summaryrefslogtreecommitdiffstats
path: root/contrib/perl5/t/lib/tie-refhash.t
blob: d80b2e10fc9d00f33bc50887e5781888a9bd8a22 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
#!/usr/bin/perl -w
# 
# Basic test suite for Tie::RefHash and Tie::RefHash::Nestable.
# 
# The testing is in two parts: first, run lots of tests on both a tied
# hash and an ordinary un-tied hash, and check they give the same
# answer.  Then there are tests for those cases where the tied hashes
# should behave differently to normal hashes, that is, when using
# references as keys.
# 

BEGIN {
    chdir 't' if -d 't';
    @INC = '.'; 
    push @INC, '../lib';
}    

use strict;
use Tie::RefHash;
use Data::Dumper;
my $numtests = 34;
my $currtest = 1;
print "1..$numtests\n";

my $ref = []; my $ref1 = [];

# Test standard hash functionality, by performing the same operations
# on a tied hash and on a normal hash, and checking that the results
# are the same.  This does of course assume that Perl hashes are not
# buggy :-)
# 
my @tests = standard_hash_tests();

my @ordinary_results = runtests(\@tests, undef);
foreach my $class ('Tie::RefHash', 'Tie::RefHash::Nestable') {
    my @tied_results = runtests(\@tests, $class);
    my $all_ok = 1;

    die if @ordinary_results != @tied_results;
    foreach my $i (0 .. $#ordinary_results) {
        my ($or, $ow, $oe) = @{$ordinary_results[$i]};
        my ($tr, $tw, $te) = @{$tied_results[$i]};
        
        my $ok = 1;
        local $^W = 0;
        $ok = 0 if (defined($or) != defined($tr)) or ($or ne $tr);
        $ok = 0 if (defined($ow) != defined($tw)) or ($ow ne $tw);
        $ok = 0 if (defined($oe) != defined($te)) or ($oe ne $te);
        
        if (not $ok) {
            print STDERR
              "failed for $class: $tests[$i]\n",
              "ordinary hash gave:\n",
              defined $or ? "\tresult:    $or\n" : "\tundef result\n",
              defined $ow ? "\twarning:   $ow\n" : "\tno warning\n",
              defined $oe ? "\texception: $oe\n" : "\tno exception\n",
              "tied $class hash gave:\n",
              defined $tr ? "\tresult:    $tr\n" : "\tundef result\n",
              defined $tw ? "\twarning:   $tw\n" : "\tno warning\n",
              defined $te ? "\texception: $te\n" : "\tno exception\n",
              "\n";
            $all_ok = 0;
        }
    }
    test($all_ok);
}

# Now test Tie::RefHash's special powers
my (%h, $h);
$h = eval { tie %h, 'Tie::RefHash' };
warn $@ if $@;
test(not $@);
test(ref($h) eq 'Tie::RefHash');
test(defined(tied(%h)) and tied(%h) =~ /^Tie::RefHash/);
$h{$ref} = 'cholet';
test($h{$ref} eq 'cholet');
test(exists $h{$ref});
test((keys %h) == 1);
test(ref((keys %h)[0]) eq 'ARRAY');
test((keys %h)[0] eq $ref);
test((values %h) == 1);
test((values %h)[0] eq 'cholet');
my $count = 0;
while (my ($k, $v) = each %h) {
    if ($count++ == 0) {
        test(ref($k) eq 'ARRAY');
        test($k eq $ref);
    }
}
test($count == 1);
delete $h{$ref};
test(not defined $h{$ref});
test(not exists($h{$ref}));
test((keys %h) == 0);
test((values %h) == 0);
undef $h;
untie %h;

# And now Tie::RefHash::Nestable's differences from Tie::RefHash.
$h = eval { tie %h, 'Tie::RefHash::Nestable' };
warn $@ if $@;
test(not $@);
test(ref($h) eq 'Tie::RefHash::Nestable');
test(defined(tied(%h)) and tied(%h) =~ /^Tie::RefHash::Nestable/);
$h{$ref}->{$ref1} = 'bungo';
test($h{$ref}->{$ref1} eq 'bungo');

# Test that the nested hash is also tied (for current implementation)
test(defined(tied(%{$h{$ref}}))
     and tied(%{$h{$ref}}) =~ /^Tie::RefHash::Nestable=/ );

test((keys %h) == 1);
test((keys %h)[0] eq $ref);
test((keys %{$h{$ref}}) == 1);
test((keys %{$h{$ref}})[0] eq $ref1);


die "expected to run $numtests tests, but ran ", $currtest - 1
  if $currtest - 1 != $numtests;

@tests = ();
undef $ref;
undef $ref1;

exit();


# Print 'ok X' if true, 'not ok X' if false
# Uses global $currtest.
# 
sub test {
    my $t = shift;
    print 'not ' if not $t;
    print 'ok ', $currtest++, "\n";
}


# Wrapper for Data::Dumper to 'dump' a scalar as an EXPR string. 
sub dumped {
    my $s = shift;
    my $d = Dumper($s);
    $d =~ s/^\$VAR1 =\s*//;
    $d =~ s/;$//;
    chomp $d;
    return $d;
}

# Crudely dump a hash into a canonical string representation (because
# hash keys can appear in any order, Data::Dumper may give different
# strings for the same hash).
# 
sub dumph {
    my $h = shift;
    my $r = '';
    foreach (sort keys %$h) {
        $r = dumped($_) . ' => ' . dumped($h->{$_}) . "\n";
    }
    return $r;
}

# Run the tests and give results.
# 
# Parameters: reference to list of tests to run
#             name of class to use for tied hash, or undef if not tied
# 
# Returns: list of [R, W, E] tuples, one for each test.
# R is the return value from running the test, W any warnings it gave,
# and E any exception raised with 'die'.  E and W will be tidied up a
# little to remove irrelevant details like line numbers :-)
# 
# Will also run a few of its own 'ok N' tests.
# 
sub runtests {
    my ($tests, $class) = @_;
    my @r;

    my (%h, $h);
    if (defined $class) {
        $h = eval { tie %h, $class };
        warn $@ if $@;
        test(not $@);
        test(ref($h) eq $class);
        test(defined(tied(%h)) and tied(%h) =~ /^\Q$class\E/);
    }

    foreach (@$tests) {
        my ($result, $warning, $exception);
        local $SIG{__WARN__} = sub { $warning .= $_[0] };
        $result = scalar(eval $_);
        if ($@)
         {
          die "$@:$_" unless defined $class;
          $exception = $@;
         }

        foreach ($warning, $exception) {
            next if not defined;
            s/ at .+ line \d+\.$//mg;
            s/ at .+ line \d+, at .*//mg;
            s/ at .+ line \d+, near .*//mg;
        }

        my (@warnings, %seen);
        foreach (split /\n/, $warning) {
            push @warnings, $_ unless $seen{$_}++;
        }
        $warning = join("\n", @warnings);

        push @r, [ $result, $warning, $exception ];
    }

    return @r;
}


# Things that should work just the same for an ordinary hash and a
# Tie::RefHash.
# 
# Each test is a code string to be eval'd, it should do something with
# %h and give a scalar return value.  The global $ref and $ref1 may
# also be used.
# 
# One thing we don't test is that the ordering from 'keys', 'values'
# and 'each' is the same.  You can't reasonably expect that.
# 
sub standard_hash_tests {
    my @r;

    # Library of standard tests on keys, values and each
    my $STD_TESTS = <<'END'
    join $;, sort keys %h;
    join $;, sort values %h;
    { my ($v, %tmp); $tmp{$v}++ while (defined($v = each %h)); dumph(\%tmp) }
    { my ($k, $v, %tmp); $tmp{"$k$;$v"}++ while (($k, $v) = each %h); dumph(\%tmp) }
END
  ;
    
    # Tests on the existence of the element 'foo'
    my $FOO_TESTS = <<'END'
    defined $h{foo};
    exists $h{foo};
    $h{foo};    
END
  ;

    # Test storing and deleting 'foo'
    push @r, split /\n/, <<"END"
    $STD_TESTS;
    $FOO_TESTS;
    \$h{foo} = undef;
    $STD_TESTS;
    $FOO_TESTS;
    \$h{foo} = 'hello';
    $STD_TESTS;
    $FOO_TESTS;
    delete  \$h{foo};
    $STD_TESTS;
    $FOO_TESTS;
END
  ;

    # Test storing and removing under ordinary keys
    my @things = ('boink', 0, 1, '', undef);
    foreach my $key (map { dumped($_) } @things) {
        foreach my $value ((map { dumped($_) } @things), '$ref') {
            push @r, split /\n/, <<"END"
            \$h{$key} = $value;
            $STD_TESTS;
            defined \$h{$key};
            exists \$h{$key};
            \$h{$key};
            delete \$h{$key};
            $STD_TESTS;
            defined \$h{$key};
            exists \$h{$key};
            \$h{$key};
END
  ;
        }
    }
    
    # Test hash slices
    my @slicetests;
    @slicetests = split /\n/, <<'END'
    @h{'b'} = ();
    @h{'c'} = ('d');
    @h{'e'} = ('f', 'g');
    @h{'h', 'i'} = ();
    @h{'j', 'k'} = ('l');
    @h{'m', 'n'} = ('o', 'p');
    @h{'q', 'r'} = ('s', 't', 'u');
END
  ;
    my @aaa = @slicetests;
    foreach (@slicetests) {
        push @r, $_;
        push @r, split(/\n/, $STD_TESTS);
    }

    # Test CLEAR
    push @r, '%h = ();', split(/\n/, $STD_TESTS);

    return @r;
}

OpenPOWER on IntegriCloud