libwww-search-perl 2.51.50-1 / usr / share / perl5 / WWW / Search / Test.pm

# $rcs = ' $Id: Test.pm,v 2.293 2015-06-06 20:55:31 Martin Exp $ ' ;

=head1 NAME

WWW::Search::Test - utilities to aid in testing WWW::Search backends

=head1 SYNOPSIS

  $oTest = new WWW::Search::Test('HotBot,Yahoo,Excite');
  $oTest->test('HotBot', 'Kingpin', 'one', $sQuery, $TEST_RANGE, 1, 10);

=head1 DESCRIPTION

See file test.pl in the WWW-Search-HotBot distribution for a detailed
"real-world" example.

=head1 METHODS AND FUNCTIONS

=cut

package WWW::Search::Test;

use strict;
use warnings;

use Bit::Vector;
use Carp;
use Config;
use Cwd;
use Data::Dumper;  # for debugging only
use Date::Manip;
use base 'Exporter';
use File::Path;
use File::Slurp;
use File::Spec::Functions qw( :ALL );
use Test::More;
use WWW::Search;

use vars qw( $MODE_DUMMY $MODE_INTERNAL $MODE_EXTERNAL $MODE_UPDATE );
use vars qw( $TEST_DUMMY $TEST_EXACTLY $TEST_BY_COUNTING $TEST_GREATER_THAN $TEST_RANGE );
use vars qw( $iTest $oSearch $sEngine );
# If set, will be used as a filename to save HTML when a test fails:
use vars qw( $sSaveOnError );

use vars qw( @EXPORT );
@EXPORT = qw( eval_test test
              no_test not_working not_working_with_tests not_working_and_abandoned
              $MODE_DUMMY $MODE_INTERNAL $MODE_EXTERNAL $MODE_UPDATE
              $TEST_DUMMY $TEST_EXACTLY $TEST_BY_COUNTING $TEST_GREATER_THAN $TEST_RANGE
              new_engine run_test run_gui_test skip_test count_results
              tm_new_engine tm_run_test tm_run_test_no_approx
            );

use vars qw( $VERSION $bogus_query $websearch );

$VERSION = do { my @r = (q$Revision: 2.293 $ =~ /\d+/g); sprintf "%d."."%03d" x $#r, @r };
$bogus_query = "Bogus" . $$ . "NoSuchWord" . time;

($MODE_DUMMY, $MODE_INTERNAL, $MODE_EXTERNAL, $MODE_UPDATE) = qw(dummy internal external update);
($TEST_DUMMY, $TEST_EXACTLY, $TEST_BY_COUNTING, $TEST_GREATER_THAN, $TEST_RANGE) = (1..10);

use constant DEBUG => 0;

=head2 find_websearch

Returns the full path of an executable WebSearch program,
or undef if none can be found.

=cut

sub find_websearch
  {
  unless ($websearch)
    {
    # Try to find a working WebSearch:
    my $sProg = 'WebSearch';
    my @asTry = ( $sProg );
    # Try local directory, in case . is not in the path:
    push @asTry, catfile(curdir, $sProg);
    push @asTry, catfile(qw( blib script ), $sProg);
    # See if WebSearch.BAT has been created/installed, and try it with
    # explicit 'perl' in front:
    push @asTry, map { ("$_.bat", "$Config{perlpath} $_") } @asTry;
    DEBUG && print STDERR Dumper(\@asTry);
 WEBSEARCH_TRY:
    foreach my $sTry (@asTry)
      {
      my $sCmd = "$sTry --VERSION";
      DEBUG && print STDERR " + W::S::T::find_websearch() cmd ==$sCmd==\n";
      # Turn off warnings:
      local $^W = 0;
      # Wrap it in an eval so we don't die if it fails:
      my @as = split(/\s/, eval{`$sCmd`});
      $websearch = shift(@as) || undef;
      last WEBSEARCH_TRY if $websearch;
      } # foreach
    # Prevent undef warning:
    $websearch ||= '';
    undef $websearch unless ($websearch =~ m/WebSearch/);
    # print STDERR "in WWW::Search::Test, websearch is $websearch\n";
    } # unless
  return $websearch;
  } # find_websearch


=head2 new

Create a new WWW::Search::Test object.
All arguments are strings, names of backends that this object will be able to test.
If no arguments are given, will be able to test all backends.

=cut

sub new
  {
  my $class = shift;
  my $sEngines = join(',', '', @_, '');

  return bless {
                debug => 0,
                engines => $sEngines,
                error_count => 0,
                mode => $MODE_DUMMY,
                verbose => 0,
                # websearch => $websearch, # why do we need this?
               }, $class;
  } # new


=head2 mode

Set / get the test mode of this object.
If an argument is given, sets the mode to that value.
Returns the current (or newly set) value.

There are three test modes available.  They are:

  $MODE_INTERNAL: parse URLs out of saved pages (as a sanity check or regression test);
  $MODE_EXTERNAL: send the query to the search engine "live", parse the results, and compare them to the previously saved results;
and
  $MODE_UPDATE: send the query to the search engine "live", parse the results, and save them for future testing.

=cut

sub mode
  {
  my $self = shift;
  my $new_mode = shift;
  if ($new_mode)
    {
    $self->{'mode'} = $new_mode;
    }
  return $self->{'mode'};
  } # mode

=head2 relevant_test

Given the name of a backend,
returns true if this Test object is able to test that backend.

=cut

sub relevant_test
  {
  my $self = shift;
  return 1 if ($self->{engines} eq ',,');
  my $e = ','.shift().',';
  # print STDERR " + relevant_test($e|", $self->{engines}, ")\n";
  return ($self->{engines} =~ m/$e/);
  } # relevant_test


=head2 eval_test

Given the name of a backend,
grabs the $TEST_CASES variable from that backend and evaluates it.

=cut

sub eval_test
  {
  my $self = shift;
  my $sSE = shift;
  return unless $self->relevant_test($sSE);
  my $o = new WWW::Search($sSE);
  my $iVersion = $o->version;
  my $code = $o->test_cases;
  $code ||= '';
  unless ($code ne '')
    {
    print "  $sSE version $iVersion contains no TEST_CASES\n";
    $self->{error_count}++;
    }
  # print STDERR " BEFORE SUBST: $code\n";
  $code =~ s!&test\(!\$self->test\(!g;
  $code =~ s/&no_test\(/\$self->no_test\(/g;
  $code =~ s/&not_working\(/\$self->not_working\(/g;
  $code =~ s/&not_working_and_abandoned\(/\$self->not_working_and_abandoned\(/g;
  $code =~ s/&not_working_with_tests\(/\$self->not_working_with_tests\(/g;
  # print STDERR " AFTER SUBST: $code\n";
  print "\n";  # put a little space between each engine's results
  eval $code; 
  warn $@ if $@;
  } # eval_test


=head2 test

Run test(s) for a backend.
Arguments are, in order: 
name of a backend to test (string, required);
name of backend maintainer (string, if undef $backend::MAINTAINER will be used);
filename for results storage/comparison (string, required);
query to be sent to backend (string, required);
test method (required, one of the following).

Several test methods are possible:

  $TEST_EXACTLY: list of URLs must match exactly (line for line, in order);
  $TEST_BY_COUNTING: test passes if number of resulting URLs is equal;
  $TEST_GREATER_THAN: test passes if we get more than N result URLs;
and
  $TEST_RANGE: like $TEST_GREATER_THAN but constrained on both ends.

=cut

sub test
  {
  my $self = shift;
  my $sSE = shift;
  my $sM = shift;
  my $file = shift;
  my $query = shift;
  my $test_method = shift;
  print STDERR " + test($sSE,$sM,$file,$query,$test_method)\n" if $self->{debug};
  my ($low_end, $high_end) = @_;
  $low_end ||= 0;
  $high_end ||= 0;
  my $sExpected = $low_end;
  if ($test_method == $TEST_GREATER_THAN)
    {
    $low_end++;
    $sExpected = "$low_end..";
    }
  if (0 < $high_end)
    {
    $sExpected = "$low_end..$high_end";
    }
  return if (!$self->relevant_test($sSE));
  print "  trial $file (", $self->{'mode'}, ")\n";
  if (($self->{'mode'} eq $MODE_INTERNAL) && ($query =~ m/$bogus_query/))
    {
    print "  skipping test on this platform.\n";
    return;
    } # if

  my $pwd = curdir();
  my @asSE = split(/::/, $sSE);
  my $path = catdir($pwd, 'Test-Pages', @asSE);
  mkpath $path;
  if ($self->{'mode'} eq $MODE_UPDATE)
    {
    # Delete all existing test result files for this Engine:
    opendir DIR, $path;
    foreach my $afile (readdir DIR)
      {
      unlink catfile($path, $afile) if $afile =~ m/^$file/;
      } # foreach
    closedir DIR;
    } # if MODE_UPDATE
  # Look at the filename argument we got:
  my ($v,$d,$f) = splitpath($file);
  # If it contains no path element (file name only):
  if ($d eq '')
    {
    # Prepend path onto file:
    $file = catfile($path, $file);
    } # if
  my $o = new WWW::Search($sSE);
  my $version = $o->version;
  print "  ($sSE $version, $sM)\n";
  print STDERR "  expect to find results in $file\n" if $self->{debug};
  my %src = (
             $MODE_INTERNAL => "--option search_from_file=$file",
             $MODE_EXTERNAL => '',
             $MODE_UPDATE => "--option search_to_file=$file",
            );
  # --max 209 added by Martin Thurn 1999-09-27.  We never want to
  # fetch more than three pages, if we can at all help it (or do we?)
  my $websearch = &find_websearch;
  $websearch ||= catfile($pwd, 'blib', 'script', 'WebSearch');
  my $cmd = $Config{'perlpath'} . " -MExtUtils::testlib $websearch ";
  $cmd .= $self->{debug} ? '--debug '.$self->{debug} : '';
  $cmd .= " --max 209 --engine $sSE ". $src{$self->{'mode'}} ." -- $query";
  print "  $cmd\n" if ($self->{verbose} || $self->{debug});
  open(TRIALSTREAM, "$cmd|") || die "$0: cannot run test ($!)\n";
  open(TRIALFILE, ">$file.trial") || die "$0: cannot open $file.trial for writing ($!)\n";
  open(OUTFILE, ">$file.out") || die "$0: cannot open $file.out for writing ($!)\n" if ($self->{'mode'} eq $MODE_UPDATE);
  my $iActual = 0;
  while (<TRIALSTREAM>)
    {
    print TRIALFILE $_;
    $iActual++;
    print OUTFILE $_ if ($self->{'mode'} eq $MODE_UPDATE);
    }
  close TRIALSTREAM;
  close TRIALFILE;
  if ($self->{'mode'} eq $MODE_UPDATE)
    {
    close OUTFILE;
    if (open TS, ">$file.README")
      {
      print TS "This set of test-result pages was created on ", scalar(localtime(time)), "\n";
      close TS;
      } # if
    my $iPageCount = &wc_l($file);
    my $iURLCount = &wc_l("$file.out");
    print "  $query --> $iURLCount urls (should be $sExpected) on $iPageCount pages\n";
    return;
    } # if

  if (-f "$file.out")
    {
    my ($e, $sMsg) = (0, '');
    if ($test_method == $TEST_GREATER_THAN)
      {
      if ($iActual <= $low_end)
        {
        $sMsg .= "expected more than $low_end, but got $iActual; ";
        $e = 1;
        }
      } # TEST_GREATER_THAN
    elsif ($test_method == $TEST_RANGE)
      {
      $sMsg .= "INTERNAL ERROR, low_end has no value; " unless defined($low_end);
      $sMsg .= "INTERNAL ERROR, high_end has no value; " unless defined($high_end);
      $sMsg .= "INTERNAL ERROR, high_end is zero; " unless 0 < $high_end;
      if ($iActual < $low_end)
        {
        $sMsg .= "expected $low_end..$high_end, but got $iActual; ";
        $e = 1;
        }
      if ($high_end < $iActual)
        {
        $sMsg .= "expected $low_end..$high_end, but got $iActual; ";
        $e = 1;
        }
      } # TEST_RANGE
    elsif ($test_method == $TEST_EXACTLY)
      {
      $e = &diff("$file.out", "$file.trial") ? 1 : 0;
      } # TEST_EXACTLY
    elsif ($test_method == $TEST_BY_COUNTING)
      {
      my $iExpected = shift;
      my $iActual = &wc_l("$file.trial");
      if ($iActual != $iExpected)
        {
        $sMsg .= "expected $iExpected, but got $iActual; ";
        $e = 1;
        }
      }
    else
      {
      $e = 0;
      $sMsg = "INTERNAL ERROR, unknown test method $test_method; ";
      }

    if ($e == 0)
      {
      print "  ok.\n";
      unlink("$file.trial");   # clean up
      } 
    elsif ($e == 1)
      {
      print "DIFFERENCE DETECTED: $query --> $sMsg\n";
      $self->{error_count}++;
      }
    else
      {
      print "INTERNAL ERROR $query --> e is $e.\n";
      $self->{error_count}++;
      }
    }
  else
    {
    print "NO SAVED OUTPUT, can not evaluate test results.\n";
    $self->{error_count}++;
    }
  } # test

=head2 no_test

Prints a message stating that this backend does not have a test suite.
Takes two arguments, the backend name and the name of the maintainer.

=cut

sub no_test
  {
  my $self = shift;
  my ($engine, $maint) = @_;
  return unless ($self->relevant_test($engine));
  print <<"NONE";
  trial none ($engine)
  This search engine does not have any tests,
  but report problems with it to $maint.
NONE
  } # no_test


=head2 not_working

Prints a message stating that this backend is known to be broken.
Takes two arguments, the backend name and the name of the maintainer.

=cut

sub not_working
  {
  my $self = shift;
  my ($engine, $maint) = @_;
  return unless ($self->relevant_test($engine));
  print <<"BROKEN";
  trial none ($engine)
  This search engine is known to be non-functional.  
  You are encouraged to investigate the problem and email its maintainer, 
  $maint.
BROKEN
  } # not_working


=head2 not_working_with_tests

Prints a message stating that this backend is known to be broken
even though it has a test suite.
Takes two arguments, the backend name and the name of the maintainer.

=cut

sub not_working_with_tests
  {
  my $self = shift;
  my ($engine, $maint) = @_;
  return if (!$self->relevant_test($engine));
  print <<"KNOWNFAILURE";
  trial none ($engine)
  Test cases for this search engine are known to fail.
  You are encouraged to investigate the problem and email its maintainer,
  $maint.
KNOWNFAILURE
  } # not_working_with_tests


=head2 not_working_and_abandoned

Prints a message stating that this backend is known to be broken
and is not being actively maintained.
Takes two arguments, the backend name and the name of the maintainer.

=cut

sub not_working_and_abandoned
  {
  my $self = shift;
  my ($engine, $maint) = @_;
  return if (!$self->relevant_test($engine));
  print <<"ADOPT";
  trial none ($engine)
  This search engine is known to be non-functional.
  You are encouraged to adopt it from its last known maintainer,
  $maint.
ADOPT
  } # not_working_and_abandoned


=head2 reset_error_count

Reset the counter of errors to zero.
You probably want to call this before each call to test() or eval_test().

=cut

sub reset_error_count
  {
  my $self = shift;
  $self->{error_count} = 0;
  } # reset_error_count


=head2 wc_l (private, not a method)

Given a filename, count the number of lines of text contained
within the file.  
(I.e. simulate running UNIX command C<wc -l> on a file)

=cut

sub wc_l
  {
  # SPECIAL CASE: If first line is "Nothing found.", report 0 lines.
  open WC, shift or return 0;
  $/ = "\n";
  my $i = 0;
  while (<WC>)
    {
    last if /Nothing found./;
    $i++;
    } # while
  return $i;
  } # wc_l


=head2 diff (private, not a method)

Given two files, returns TRUE if contents are line-by-line
different, or FALSE if contents are line-by-line same.
(I.e. like the UNIX command diff, but just reports true or false)

=cut

sub diff
  {
  open DIFF1, shift or return 91;
  open DIFF2, shift or return 92;
  my $iResult = 0;
  $/ = "\n";
  while ((defined(my $s1 = <DIFF1>)) &&
         ($iResult ne 1))
    {
    my $s2 = <DIFF2>;
    unless (defined($s2))
      {
      $iResult = 1;
      last;
      }
    chomp $s1;
    chomp $s2;
    if ($s1 ne $s2)
      {
      $iResult = 1;
      last;
      }
    } # while
  close DIFF1;
  close DIFF2;
  return $iResult;
  } # diff


=head2 Shortcuts for running backend tests

WWW::Search::Test keeps its own count of test numbers,
so if you want to mix-and-match these functions with your own tests,
use the $WWW::Search::Test::iTest counter.

=head2 new_engine

One argument: the name of a backend suitable to be passed to WWW::Search::new().
Prints 'ok' or 'not ok' and the test number.
Creates a WWW::Search object internally,
to be used for all subsequent calls to run_test and run_gui_test (see below).

=cut

sub new_engine
  {
  $iTest++;
  $sEngine = shift;
  $oSearch = new WWW::Search($sEngine);
  print ref($oSearch) ? '' : 'not ';
  print "ok $iTest\n";
  $oSearch->env_proxy('yes');
  } # new_engine

=head2 tm_new_engine

Same as new_engine(), but uses Test::More instead of just printing 'ok'.

=cut

sub tm_new_engine
  {
  my $sEngine = shift;
  $oSearch = new WWW::Search($sEngine);
  Test::More::ok(ref($oSearch), "instantiate WWW::Search::$sEngine object");
  $oSearch->env_proxy('yes');
  } # tm_new_engine

=head2 run_test

Three arguments: a query string, NOT escaped;
a minimum number of expected results; and
a maximum number of expected results.
Optional fourth argument: integer value to be used as the search_debug.
Optional fifth argument: send any true value to dump the search results.
Optional sixth argument: reference to hash of search options (see backend documentation).
Optional seventh argument: send any true value to NOT escape the query string.

If the minimum is undef, assumes zero.
If the maximum is undef, does not check.

Prints 'ok' or 'not ok' and the test number.

=cut

sub run_test
  {
  return &_run_our_test('normal', @_);
  } # run_test

=head2 run_gui_test

Same as run_test(), but calls gui_query() instead of native_query().

=cut

sub run_gui_test
  {
  return &_run_our_test('gui', @_);
  } # run_gui_test


=head2 tm_run_test

Same as run_test(), but uses Test::More rather than just printing 'ok'.

Note: If you use this function inside a TODO block,
you must set global variable $TODO rather than a local $TODO,
and you must set the global $TODO back to empty-string (or undef) at the end of your TODO block.
For example:

  TODO:
    {
    $TODO = 'I have not fixed this yet';
    tm_run_test(...);
    $TODO = '';
    } # end of TODO block

=cut

sub tm_run_test
  {
  _tm_run_test(@_, 1);
  } # tm_run_test

sub _tm_run_test
  {
  # Last argument is boolean, whether to check approx_result_count:
  my $iApprox = pop(@_) || 0;
  # Remaining args, same as count_results():
  my ($sType, $sQuery, $iMin, $iMax) = @_;
  my $iCount = count_results(@_);
  my $iAnyFailure = 0;
  $iAnyFailure++ unless Test::More::is($oSearch->response->code, 200, 'got valid HTTP response');
  if (defined $iMin)
    {
    $iAnyFailure++ unless Test::More::cmp_ok($iMin, '<=', $iCount,
                                             qq{lower-bound num-hits for query=$sQuery});
    if ($iApprox)
      {
      $iAnyFailure++ unless Test::More::cmp_ok($iMin, '<=', $oSearch->approximate_result_count,
                                               qq{lower-bound approximate_result_count});
      } # if
    } # if
  if (defined $iMax)
    {
    $iAnyFailure++ unless Test::More::cmp_ok($iCount, '<=', $iMax,
                                             qq{upper-bound num-hits for query=$sQuery});
    if ($iApprox)
      {
      $iAnyFailure++ unless Test::More::cmp_ok($oSearch->approximate_result_count, '<=', $iMax,
                                               qq{upper-bound approximate_result_count});
      } # if
    } # if
  $sSaveOnError ||= q'';
  if ($iAnyFailure && ($sSaveOnError ne q''))
    {
    write_file($sSaveOnError, { err_mode => 'quiet'}, $oSearch->response->content);
    Test::More::diag(qq'HTML was saved in $sSaveOnError');
    } # if
  } # _tm_run_test

=head2 tm_run_test_no_approx

Same as tm_run_test, but does NOT check the approximate_result_count.

=cut

sub tm_run_test_no_approx
  {
  _tm_run_test(@_, 0);
  } # tm_run_test_no_approx

=head2 count_results

Run a query, and return the actual (not approximate) number of hits.
Required first argument determines which backend query method to call: 'gui' to call gui_query(), anything else to call native_query().
Remaining arguments are same as all the run_test() arguments.

=cut

sub count_results
  {
  my ($sType, $sQuery, $iMin, $iMax, $iDebug, $iPrintResults, $rh, $iDoNotEscape) = @_;
  # print STDERR qq{ DDD count_results raw args($sType,$sQuery,$iMin,$iMax,$iDebug,$iPrintResults,$rh,$iDoNotEscape)\n};
  $iDebug ||= 0;
  $iPrintResults ||= 0;
  $rh->{'search_debug'} = $iDebug;
  carp ' --- min/max values out of order?' if (defined($iMin) && defined($iMax) && ($iMax < $iMin));
  $oSearch->reset_search;
  $iMin ||= 0;
  # While $iMax is the number the user wants to compare, $iMaxAbs is
  # the actual number we apply to the search:
  my $iMaxAbs;
  if (! defined($iMax))
    {
    # User said upper limit is 'undef'; just make sure we get the
    # mininum:
    $iMaxAbs = $iMin + 1;
    } # if
  else
    {
    # Give a little breathing room, so we'll notice if there are too
    # many returned:
    $iMaxAbs = $iMax + 1;
    }
  $oSearch->maximum_to_retrieve($iMaxAbs);
  $iTest++;
  $sQuery = WWW::Search::escape_query($sQuery) unless $iDoNotEscape;
  # print STDERR " + in WWW::Search::Test::count_results, iDebug = $iDebug\n";
  if ($sType eq 'gui')
    {
    $oSearch->gui_query($sQuery, $rh);
    }
  else
    {
    $oSearch->native_query($sQuery, $rh);
    }
  $oSearch->login($ENV{WWW_SEARCH_USERNAME}, $ENV{WWW_SEARCH_PASSWORD});
  my @aoResults = $oSearch->results();
  if ($iPrintResults)
    {
    my $i = 1;
    foreach my $oResult (@aoResults)
      {
      print $i++, '. ', $oResult->url, "\n";
      foreach my $sField (qw( title description score change_date index_date size company location source ))
        {
        print "  $sField==", $oResult->$sField, "==\n" if defined($oResult->$sField);
        } # foreach
      } # foreach
    } # if
  return scalar(@aoResults);
  } # count_results


sub _run_our_test
  {
  my ($sType, $sQuery, $iMin, $iMax, $iDebug, $iPrintResults) = @_;
  my $iResults = &count_results(@_);
  my $sExpect;
  if (! defined($iMax))
    {
    $sExpect = "more than $iMin";
    }
  elsif (! defined($iMin))
    {
    $sExpect = "fewer than $iMax";
    }
  else
    {
    $sExpect = "$iMin..$iMax";
    }
  $iMax = 999999 unless defined ($iMax);
  if (($iResults < $iMin) || ($iMax < $iResults))
    {
    print STDERR " --- got $iResults results for $sType $sEngine query '$sQuery', but expected $sExpect\n";
    print STDOUT 'not ';
    } # if
  print STDOUT "ok $iTest\n";
  } # _run_our_test


=head2 skip_test

You can call this function instead of run_test() or run_gui_test()
if the current test must be skipped for any reason.

=cut

sub skip_test
  {
  $iTest++;
  print STDOUT "skip $iTest\n";
  } # skip_test


=head2 test_most_results

Given an arrayref of things to test,
runs all those things against all the results of the most-recently executed test search.

=cut

sub test_most_results
  {
  my $rara = shift;
  my $fPct = shift || 0.80;
  my $iCount = scalar(@$rara);
  my $iAnyFailed = my $iResult = 0;
  my %hioExemplar;
  my %hiiFailed;
  # Create a bit vector large enough to hold one bit for each test:
  my $oV = new Bit::Vector($iCount);
  # Turn on all the bits (we will turn off bits when tests fail):
  $oV->Fill;
  my $iVall = $oV->to_Dec;
  my $sCodeAll = q{};
  my $iTest = 0;
 TEST:
  foreach my $ra (@$rara)
    {
    # print STDERR " DDD ra is ", Dumper($ra);
    my ($sField, $sCmp, $sValue, $sDesc) = @$ra;
    $sDesc ||= qq{test #$iTest};
    my $sCode;
    if ($sCmp eq 'like')
      {
      $sCode = "(\$oResult->$sField =~ m!$sValue!)";
      } # if
    elsif ($sCmp eq 'unlike')
      {
      $sCode = "(\$oResult->$sField !~ m!$sValue!)";
      } # if
    elsif ($sCmp eq 'date')
      {
      $sCode = "((ParseDate(\$oResult->$sField) || '') ne q{})";
      } # if
    else
      {
      $sCode = "(\$oResult->$sField $sCmp $sValue)";
      }
    $sCode = <<"ENDCODE";
if (! $sCode)
  {
  \$oV->Bit_Off($iTest);
  \$hiiFailed{'$sDesc'}++;
  } # if
ENDCODE
    $sCodeAll .= $sCode;
    $iTest++;
    } # foreach TEST
  $sCodeAll .= "1;\n";
  # print STDERR " DDD the test is ===$sCodeAll===\n";
 RESULT:
  foreach my $oResult ($oSearch->results())
    {
    $iResult++;
    # Turn on all the bits (we will turn off bits when tests fail):
    $oV->Fill;
    # print STDERR " DDD eval the test...\n";
    if (! eval $sCodeAll)
      {
      print STDERR $@;
      } # if
    # Now look at the value of the Bit::Vector after running the tests:
    my $iV = $oV->to_Dec;
    if ($iV < $iVall)
      {
      # At least one of the bits got turned off (i.e. a test failed):
      $hioExemplar{$iV} = $oResult;
      $iAnyFailed++;
      # For debugging:
      # print STDERR Dumper($oResult);
      # last RESULT;
      } # if
    } # foreach RESULT
  ok($iResult, qq{got more than zero results ($iResult, to be exact)});
  # Now make sure all the sub-tests passed at least N% of the time.
  # We only need to look at sub-tests that had any failures (sub-tests
  # with no failures are 100% correct, so there's no need to check
  # them 8-)
  while (my ($sItem, $iFailed) = each %hiiFailed)
    {
    my $fPctFailed = ($iFailed / $iResult);
    ok($fPctFailed < (1 - $fPct), sprintf(qq{%0.1f%% of '%s' tests failed}, $fPctFailed * 100, $sItem));
    } # while
  if ($iAnyFailed)
    {
    Test::More::diag(" Here are result(s) that exemplify test failure(s):");
    foreach my $oResult (values %hioExemplar)
      {
      Test::More::diag(Dumper($oResult));
      } # while
    } # if
  } # test_most_results

1;

__END__