is not lexically scoped: its effect is global to the program
+it appears in.  That means the following will probably not do what you
+expect, because I<both> pragmas take effect at compile time, before
+I<either> C<sort()> happens.
+
+  { use sort "_quicksort";
+    print sort::current . "\n";
+    @a = sort @b;
+  }
+  { use sort "stable";
+    print sort::current . "\n";
+    @c = sort @d;
+  }
+  # prints:
+  # quicksort stable
+  # quicksort stable
+
+You can achieve the effect you probably wanted by using C<eval()>
+to defer the pragmas until run time.  Use the quoted argument
+form of C<eval()>, I<not> the BLOCK form, as in
+
+  eval { use sort "_quicksort" }; # WRONG
+
+or the effect will still be at compile time.
+Reset to default options before selecting other subpragmas
+(in case somebody carelessly left them on) and after sorting,
+as a courtesy to others.
+
+  { eval 'use sort qw(defaults _quicksort)'; # force quicksort
+    eval 'no sort "stable"';      # stability not wanted
+    print sort::current . "\n";
+    @a = sort @b;
+    eval 'use sort "defaults"';   # clean up, for others
+  }
+  { eval 'use sort qw(defaults stable)';     # force stability
+    print sort::current . "\n";
+    @c = sort @d;
+    eval 'use sort "defaults"';   # clean up, for others
+  }
+  # prints:
+  # quicksort
+  # stable
+
+Scoping for this pragma may change in future versions.
 
 =cut
 

diff --git a/lib/sort.t b/lib/sort.t
index 9903765..8828083 100644 (file)
--- a/lib/sort.t
+++ b/lib/sort.t
@@ -28,7 +28,8 @@ use warnings;
 use Test::More tests => @TestSizes * 2 # sort() tests
                        * 4             # number of pragmas to test
                        + 1             # extra test for qsort instability
-                       + 3;            # tests for sort::current
+                       + 3             # tests for sort::current
+                       + 3;            # tests for "defaults" and "no sort"
 
 # Generate array of specified size for testing sort.
 #
@@ -160,3 +161,23 @@ eval q{
     main(0);
 };
 die $@ if $@;
+
+# Tests added to check "defaults" subpragma, and "no sort"
+
+eval q{
+    no sort qw(_qsort);
+    is(sort::current(), 'stable', 'sort::current after no _qsort');
+};
+die $@ if $@;
+
+eval q{
+    use sort qw(defaults _qsort);
+    is(sort::current(), 'quicksort', 'sort::current after defaults _qsort');
+};
+die $@ if $@;
+
+eval q{
+    use sort qw(defaults stable);
+    is(sort::current(), 'stable', 'sort::current after defaults stable');
+};
+die $@ if $@;