summaryrefslogtreecommitdiff
path: root/triehash.pl
blob: c722cc62d861ca615f83015ff7b590479345b055 (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
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
#!/usr/bin/perl -w
#
# Copyright (C) 2016 Julian Andres Klode <jak@jak-linux.org>
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
# THE SOFTWARE.


=head1 NAME

triehash - Generate a perfect hash function derived from a trie.

=cut

use strict;
use warnings;
use Getopt::Long;

=head1 SYNOPSIS

B<triehash> [S<I<option>>] [S<I<input file>>]

=head1 DESCRIPTION

triehash takes a list of words in input file and generates a function and
an enumeration to describe the word

=head1 INPUT FILE FORMAT

The file consists of multiple lines of the form:

    [label ~ ] word [= value]

This maps word to value, and generates an enumeration with entries of the form:

    label = value

If I<label> is undefined, the word will be used, the minus character will be
replaced by an underscore. If value is undefined it is counted upwards from
the last value.

There may also be one line of the format

    [ label ~] = value

Which defines the value to be used for non-existing keys. Note that this also
changes default value for other keys, as for normal entries. So if you place

    = 0

at the beginning of the file, unknown strings map to 0, and the other strings
map to values starting with 1. If label is not specified, the default is
I<Unknown>.

=head1 OPTIONS

=over 4

=item B<-C>I<.c file> B<--code>=I<.c file>

Generate code in the given file.

=item B<-H>I<header file> B<--header>=I<header file>

Generate a header in the given file, containing a declaration of the hash
function and an enumeration.

=item B<--enum-name=>I<word>

The name of the enumeration.

=item B<--function-name=>I<word>

The name of the function.

=item B<--namespace=>I<name>

Put the function and enum into a namespace (C++)

=item B<--class=>I<name>

Put the function and enum into a class (C++)

=item B<--enum-class>

Generate an enum class instead of an enum (C++)

=item B<--counter-name=>I<name>

Use I<name> for a counter that is set to the latest entry in the enumeration
+ 1. This can be useful for defining array sizes.

=item B<--extern-c>

Wrap everything into an extern "C" block. Not compatible with the C++
options, as a header with namespaces, classes, or enum classes is not
valid C.

=item B<--multi-byte>=I<value>

Generate code reading multiple bytes at once. The value is a string of power
of twos to enable. The default value is 320 meaning that 8, 4, and single byte
reads are enabled. Specify 0 to disable multi-byte completely, or add 2 if you
also want to allow 2-byte reads. 2-byte reads are disabled by default because
they negatively affect performance on older Intel architectures.

This generates code for both multiple bytes and single byte reads, but only
enables the multiple byte reads of GNU C compatible compilers, as the following
extensions are used:

=over 8

=item Byte-aligned integers

We must be able to generate integers that are aligned to a single byte using:

    typedef uint64_t __attribute__((aligned (1))) triehash_uu64;

=item Byte-order

The macros __BYTE_ORDER__ and __ORDER_LITTLE_ENDIAN__ must be defined.

=back

We forcefully disable multi-byte reads on platforms where the variable
I<__ARM_ARCH> is defined and I<__ARM_FEATURE_UNALIGNED> is not defined,
as there is a measurable overhead from emulating the unaligned reads on
ARM.

=item B<--language=>I<language>

Generate a file in the specified language. Currently known are 'C' and 'tree',
the latter generating a tree.

=item B<--include=>I<header>

Add the header to the include statements of the header file. The value must
be surrounded by quotes or angle brackets for C code. May be specified multiple
times.

=back

=cut

my $unknown = -1;
my $unknown_label = "Unknown";
my $counter_start = 0;
my $enum_name = "PerfectKey";
my $function_name = "PerfectHash";
my $enum_class = 0;

my $code_name = "-";
my $header_name = "-";
my $code;
my $header;
my $ignore_case = 0;
my $multi_byte = "320";
my $language = 'C';
my $counter_name = undef;
my @includes = ();


Getopt::Long::config('default',
                     'bundling',
                     'no_getopt_compat',
                     'no_auto_abbrev',
                     'permute',
                     'auto_help');

GetOptions ("code|C=s" => \$code_name,
            "header|H=s"   => \$header_name,
            "function-name=s" => \$function_name,
            "ignore-case" => \$ignore_case,
            "enum-name=s" => \$enum_name,
            "language|l=s" => \$language,
            "multi-byte=s" => \$multi_byte,
            "enum-class" => \$enum_class,
            "include=s" => \@includes,
            "counter-name=s" => \$counter_name)
    or die("Could not parse options!");


package Trie {

    sub new {
        my $class = shift;
        my $self = {};
        bless $self, $class;

        $self->{children} = {};
        $self->{value} = undef;
        $self->{label} = undef;

        return $self;
    }

    # Return the largest power of 2 smaller or equal to the argument
    sub alignpower2 {
        my ($self, $length) = @_;

        return 8 if ($length >= 8 && $multi_byte =~ /3/);
        return 4 if ($length >= 4 && $multi_byte =~ /2/);
        return 2 if ($length >= 2 && $multi_byte =~ /1/);

        return 1;
    }

    # Split the key into a head block and a tail
    sub split_key {
        my ($self, $key) = @_;
        my $length = length $key;
        my $split = $self->alignpower2($length);

        return (substr($key, 0, $split), substr($key, $split));
    }

    sub insert {
        my ($self, $key, $label, $value) = @_;

        if (length($key) == 0) {
            $self->{label} = $label;
            $self->{value} = $value;
            return;
        }

        my ($child, $tail) = $self->split_key($key);

        $self->{children}{$child} = Trie->new if (!defined($self->{children}{$child}));

        $self->{children}{$child}->insert($tail, $label, $value);
    }

    sub filter_depth {
        my ($self, $togo) = @_;

        my $new = Trie->new;

        if ($togo != 0) {
            my $found = 0;
            foreach my $key (sort keys %{$self->{children}}) {
                if ($togo > length($key) || defined $self->{children}{$key}->{value}) {
                    my $child = $self->{children}{$key}->filter_depth($togo - length($key));

                    $new->{children}{$key}= $child if defined $child;
                    $found = 1 if defined $child;
                }
            }
            return undef if (!$found);
        } else {
            $new->{value} = $self->{value};
            $new->{label} = $self->{label};
        }

        return $new;
    }

    # Reinsert all value nodes into the specified $trie, prepending $prefix
    # to their $paths.
    sub reinsert_value_nodes_into {
        my ($self, $trie, $prefix) = @_;

        $trie->insert($prefix, $self->{label}, $self->{value}) if (defined $self->{value});

        foreach my $key (sort keys %{$self->{children}}) {
            $self->{children}{$key}->reinsert_value_nodes_into($trie, $prefix . $key);
        }
    }

    # Find an earlier split due a an ambiguous character
    sub find_ealier_split {
        my ($self, $key) = @_;

        if ($ignore_case) {
            for my $i (0..length($key)-1) {
                # If the key starts with an ambiguous character, we need to
                # take only it. Otherwise, we need to take everything
                # before the character.
                return $self->alignpower2($i || 1) if (main::ambiguous(substr($key, $i, 1)));
            }
        }
        return $self->alignpower2(length $key);
    }

    # Rebuild the trie, splitting at ambigous chars, and unifying key lengths
    sub rebuild_tree {
        my $self = shift;
        # Determine if/where we need to split before an ambiguous character
        my $new_split = 99999999999999999;
        foreach my $key (sort keys %{$self->{children}}) {
            my $special_length = $self->find_ealier_split($key);
            $new_split = $special_length if ($special_length < $new_split);
        }

        # Start building a new uniform trie
        my $newself = Trie->new;
        $newself->{label} = $self->{label};
        $newself->{value} = $self->{value};
        $newself->{children} = {};

        foreach my $key (sort keys %{$self->{children}}) {
            my $head = substr($key, 0, $new_split);
            my $tail = substr($key, $new_split);
            # Rebuild the child node at $head, pushing $tail downwards
            $newself->{children}{$head} //= Trie->new;
            $self->{children}{$key}->reinsert_value_nodes_into($newself->{children}{$head}, $tail);
            # We took up to one special character of each key label. There might
            # be more, so we need to rebuild recursively.
            $newself->{children}{$head} = $newself->{children}{$head}->rebuild_tree();
        }

        return $newself;
    }
}

# Code generator for C and C++
package CCodeGen {
    my $static = ($code_name eq $header_name) ? "static" : "";
    my $enum_specifier = $enum_class ? "enum class" : "enum";

    sub new {
        my $class = shift;
        my $self = {};
        bless $self, $class;

        return $self;
    }

    sub open_output {
        my $self = shift;
        if ($code_name ne "-") {
            open($code, '>', $code_name) or die "Cannot open $code_name: $!" ;
        } else {
            $code = *STDOUT;
        }
        if($code_name eq $header_name) {
            $header = $code;
        } elsif ($header_name ne "-") {
            open($header, '>', $header_name) or die "Cannot open $header_name: $!" ;
        } else {
            $header = *STDOUT;
        }
    }

    sub word_to_label {
        my ($class, $word) = @_;

        $word =~ s/_/__/g;
        $word =~ s/-/_/g;
        return $word;
    }

    # Return a case label, by shifting and or-ing bytes in the word
    sub case_label {
        my ($self, $key) = @_;

        return sprintf("'%s'", substr($key, 0, 1)) if not $multi_byte;

        my $output = '0';

        for my $i (0..length($key)-1) {
            $output .= sprintf("| onechar('%s', %d, %d)", substr($key, $i, 1), 8 * $i, 8*length($key));
        }

        return $output;
    }

    # Return an appropriate read instruction for $length bytes from $offset
    sub switch_key {
        my ($self, $offset, $length) = @_;

        return "string[$offset]" if $length == 1;
        return sprintf("*((triehash_uu%s*) &string[$offset])", $length * 8);
    }

    sub print_table {
        my ($self, $trie, $fh, $indent, $index) = @_;
        $indent //= 0;
        $index //= 0;

        if (defined $trie->{value}) {
            printf $fh ("    " x $indent . "return %s;\n", ($enum_class ? "${enum_name}::" : "").$trie->{label});
            return;
        }

        # The difference between lowercase and uppercase alphabetical characters
        # is that they have one bit flipped. If we have alphabetical characters
        # in the search space, and the entire search space works fine if we
        # always turn on the flip, just OR the character we are switching over
        # with the bit.
        my $want_use_bit = 0;
        my $can_use_bit = 1;
        my $key_length = 0;
        foreach my $key (sort keys %{$trie->{children}}) {
            $can_use_bit &= not main::ambiguous($key);
            $want_use_bit |= ($key =~ /^[a-zA-Z]+$/);
            $key_length = length($key);
        }

        if ($ignore_case && $can_use_bit && $want_use_bit) {
            printf $fh (("    " x $indent) . "switch(%s | 0x%s) {\n", $self->switch_key($index, $key_length), "20" x $key_length);
        } else {
            printf $fh (("    " x $indent) . "switch(%s) {\n", $self->switch_key($index, $key_length));
        }

        my $notfirst = 0;
        foreach my $key (sort keys %{$trie->{children}}) {
            if ($notfirst) {
                printf $fh ("    " x $indent . "    break;\n");
            }
            if ($ignore_case) {
                printf $fh ("    " x $indent . "case %s:\n", $self->case_label(lc($key)));
                printf $fh ("    " x $indent . "case %s:\n", $self->case_label(uc($key))) if lc($key) ne uc($key) && !($can_use_bit && $want_use_bit);
            } else {
                printf $fh ("    " x $indent . "case %s:\n", $self->case_label($key));
            }

            $self->print_table($trie->{children}{$key}, $fh, $indent + 1, $index + length($key));

            $notfirst=1;
        }

        printf $fh ("    " x $indent . "}\n");
    }

    sub print_words {
        my ($self, $trie, $fh, $indent, $sofar) = @_;

        $indent //= 0;
        $sofar //= "";


        printf $fh ("    " x $indent."%s = %s,\n", $trie->{label}, $trie->{value}) if defined $trie->{value};

        foreach my $key (sort keys %{$trie->{children}}) {
            $self->print_words($trie->{children}{$key}, $fh, $indent, $sofar . $key);
        }
    }

    sub print_functions {
        my ($self, $trie, %lengths) = @_;
        foreach my $local_length (sort { $a <=> $b } (keys %lengths)) {
            print $code ("static enum ${enum_name} ${function_name}${local_length}(const char *string)\n");
            print $code ("{\n");
            $self->print_table($trie->filter_depth($local_length)->rebuild_tree(), $code, 1);
            printf $code ("    return %s$unknown_label;\n", ($enum_class ? "${enum_name}::" : ""));
            print $code ("}\n");
        }
    }

    sub main {
        my ($self, $trie, $num_values, %lengths) = @_;
        print $header ("#ifndef TRIE_HASH_${function_name}\n");
        print $header ("#define TRIE_HASH_${function_name}\n");
        print $header ("#include <stddef.h>\n");
        print $header ("#include <stdint.h>\n");
        foreach my $include (@includes) {
            print $header ("#include $include\n");
        }
        printf $header ("enum { $counter_name = $num_values };\n") if (defined($counter_name));
        print $header ("${enum_specifier} ${enum_name} {\n");
        $self->print_words($trie, $header, 1);
        printf $header ("    $unknown_label = $unknown,\n");
        print $header ("};\n");
        print $header ("$static enum ${enum_name} ${function_name}(const char *string, size_t length);\n");

        print $code ("#include \"$header_name\"\n") if ($header_name ne $code_name);

        if ($multi_byte) {
            print $code ("#ifdef __GNUC__\n");
            for (my $i=16; $i <= 64; $i *= 2) {
                print $code ("typedef uint${i}_t __attribute__((aligned (1))) triehash_uu${i};\n");
                print $code ("typedef char static_assert${i}[__alignof__(triehash_uu${i}) == 1 ? 1 : -1];\n");
            }

            print $code ("#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__\n");
            print $code ("#define onechar(c, s, l) (((uint64_t)(c)) << (s))\n");
            print $code ("#else\n");
            print $code ("#define onechar(c, s, l) (((uint64_t)(c)) << (l-8-s))\n");
            print $code ("#endif\n");
            print $code ("#if (!defined(__ARM_ARCH) || defined(__ARM_FEATURE_UNALIGNED)) && !defined(TRIE_HASH_NO_MULTI_BYTE)\n");
            print $code ("#define TRIE_HASH_MULTI_BYTE\n");
            print $code ("#endif\n");
            print $code ("#endif /*GNUC */\n");

            print $code ("#ifdef TRIE_HASH_MULTI_BYTE\n");
            $self->print_functions($trie, %lengths);
            $multi_byte = 0;
            print $code ("#else\n");
            $self->print_functions($trie, %lengths);
            print $code ("#endif /* TRIE_HASH_MULTI_BYTE */\n");
        } else {
            $self->print_functions($trie, %lengths);
        }

        print $code ("$static enum ${enum_name} ${function_name}(const char *string, size_t length)\n");
        print $code ("{\n");
        print $code ("    switch (length) {\n");
        foreach my $local_length (sort { $a <=> $b } (keys %lengths)) {
            print $code ("    case $local_length:\n");
            print $code ("        return ${function_name}${local_length}(string);\n");
        }
        print $code ("    default:\n");
        printf $code ("        return %s$unknown_label;\n", ($enum_class ? "${enum_name}::" : ""));
        print $code ("    }\n");
        print $code ("}\n");

        # Print end of header here, in case header and code point to the same file
        print $header ("#endif                       /* TRIE_HASH_${function_name} */\n");
    }
}

# Check if the word can be reached by exactly one word in (alphabet OR 0x20).
sub ambiguous {
    my $word = shift;

    foreach my $char (split //, $word) {
        # Setting the lowercase flag in the character produces a different
        # character, the character would thus not be matched.
        return 1 if ((ord($char) | 0x20) != ord(lc($char)));

        # A word is also ambiguous if any character in lowercase can be reached
        # by ORing 0x20 from another character in the charset that is not a
        # lowercase character of the current character.
        # Assume that we have UTF-8 and the most significant bit can be set
        for my $i (0..255) {
            return 1 if (($i | 0x20) == ord(lc($char)) && lc(chr($i)) ne lc($char));
        }
    }

    return 0;
}

sub build_trie {
    my $codegen = shift;
    my $trie = Trie->new;

    my $counter = $counter_start;
    my %lengths;

    open(my $input, '<', $ARGV[0]) or die "Cannot open ".$ARGV[0].": $!";
    while (my $line = <$input>) {
        my ($label, $word, $value) = $line =~/\s*(?:([^~\s]+)\s*~)?(?:\s*([^~=\s]+)\s*)?(?:=\s*([^\s]+)\s+)?\s*/;

        if (defined $word) {
            $counter = $value if defined($value);
            $label //= $codegen->word_to_label($word);

            $trie->insert($word, $label, $counter);
            $lengths{length($word)} = 1;
            $counter++;
        } elsif (defined $value) {
            $unknown = $value;
            $unknown_label = $label if defined($label);
            $counter = $value + 1;
        } else {
            die "Invalid line: $line";
        }
    }

    return ($trie, $counter, %lengths);
}

# Generates an ASCII art tree
package TreeCodeGen {

    sub new {
        my $class = shift;
        my $self = {};
        bless $self, $class;

        return $self;
    }

    sub word_to_label {
        my ($self, $word) = @_;
        return $word;
    }

    sub main {
        my ($self, $trie, $counter, %lengths) = @_;
        printf $code ("┌────────────────────────────────────────────────────┐\n");
        printf $code ("│                   Initial trie                     │\n");
        printf $code ("└────────────────────────────────────────────────────┘\n");
        $self->print($trie);
        printf $code ("┌────────────────────────────────────────────────────┐\n");
        printf $code ("│                   Rebuilt trie                     │\n");
        printf $code ("└────────────────────────────────────────────────────┘\n");
        $self->print($trie->rebuild_tree());

        foreach my $local_length (sort { $a <=> $b } (keys %lengths)) {
            printf $code ("┌────────────────────────────────────────────────────┐\n");
            printf $code ("│              Trie for words of length %-4d         │\n", $local_length);
            printf $code ("└────────────────────────────────────────────────────┘\n");
            $self->print($trie->filter_depth($local_length)->rebuild_tree());
        }
    }

    sub open_output {
        my $self = shift;
        if ($code_name ne "-") {
            open($code, '>', $code_name) or die "Cannot open ".$ARGV[0].": $!" ;
        } else {
            $code = *STDOUT;
        }
    }

    # Print a trie
    sub print {
        my ($self, $trie, $depth) = @_;
        $depth //= 0;

        print $code (" → ") if defined($trie->{label});
        print $code ($trie->{label} // "", "\n");
        foreach my $key (sort keys %{$trie->{children}}) {
            print $code ("│   " x ($depth), "├── $key");
            $self->print($trie->{children}{$key}, $depth + 1);
        }
    }
}

my %codegens = (
    C => "CCodeGen",
    tree => "TreeCodeGen",
);


defined($codegens{$language}) or die "Unknown language $language. Valid choices: ", join(", ", keys %codegens);
my $codegen = $codegens{$language}->new();
my ($trie, $counter, %lengths) = build_trie($codegen);

$codegen->open_output();
$codegen->main($trie, $counter, %lengths);


=head1 LICENSE

triehash is available under the MIT/Expat license, see the source code
for more information.

=head1 AUTHOR

Julian Andres Klode <jak@jak-linux.org>

=cut