Commit | Line | Data |
3ef515df |
1 | #!./perl |
67d7b5ef |
2 | BEGIN { |
037b88d6 |
3 | # fiddle with @INC iff I am a part of perl dist |
4 | if ($^X =~ m/\bminiperl$/o){ |
5 | warn "Fixing \@INC for perl core.\n"; |
6 | unshift @INC, qw(../../lib ../../../lib ../../../../lib); |
7 | $ENV{PATH} .= ';../..;../../..;../../../..' if $^O eq 'MSWin32'; |
8 | } |
67d7b5ef |
9 | } |
10 | use strict; |
11 | use Getopt::Std; |
12 | my @orig_ARGV = @ARGV; |
3ef515df |
13 | our $VERSION = do { my @r = (q$Revision: 1.10 $ =~ /\d+/g); sprintf "%d."."%02d" x $#r, @r }; |
67d7b5ef |
14 | |
15 | |
16 | # These may get re-ordered. |
17 | # RAW is a do_now as inserted by &enter |
18 | # AGG is an aggreagated do_now, as built up by &process |
19 | use constant { |
20 | RAW_NEXT => 0, |
21 | RAW_IN_LEN => 1, |
22 | RAW_OUT_BYTES => 2, |
23 | RAW_FALLBACK => 3, |
24 | |
25 | AGG_MIN_IN => 0, |
26 | AGG_MAX_IN => 1, |
27 | AGG_OUT_BYTES => 2, |
28 | AGG_NEXT => 3, |
29 | AGG_IN_LEN => 4, |
30 | AGG_OUT_LEN => 5, |
31 | AGG_FALLBACK => 6, |
32 | }; |
33 | # (See the algorithm in encengine.c - we're building structures for it) |
34 | |
35 | # There are two sorts of structures. |
36 | # "do_now" (an array, two variants of what needs storing) is whatever we need |
37 | # to do now we've read an input byte. |
38 | # It's housed in a "do_next" (which is how we got to it), and in turn points |
39 | # to a "do_next" which contains all the "do_now"s for the next input byte. |
40 | |
41 | # There will be a "do_next" which is the start state. |
42 | # For a single byte encoding it's the only "do_next" - each "do_now" points |
43 | # back to it, and each "do_now" will cause bytes. There is no state. |
44 | |
45 | # For a multi-byte encoding where all characters in the input are the same |
46 | # length, then there will be a tree of "do_now"->"do_next"->"do_now" |
47 | # branching out from the start state, one step for each input byte. |
48 | # The leaf "do_now"s will all be at the same distance from the start state, |
49 | # only the leaf "do_now"s cause output bytes, and they in turn point back to |
50 | # the start state. |
51 | |
52 | # For an encoding where there are varaible length input byte sequences, you |
53 | # will encounter a leaf "do_now" sooner for the shorter input sequences, but |
54 | # as before the leaves will point back to the start state. |
55 | |
56 | # The system will cope with escape encodings (imagine them as a mostly |
57 | # self-contained tree for each escape state, and cross links between trees |
58 | # at the state-switching characters) but so far no input format defines these. |
59 | |
60 | # The system will also cope with having output "leaves" in the middle of |
61 | # the bifurcating branches, not just at the extremities, but again no |
62 | # input format does this yet. |
63 | |
64 | # There are two variants of the "do_now" structure. The first, smaller variant |
65 | # is generated by &enter as the input file is read. There is one structure |
66 | # for each input byte. Say we are mapping a single byte encoding to a |
67 | # single byte encoding, with "ABCD" going "abcd". There will be |
68 | # 4 "do_now"s, {"A" => [...,"a",...], "B" => [...,"b",...], "C"=>..., "D"=>...} |
69 | |
70 | # &process then walks the tree, building aggregate "do_now" structres for |
71 | # adjacent bytes where possible. The aggregate is for a contiguous range of |
72 | # bytes which each produce the same length of output, each move to the |
73 | # same next state, and each have the same fallback flag. |
74 | # So our 4 RAW "do_now"s above become replaced by a single structure |
75 | # containing: |
76 | # ["A", "D", "abcd", 1, ...] |
77 | # ie, for an input byte $_ in "A".."D", output 1 byte, found as |
78 | # substr ("abcd", (ord $_ - ord "A") * 1, 1) |
79 | # which maps very nicely into pointer arithmetic in C for encengine.c |
80 | |
81 | sub encode_U |
82 | { |
83 | # UTF-8 encode long hand - only covers part of perl's range |
84 | ## my $uv = shift; |
85 | # chr() works in native space so convert value from table |
86 | # into that space before using chr(). |
87 | my $ch = chr(utf8::unicode_to_native($_[0])); |
88 | # Now get core perl to encode that the way it likes. |
89 | utf8::encode($ch); |
90 | return $ch; |
91 | } |
92 | |
93 | sub encode_S |
94 | { |
95 | # encode single byte |
96 | ## my ($ch,$page) = @_; return chr($ch); |
97 | return chr $_[0]; |
98 | } |
99 | |
100 | sub encode_D |
101 | { |
102 | # encode double byte MS byte first |
103 | ## my ($ch,$page) = @_; return chr($page).chr($ch); |
104 | return chr ($_[1]) . chr $_[0]; |
105 | } |
106 | |
107 | sub encode_M |
108 | { |
109 | # encode Multi-byte - single for 0..255 otherwise double |
110 | ## my ($ch,$page) = @_; |
111 | ## return &encode_D if $page; |
112 | ## return &encode_S; |
113 | return chr ($_[1]) . chr $_[0] if $_[1]; |
114 | return chr $_[0]; |
115 | } |
116 | |
117 | my %encode_types = (U => \&encode_U, |
118 | S => \&encode_S, |
119 | D => \&encode_D, |
120 | M => \&encode_M, |
121 | ); |
122 | |
123 | # Win32 does not expand globs on command line |
124 | eval "\@ARGV = map(glob(\$_),\@ARGV)" if ($^O eq 'MSWin32'); |
125 | |
126 | my %opt; |
127 | # I think these are: |
128 | # -Q to disable the duplicate codepoint test |
129 | # -S make mapping errors fatal |
130 | # -q to remove comments written to output files |
131 | # -O to enable the (brute force) substring optimiser |
132 | # -o <output> to specify the output file name (else it's the first arg) |
133 | # -f <inlist> to give a file with a list of input files (else use the args) |
134 | # -n <name> to name the encoding (else use the basename of the input file. |
135 | getopts('M:SQqOo:f:n:',\%opt); |
136 | |
137 | $opt{M} and make_makefile_pl($opt{M}, @ARGV); |
138 | |
139 | # This really should go first, else the die here causes empty (non-erroneous) |
140 | # output files to be written. |
141 | my @encfiles; |
142 | if (exists $opt{'f'}) { |
143 | # -F is followed by name of file containing list of filenames |
144 | my $flist = $opt{'f'}; |
145 | open(FLIST,$flist) || die "Cannot open $flist:$!"; |
146 | chomp(@encfiles = <FLIST>); |
147 | close(FLIST); |
148 | } else { |
149 | @encfiles = @ARGV; |
150 | } |
151 | |
152 | my $cname = (exists $opt{'o'}) ? $opt{'o'} : shift(@ARGV); |
153 | chmod(0666,$cname) if -f $cname && !-w $cname; |
154 | open(C,">$cname") || die "Cannot open $cname:$!"; |
155 | |
156 | my $dname = $cname; |
157 | my $hname = $cname; |
158 | |
159 | my ($doC,$doEnc,$doUcm,$doPet); |
160 | |
161 | if ($cname =~ /\.(c|xs)$/) |
162 | { |
163 | $doC = 1; |
e7cbefb8 |
164 | $dname =~ s/(\.[^\.]*)?$/.exh/; |
67d7b5ef |
165 | chmod(0666,$dname) if -f $cname && !-w $dname; |
166 | open(D,">$dname") || die "Cannot open $dname:$!"; |
167 | $hname =~ s/(\.[^\.]*)?$/.h/; |
168 | chmod(0666,$hname) if -f $cname && !-w $hname; |
169 | open(H,">$hname") || die "Cannot open $hname:$!"; |
170 | |
171 | foreach my $fh (\*C,\*D,\*H) |
172 | { |
173 | print $fh <<"END" unless $opt{'q'}; |
174 | /* |
175 | !!!!!!! DO NOT EDIT THIS FILE !!!!!!! |
176 | This file was autogenerated by: |
177 | $^X $0 @orig_ARGV |
178 | */ |
179 | END |
180 | } |
181 | |
182 | if ($cname =~ /(\w+)\.xs$/) |
183 | { |
184 | print C "#include <EXTERN.h>\n"; |
185 | print C "#include <perl.h>\n"; |
186 | print C "#include <XSUB.h>\n"; |
187 | print C "#define U8 U8\n"; |
188 | } |
189 | print C "#include \"encode.h\"\n"; |
190 | |
191 | } |
192 | elsif ($cname =~ /\.enc$/) |
193 | { |
194 | $doEnc = 1; |
195 | } |
196 | elsif ($cname =~ /\.ucm$/) |
197 | { |
198 | $doUcm = 1; |
199 | } |
200 | elsif ($cname =~ /\.pet$/) |
201 | { |
202 | $doPet = 1; |
203 | } |
204 | |
205 | my %encoding; |
206 | my %strings; |
207 | my $saved = 0; |
208 | my $subsave = 0; |
209 | my $strings = 0; |
210 | |
211 | sub cmp_name |
212 | { |
213 | if ($a =~ /^.*-(\d+)/) |
214 | { |
215 | my $an = $1; |
216 | if ($b =~ /^.*-(\d+)/) |
217 | { |
218 | my $r = $an <=> $1; |
219 | return $r if $r; |
220 | } |
221 | } |
222 | return $a cmp $b; |
223 | } |
224 | |
225 | |
226 | foreach my $enc (sort cmp_name @encfiles) |
227 | { |
228 | my ($name,$sfx) = $enc =~ /^.*?([\w-]+)\.(enc|ucm)$/; |
229 | $name = $opt{'n'} if exists $opt{'n'}; |
230 | if (open(E,$enc)) |
231 | { |
232 | if ($sfx eq 'enc') |
233 | { |
234 | compile_enc(\*E,lc($name)); |
235 | } |
236 | else |
237 | { |
238 | compile_ucm(\*E,lc($name)); |
239 | } |
240 | } |
241 | else |
242 | { |
243 | warn "Cannot open $enc for $name:$!"; |
244 | } |
245 | } |
246 | |
247 | if ($doC) |
248 | { |
249 | print STDERR "Writing compiled form\n"; |
250 | foreach my $name (sort cmp_name keys %encoding) |
251 | { |
252 | my ($e2u,$u2e,$erep,$min_el,$max_el) = @{$encoding{$name}}; |
253 | output(\*C,$name.'_utf8',$e2u); |
254 | output(\*C,'utf8_'.$name,$u2e); |
255 | push(@{$encoding{$name}},outstring(\*C,$e2u->{Cname}.'_def',$erep)); |
256 | } |
257 | foreach my $enc (sort cmp_name keys %encoding) |
258 | { |
259 | my ($e2u,$u2e,$rep,$min_el,$max_el,$rsym) = @{$encoding{$enc}}; |
260 | my @info = ($e2u->{Cname},$u2e->{Cname},$rsym,length($rep),$min_el,$max_el); |
261 | my $sym = "${enc}_encoding"; |
262 | $sym =~ s/\W+/_/g; |
263 | print C "encode_t $sym = \n"; |
264 | print C " {",join(',',@info,"{\"$enc\",(const char *)0}"),"};\n\n"; |
265 | } |
266 | |
267 | foreach my $enc (sort cmp_name keys %encoding) |
268 | { |
269 | my $sym = "${enc}_encoding"; |
270 | $sym =~ s/\W+/_/g; |
271 | print H "extern encode_t $sym;\n"; |
272 | print D " Encode_XSEncoding(aTHX_ &$sym);\n"; |
273 | } |
274 | |
275 | if ($cname =~ /(\w+)\.xs$/) |
276 | { |
277 | my $mod = $1; |
278 | print C <<'END'; |
279 | |
280 | static void |
281 | Encode_XSEncoding(pTHX_ encode_t *enc) |
282 | { |
283 | dSP; |
284 | HV *stash = gv_stashpv("Encode::XS", TRUE); |
285 | SV *sv = sv_bless(newRV_noinc(newSViv(PTR2IV(enc))),stash); |
286 | int i = 0; |
287 | PUSHMARK(sp); |
288 | XPUSHs(sv); |
289 | while (enc->name[i]) |
290 | { |
291 | const char *name = enc->name[i++]; |
292 | XPUSHs(sv_2mortal(newSVpvn(name,strlen(name)))); |
293 | } |
294 | PUTBACK; |
295 | call_pv("Encode::define_encoding",G_DISCARD); |
296 | SvREFCNT_dec(sv); |
297 | } |
298 | |
299 | END |
300 | |
301 | print C "\nMODULE = Encode::$mod\tPACKAGE = Encode::$mod\n\n"; |
302 | print C "BOOT:\n{\n"; |
303 | print C "#include \"$dname\"\n"; |
304 | print C "}\n"; |
305 | } |
306 | # Close in void context is bad, m'kay |
307 | close(D) or warn "Error closing '$dname': $!"; |
308 | close(H) or warn "Error closing '$hname': $!"; |
309 | |
310 | my $perc_saved = $strings/($strings + $saved) * 100; |
311 | my $perc_subsaved = $strings/($strings + $subsave) * 100; |
312 | printf STDERR "%d bytes in string tables\n",$strings; |
313 | printf STDERR "%d bytes (%.3g%%) saved spotting duplicates\n", |
314 | $saved, $perc_saved if $saved; |
315 | printf STDERR "%d bytes (%.3g%%) saved using substrings\n", |
316 | $subsave, $perc_subsaved if $subsave; |
317 | } |
318 | elsif ($doEnc) |
319 | { |
320 | foreach my $name (sort cmp_name keys %encoding) |
321 | { |
322 | my ($e2u,$u2e,$erep,$min_el,$max_el) = @{$encoding{$name}}; |
323 | output_enc(\*C,$name,$e2u); |
324 | } |
325 | } |
326 | elsif ($doUcm) |
327 | { |
328 | foreach my $name (sort cmp_name keys %encoding) |
329 | { |
330 | my ($e2u,$u2e,$erep,$min_el,$max_el) = @{$encoding{$name}}; |
331 | output_ucm(\*C,$name,$u2e,$erep,$min_el,$max_el); |
332 | } |
333 | } |
334 | |
335 | # writing half meg files and then not checking to see if you just filled the |
336 | # disk is bad, m'kay |
337 | close(C) or die "Error closing '$cname': $!"; |
338 | |
339 | # End of the main program. |
340 | |
341 | sub compile_ucm |
342 | { |
343 | my ($fh,$name) = @_; |
344 | my $e2u = {}; |
345 | my $u2e = {}; |
346 | my $cs; |
347 | my %attr; |
348 | while (<$fh>) |
349 | { |
350 | s/#.*$//; |
351 | last if /^\s*CHARMAP\s*$/i; |
352 | if (/^\s*<(\w+)>\s+"?([^"]*)"?\s*$/i) # " # Grrr |
353 | { |
354 | $attr{$1} = $2; |
355 | } |
356 | } |
357 | if (!defined($cs = $attr{'code_set_name'})) |
358 | { |
359 | warn "No <code_set_name> in $name\n"; |
360 | } |
361 | else |
362 | { |
363 | $name = $cs unless exists $opt{'n'}; |
364 | } |
365 | my $erep; |
366 | my $urep; |
367 | my $max_el; |
368 | my $min_el; |
369 | if (exists $attr{'subchar'}) |
370 | { |
371 | my @byte; |
372 | $attr{'subchar'} =~ /^\s*/cg; |
373 | push(@byte,$1) while $attr{'subchar'} =~ /\G\\x([0-9a-f]+)/icg; |
374 | $erep = join('',map(chr(hex($_)),@byte)); |
375 | } |
376 | print "Reading $name ($cs)\n"; |
377 | my $nfb = 0; |
378 | my $hfb = 0; |
379 | while (<$fh>) |
380 | { |
381 | s/#.*$//; |
382 | last if /^\s*END\s+CHARMAP\s*$/i; |
383 | next if /^\s*$/; |
384 | my ($u,@byte); |
385 | my $fb = ''; |
386 | $u = $1 if (/^<U([0-9a-f]+)>\s+/igc); |
387 | push(@byte,$1) while /\G\\x([0-9a-f]+)/igc; |
388 | $fb = $1 if /\G\s*(\|[0-3])/gc; |
389 | # warn "$_: $u @byte | $fb\n"; |
390 | die "Bad line:$_" unless /\G\s*(#.*)?$/gc; |
391 | if (defined($u)) |
392 | { |
393 | my $uch = encode_U(hex($u)); |
394 | my $ech = join('',map(chr(hex($_)),@byte)); |
395 | my $el = length($ech); |
396 | $max_el = $el if (!defined($max_el) || $el > $max_el); |
397 | $min_el = $el if (!defined($min_el) || $el < $min_el); |
398 | if (length($fb)) |
399 | { |
400 | $fb = substr($fb,1); |
401 | $hfb++; |
402 | } |
403 | else |
404 | { |
405 | $nfb++; |
406 | $fb = '0'; |
407 | } |
408 | # $fb is fallback flag |
409 | # 0 - round trip safe |
410 | # 1 - fallback for unicode -> enc |
411 | # 2 - skip sub-char mapping |
412 | # 3 - fallback enc -> unicode |
413 | enter($u2e,$uch,$ech,$u2e,$fb+0) if ($fb =~ /[01]/); |
414 | enter($e2u,$ech,$uch,$e2u,$fb+0) if ($fb =~ /[03]/); |
415 | } |
416 | else |
417 | { |
418 | warn $_; |
419 | } |
420 | } |
421 | if ($nfb && $hfb) |
422 | { |
423 | die "$nfb entries without fallback, $hfb entries with\n"; |
424 | } |
425 | $encoding{$name} = [$e2u,$u2e,$erep,$min_el,$max_el]; |
426 | } |
427 | |
428 | |
429 | |
430 | sub compile_enc |
431 | { |
432 | my ($fh,$name) = @_; |
433 | my $e2u = {}; |
434 | my $u2e = {}; |
435 | |
436 | my $type; |
437 | while ($type = <$fh>) |
438 | { |
439 | last if $type !~ /^\s*#/; |
440 | } |
441 | chomp($type); |
442 | return if $type eq 'E'; |
443 | # Do the hash lookup once, rather than once per function call. 4% speedup. |
444 | my $type_func = $encode_types{$type}; |
445 | my ($def,$sym,$pages) = split(/\s+/,scalar(<$fh>)); |
446 | warn "$type encoded $name\n"; |
447 | my $rep = ''; |
448 | # Save a defined test by setting these to defined values. |
449 | my $min_el = ~0; # A very big integer |
450 | my $max_el = 0; # Anything must be longer than 0 |
451 | { |
452 | my $v = hex($def); |
453 | $rep = &$type_func($v & 0xFF, ($v >> 8) & 0xffe); |
454 | } |
455 | my $errors; |
456 | my $seen; |
457 | # use -Q to silence the seen test. Makefile.PL uses this by default. |
458 | $seen = {} unless $opt{Q}; |
459 | do |
460 | { |
461 | my $line = <$fh>; |
462 | chomp($line); |
463 | my $page = hex($line); |
464 | my $ch = 0; |
465 | my $i = 16; |
466 | do |
467 | { |
468 | # So why is it 1% faster to leave the my here? |
469 | my $line = <$fh>; |
470 | $line =~ s/\r\n$/\n/; |
471 | die "$.:${line}Line should be exactly 65 characters long including |
472 | newline (".length($line).")" unless length ($line) == 65; |
473 | # Split line into groups of 4 hex digits, convert groups to ints |
474 | # This takes 65.35 |
475 | # map {hex $_} $line =~ /(....)/g |
476 | # This takes 63.75 (2.5% less time) |
477 | # unpack "n*", pack "H*", $line |
478 | # There's an implicit loop in map. Loops are bad, m'kay. Ops are bad, m'kay |
479 | # Doing it as while ($line =~ /(....)/g) took 74.63 |
480 | foreach my $val (unpack "n*", pack "H*", $line) |
481 | { |
482 | next if $val == 0xFFFD; |
483 | my $ech = &$type_func($ch,$page); |
484 | if ($val || (!$ch && !$page)) |
485 | { |
486 | my $el = length($ech); |
487 | $max_el = $el if $el > $max_el; |
488 | $min_el = $el if $el < $min_el; |
489 | my $uch = encode_U($val); |
490 | if ($seen) { |
491 | # We're doing the test. |
492 | # We don't need to read this quickly, so storing it as a scalar, |
493 | # rather than 3 (anon array, plus the 2 scalars it holds) saves |
494 | # RAM and may make us faster on low RAM systems. [see __END__] |
495 | if (exists $seen->{$uch}) |
496 | { |
497 | warn sprintf("U%04X is %02X%02X and %04X\n", |
498 | $val,$page,$ch,$seen->{$uch}); |
499 | $errors++; |
500 | } |
501 | else |
502 | { |
503 | $seen->{$uch} = $page << 8 | $ch; |
504 | } |
505 | } |
506 | # Passing 2 extra args each time is 3.6% slower! |
507 | # Even with having to add $fallback ||= 0 later |
508 | enter_fb0($e2u,$ech,$uch); |
509 | enter_fb0($u2e,$uch,$ech); |
510 | } |
511 | else |
512 | { |
513 | # No character at this position |
514 | # enter($e2u,$ech,undef,$e2u); |
515 | } |
516 | $ch++; |
517 | } |
518 | } while --$i; |
519 | } while --$pages; |
520 | die "\$min_el=$min_el, \$max_el=$max_el - seems we read no lines" |
521 | if $min_el > $max_el; |
522 | die "$errors mapping conflicts\n" if ($errors && $opt{'S'}); |
523 | $encoding{$name} = [$e2u,$u2e,$rep,$min_el,$max_el]; |
524 | } |
525 | |
526 | # my ($a,$s,$d,$t,$fb) = @_; |
527 | sub enter { |
528 | my ($current,$inbytes,$outbytes,$next,$fallback) = @_; |
529 | # state we shift to after this (multibyte) input character defaults to same |
530 | # as current state. |
531 | $next ||= $current; |
532 | # Making sure it is defined seems to be faster than {no warnings;} in |
533 | # &process, or passing it in as 0 explicity. |
534 | # XXX $fallback ||= 0; |
535 | |
536 | # Start at the beginning and work forwards through the string to zero. |
537 | # effectively we are removing 1 character from the front each time |
538 | # but we don't actually edit the string. [this alone seems to be 14% speedup] |
539 | # Hence -$pos is the length of the remaining string. |
540 | my $pos = -length $inbytes; |
541 | while (1) { |
542 | my $byte = substr $inbytes, $pos, 1; |
543 | # RAW_NEXT => 0, |
544 | # RAW_IN_LEN => 1, |
545 | # RAW_OUT_BYTES => 2, |
546 | # RAW_FALLBACK => 3, |
547 | # to unicode an array would seem to be better, because the pages are dense. |
548 | # from unicode can be very sparse, favouring a hash. |
549 | # hash using the bytes (all length 1) as keys rather than ord value, |
550 | # as it's easier to sort these in &process. |
551 | |
552 | # It's faster to always add $fallback even if it's undef, rather than |
553 | # choosing between 3 and 4 element array. (hence why we set it defined |
554 | # above) |
555 | my $do_now = $current->{Raw}{$byte} ||= [{},-$pos,'',$fallback]; |
556 | # When $pos was -1 we were at the last input character. |
557 | unless (++$pos) { |
558 | $do_now->[RAW_OUT_BYTES] = $outbytes; |
559 | $do_now->[RAW_NEXT] = $next; |
560 | return; |
561 | } |
562 | # Tail recursion. The intermdiate state may not have a name yet. |
563 | $current = $do_now->[RAW_NEXT]; |
564 | } |
565 | } |
566 | |
567 | # This is purely for optimistation. It's just &enter hard coded for $fallback |
568 | # of 0, using only a 3 entry array ref to save memory for every entry. |
569 | sub enter_fb0 { |
570 | my ($current,$inbytes,$outbytes,$next) = @_; |
571 | $next ||= $current; |
572 | |
573 | my $pos = -length $inbytes; |
574 | while (1) { |
575 | my $byte = substr $inbytes, $pos, 1; |
576 | my $do_now = $current->{Raw}{$byte} ||= [{},-$pos,'']; |
577 | unless (++$pos) { |
578 | $do_now->[RAW_OUT_BYTES] = $outbytes; |
579 | $do_now->[RAW_NEXT] = $next; |
580 | return; |
581 | } |
582 | $current = $do_now->[RAW_NEXT]; |
583 | } |
584 | } |
585 | |
586 | |
587 | sub outstring |
588 | { |
589 | my ($fh,$name,$s) = @_; |
590 | my $sym = $strings{$s}; |
591 | if ($sym) |
592 | { |
593 | $saved += length($s); |
594 | } |
595 | else |
596 | { |
597 | if ($opt{'O'}) { |
598 | foreach my $o (keys %strings) |
599 | { |
600 | next unless (my $i = index($o,$s)) >= 0; |
601 | $sym = $strings{$o}; |
602 | # gcc things that 0x0e+0x10 (anything with e+) starts to look like |
603 | # a hexadecimal floating point constant. Silly gcc. Only p |
604 | # introduces a floating point constant. Put the space in to stop it |
605 | # getting confused. |
606 | $sym .= sprintf(" +0x%02x",$i) if ($i); |
607 | $subsave += length($s); |
608 | return $strings{$s} = $sym; |
609 | } |
610 | } |
611 | $strings{$s} = $sym = $name; |
612 | $strings += length($s); |
613 | my $definition = sprintf "static const U8 %s[%d] = { ",$name,length($s); |
614 | # Maybe we should assert that these are all <256. |
615 | $definition .= join(',',unpack "C*",$s); |
616 | # We have a single long line. Split it at convenient commas. |
617 | $definition =~ s/(.{74,77},)/$1\n/g; |
618 | print $fh "$definition };\n\n"; |
619 | } |
620 | return $sym; |
621 | } |
622 | |
623 | sub process |
624 | { |
625 | my ($name,$a) = @_; |
626 | $name =~ s/\W+/_/g; |
627 | $a->{Cname} = $name; |
628 | my $raw = $a->{Raw}; |
629 | my ($l, $agg_max_in, $agg_next, $agg_in_len, $agg_out_len, $agg_fallback); |
630 | my @ent; |
631 | $agg_max_in = 0; |
632 | foreach my $key (sort keys %$raw) { |
633 | # RAW_NEXT => 0, |
634 | # RAW_IN_LEN => 1, |
635 | # RAW_OUT_BYTES => 2, |
636 | # RAW_FALLBACK => 3, |
637 | my ($next, $in_len, $out_bytes, $fallback) = @{$raw->{$key}}; |
638 | # Now we are converting from raw to aggregate, switch from 1 byte strings |
639 | # to numbers |
640 | my $b = ord $key; |
641 | $fallback ||= 0; |
642 | if ($l && |
643 | # If this == fails, we're going to reset $agg_max_in below anyway. |
644 | $b == ++$agg_max_in && |
645 | # References in numeric context give the pointer as an int. |
646 | $agg_next == $next && |
647 | $agg_in_len == $in_len && |
648 | $agg_out_len == length $out_bytes && |
649 | $agg_fallback == $fallback |
650 | # && length($l->[AGG_OUT_BYTES]) < 16 |
651 | ) { |
652 | # my $i = ord($b)-ord($l->[AGG_MIN_IN]); |
653 | # we can aggregate this byte onto the end. |
654 | $l->[AGG_MAX_IN] = $b; |
655 | $l->[AGG_OUT_BYTES] .= $out_bytes; |
656 | } else { |
657 | # AGG_MIN_IN => 0, |
658 | # AGG_MAX_IN => 1, |
659 | # AGG_OUT_BYTES => 2, |
660 | # AGG_NEXT => 3, |
661 | # AGG_IN_LEN => 4, |
662 | # AGG_OUT_LEN => 5, |
663 | # AGG_FALLBACK => 6, |
664 | # Reset the last thing we saw, plus set 5 lexicals to save some derefs. |
665 | # (only gains .6% on euc-jp -- is it worth it?) |
666 | push @ent, $l = [$b, $agg_max_in = $b, $out_bytes, $agg_next = $next, |
667 | $agg_in_len = $in_len, $agg_out_len = length $out_bytes, |
668 | $agg_fallback = $fallback]; |
669 | } |
670 | if (exists $next->{Cname}) { |
671 | $next->{'Forward'} = 1 if $next != $a; |
672 | } else { |
673 | process(sprintf("%s_%02x",$name,$b),$next); |
674 | } |
675 | } |
676 | # encengine.c rules say that last entry must be for 255 |
677 | if ($agg_max_in < 255) { |
678 | push @ent, [1+$agg_max_in, 255,undef,$a,0,0]; |
679 | } |
680 | $a->{'Entries'} = \@ent; |
681 | } |
682 | |
683 | sub outtable |
684 | { |
685 | my ($fh,$a) = @_; |
686 | my $name = $a->{'Cname'}; |
687 | # String tables |
688 | foreach my $b (@{$a->{'Entries'}}) |
689 | { |
690 | next unless $b->[AGG_OUT_LEN]; |
691 | my $s = $b->[AGG_MIN_IN]; |
692 | my $e = $b->[AGG_MAX_IN]; |
693 | outstring($fh,sprintf("%s__%02x_%02x",$name,$s,$e),$b->[AGG_OUT_BYTES]); |
694 | } |
695 | if ($a->{'Forward'}) |
696 | { |
697 | print $fh "\nstatic encpage_t $name\[",scalar(@{$a->{'Entries'}}),"];\n"; |
698 | } |
699 | $a->{'Done'} = 1; |
700 | foreach my $b (@{$a->{'Entries'}}) |
701 | { |
702 | my ($s,$e,$out,$t,$end,$l) = @$b; |
703 | outtable($fh,$t) unless $t->{'Done'}; |
704 | } |
705 | print $fh "\nstatic encpage_t $name\[",scalar(@{$a->{'Entries'}}),"] = {\n"; |
706 | foreach my $b (@{$a->{'Entries'}}) |
707 | { |
708 | my ($sc,$ec,$out,$t,$end,$l,$fb) = @$b; |
709 | $end |= 0x80 if $fb; |
710 | print $fh "{"; |
711 | if ($l) |
712 | { |
713 | printf $fh outstring($fh,'',$out); |
714 | } |
715 | else |
716 | { |
717 | print $fh "0"; |
718 | } |
719 | print $fh ",",$t->{Cname}; |
720 | printf $fh ",0x%02x,0x%02x,$l,$end},\n",$sc,$ec; |
721 | } |
722 | print $fh "};\n"; |
723 | } |
724 | |
725 | sub output |
726 | { |
727 | my ($fh,$name,$a) = @_; |
728 | process($name,$a); |
729 | # Sub-tables |
730 | outtable($fh,$a); |
731 | } |
732 | |
733 | sub output_enc |
734 | { |
735 | my ($fh,$name,$a) = @_; |
736 | die "Changed - fix me for new structure"; |
737 | foreach my $b (sort keys %$a) |
738 | { |
739 | my ($s,$e,$out,$t,$end,$l,$fb) = @{$a->{$b}}; |
740 | } |
741 | } |
742 | |
743 | sub decode_U |
744 | { |
745 | my $s = shift; |
746 | } |
747 | |
748 | my @uname; |
749 | sub char_names |
750 | { |
751 | my $s = do "unicore/Name.pl"; |
752 | die "char_names: unicore/Name.pl: $!\n" unless defined $s; |
753 | pos($s) = 0; |
754 | while ($s =~ /\G([0-9a-f]+)\t([0-9a-f]*)\t(.*?)\s*\n/igc) |
755 | { |
756 | my $name = $3; |
757 | my $s = hex($1); |
758 | last if $s >= 0x10000; |
759 | my $e = length($2) ? hex($2) : $s; |
760 | for (my $i = $s; $i <= $e; $i++) |
761 | { |
762 | $uname[$i] = $name; |
763 | # print sprintf("U%04X $name\n",$i); |
764 | } |
765 | } |
766 | } |
767 | |
768 | sub output_ucm_page |
769 | { |
770 | my ($cmap,$a,$t,$pre) = @_; |
771 | # warn sprintf("Page %x\n",$pre); |
772 | my $raw = $t->{Raw}; |
773 | foreach my $key (sort keys %$raw) { |
774 | # RAW_NEXT => 0, |
775 | # RAW_IN_LEN => 1, |
776 | # RAW_OUT_BYTES => 2, |
777 | # RAW_FALLBACK => 3, |
778 | my ($next, $in_len, $out_bytes, $fallback) = @{$raw->{$key}}; |
779 | my $u = ord $key; |
780 | $fallback ||= 0; |
781 | |
782 | if ($next != $a && $next != $t) { |
783 | output_ucm_page($cmap,$a,$next,(($pre|($u &0x3F)) << 6)&0xFFFF); |
784 | } elsif (length $out_bytes) { |
785 | if ($pre) { |
786 | $u = $pre|($u &0x3f); |
787 | } |
788 | my $s = sprintf "<U%04X> ",$u; |
789 | #foreach my $c (split(//,$out_bytes)) { |
790 | # $s .= sprintf "\\x%02X",ord($c); |
791 | #} |
792 | # 9.5% faster changing that loop to this: |
793 | $s .= sprintf +("\\x%02X" x length $out_bytes), unpack "C*", $out_bytes; |
794 | $s .= sprintf " |%d # %s\n",($fallback ? 1 : 0),$uname[$u]; |
795 | push(@$cmap,$s); |
796 | } else { |
797 | warn join(',',$u, @{$raw->{$key}},$a,$t); |
798 | } |
799 | } |
800 | } |
801 | |
802 | sub output_ucm |
803 | { |
804 | my ($fh,$name,$h,$rep,$min_el,$max_el) = @_; |
805 | print $fh "# $0 @orig_ARGV\n" unless $opt{'q'}; |
806 | print $fh "<code_set_name> \"$name\"\n"; |
807 | char_names(); |
808 | if (defined $min_el) |
809 | { |
810 | print $fh "<mb_cur_min> $min_el\n"; |
811 | } |
812 | if (defined $max_el) |
813 | { |
814 | print $fh "<mb_cur_max> $max_el\n"; |
815 | } |
816 | if (defined $rep) |
817 | { |
818 | print $fh "<subchar> "; |
819 | foreach my $c (split(//,$rep)) |
820 | { |
821 | printf $fh "\\x%02X",ord($c); |
822 | } |
823 | print $fh "\n"; |
824 | } |
825 | my @cmap; |
826 | output_ucm_page(\@cmap,$h,$h,0); |
827 | print $fh "#\nCHARMAP\n"; |
828 | foreach my $line (sort { substr($a,8) cmp substr($b,8) } @cmap) |
829 | { |
830 | print $fh $line; |
831 | } |
832 | print $fh "END CHARMAP\n"; |
833 | } |
834 | |
3ef515df |
835 | use vars qw( |
836 | $_Enc2xs |
837 | $_Version |
838 | $_Inc |
839 | $_Name |
840 | $_TableFiles |
841 | $_Now |
842 | ); |
843 | |
67d7b5ef |
844 | sub make_makefile_pl |
845 | { |
846 | eval { require Encode; }; |
847 | $@ and die "You need to install Encode to use enc2xs -M\nerror: $@\n"; |
3ef515df |
848 | # our used for variable expanstion |
849 | $_Enc2xs = $0; |
850 | $_Version = $VERSION; |
851 | $_Inc = $INC{"Encode.pm"}; $_Inc =~ s/\.pm$//o; |
852 | $_Name = shift; |
853 | $_TableFiles = join(",", map {qq('$_')} @_); |
854 | $_Now = scalar localtime(); |
855 | warn "Generating Makefile.PL\n"; |
856 | _print_expand("$_Inc/Makefile_PL.e2x", "Makefile.PL"); |
857 | warn "Generating $_Name.pm\n"; |
858 | _print_expand("$_Inc/_PM.e2x", "$_Name.pm"); |
859 | warn "Generating t/$_Name.t\n"; |
860 | _print_expand("$_Inc/_T.e2x", "t/$_Name.t"); |
861 | warn "Generating README\n"; |
862 | _print_expand("$_Inc/README.e2x", "README"); |
863 | warn "Generating t/$_Name.t\n"; |
864 | _print_expand("$_Inc/Changes.e2x", "Changes"); |
865 | exit; |
866 | } |
867 | |
868 | sub _print_expand{ |
67d7b5ef |
869 | eval { require File::Basename; }; |
870 | $@ and die "File::Basename needed. Are you on miniperl?;\nerror: $@\n"; |
871 | File::Basename->import(); |
3ef515df |
872 | my ($src, $dst) = @_; |
873 | open my $in, $src or die "$src : $!"; |
874 | if ((my $d = dirname($dst)) ne '.'){ |
875 | -d $d or mkdir $d, 0755 or die "mkdir $d : $!"; |
876 | } |
877 | open my $out, ">$dst" or die "$!"; |
878 | my $asis = 0; |
879 | while (<$in>){ |
880 | if (/^#### END_OF_HEADER/){ |
881 | $asis = 1; next; |
882 | } |
883 | s/(\$_[A-Z][A-Za-z0-9]+)_/$1/gee unless $asis; |
884 | print $out $_; |
67d7b5ef |
885 | } |
67d7b5ef |
886 | } |
67d7b5ef |
887 | __END__ |
888 | |
889 | =head1 NAME |
890 | |
891 | enc2xs -- Perl Encode Module Generator |
892 | |
893 | =head1 SYNOPSIS |
894 | |
895 | enc2xs -M ModName mapfiles... |
896 | enc2xs -[options] |
897 | |
898 | =head1 DESCRIPTION |
899 | |
900 | F<enc2xs> builds a Perl extension for use by Encode from either |
901 | Unicode Character Mapping files (.ucm) or Tcl Encoding Files |
902 | (.enc) Besides internally used during the build process of Encode |
903 | module, you can use F<enc2xs> to add your own encoding to perl. No |
904 | knowledge on XS is necessary. |
905 | |
906 | =head1 Quick Guide |
907 | |
908 | If what you want to know as little about Perl possible but needs to |
909 | add a new encoding, just read this chapter and forget the rest. |
910 | |
911 | =over 4 |
912 | |
913 | =item 0. |
914 | |
915 | Have a .ucm file ready. You can get it from somewhere or you can |
916 | write your own from scratch or you can grab one from Encode |
917 | distribution and customize. For UCM format, see the next Chapter. |
918 | In the example below, I'll call my theoretical encoding myascii, |
919 | defined inI<my.ucm>. C<$> is a shell prompt. |
920 | |
921 | $ ls -F |
922 | my.ucm |
923 | |
924 | =item 1. |
925 | |
926 | Issue a command as follows; |
927 | |
928 | $ enc2xs -M My my.ucm |
3ef515df |
929 | generating Makefile.PL |
930 | generating My.pm |
931 | generating README |
932 | generating Changes |
67d7b5ef |
933 | |
934 | Now take a look at your current directory. It should look like this. |
935 | |
936 | $ ls -F |
937 | Makefile.PL My.pm my.ucm t/ |
938 | |
939 | The following files are created. |
940 | |
941 | Makefle.PL - MakeMaker script |
942 | My.pm - Encode Submodule |
943 | t/My.t - test file |
944 | |
037b88d6 |
945 | =item 1.1. |
946 | |
947 | If you want *.ucm installed together with the modules, do as follows; |
948 | |
949 | $ mkdir Encode |
950 | $ mv *.ucm Encode |
951 | $ enc2xs -M My Encode/*ucm |
952 | |
67d7b5ef |
953 | =item 2. |
954 | |
955 | Edit the files generated. You don't have to if you have no time AND no |
956 | intention to give it to someone else. But it is a good idea to edit |
957 | pod and add more tests. |
958 | |
959 | =item 3. |
960 | |
961 | Now issue a command all Perl Mongers love; |
962 | |
963 | $ perl5.7.3 Makefile.PL |
964 | Writing Makefile for Encode::My |
965 | |
966 | =item 4. |
967 | |
968 | Now all you have to do is make. |
969 | |
970 | $ make |
971 | cp My.pm blib/lib/Encode/My.pm |
972 | /usr/local/bin/perl /usr/local/bin/enc2xs -Q -O \ |
973 | -o encode_t.c -f encode_t.fnm |
974 | Reading myascii (myascii) |
975 | Writing compiled form |
976 | 128 bytes in string tables |
977 | 384 bytes (25%) saved spotting duplicates |
978 | 1 bytes (99.2%) saved using substrings |
979 | .... |
980 | chmod 644 blib/arch/auto/Encode/My/My.bs |
981 | $ |
982 | |
983 | The time it takes varies how fast your machine is and how large your |
984 | encoding is. Unless you are working on something big like euc-tw, it |
985 | won't take too long. |
986 | |
987 | =item 5. |
988 | |
989 | You can "make install" already but you should test first. |
990 | |
991 | $ make test |
992 | PERL_DL_NONLAZY=1 /usr/local/bin/perl -Iblib/arch -Iblib/lib \ |
993 | -e 'use Test::Harness qw(&runtests $verbose); \ |
994 | $verbose=0; runtests @ARGV;' t/*.t |
995 | t/My....ok |
996 | All tests successful. |
997 | Files=1, Tests=2, 0 wallclock secs |
998 | ( 0.09 cusr + 0.01 csys = 0.09 CPU) |
999 | |
1000 | =item 6. |
1001 | |
1002 | If you are content with the test result, just "make install" |
1003 | |
1004 | =back |
1005 | |
1006 | =head1 The Unicode Character Map |
1007 | |
1008 | Encode uses The Unicode Character Map (UCM) for source character |
1009 | mappings. This format is used by ICU package of IBM and adopted by |
1010 | Nick Ing-Simmons. Since UCM is more flexible than Tcl's Encoding Map |
1011 | and far more user-friendly, This is the recommended formet for |
1012 | Encode now. |
1013 | |
1014 | UCM file looks like this. |
1015 | |
1016 | # |
1017 | # Comments |
1018 | # |
1019 | <code_set_name> "US-ascii" # Required |
1020 | <code_set_alias> "ascii" # Optional |
1021 | <mb_cur_min> 1 # Required; usually 1 |
1022 | <mb_cur_max> 1 # Max. # of bytes/char |
1023 | <subchar> \x3F # Substitution char |
1024 | # |
1025 | CHARMAP |
1026 | <U0000> \x00 |0 # <control> |
1027 | <U0001> \x01 |0 # <control> |
1028 | <U0002> \x02 |0 # <control> |
1029 | .... |
1030 | <U007C> \x7C |0 # VERTICAL LINE |
1031 | <U007D> \x7D |0 # RIGHT CURLY BRACKET |
1032 | <U007E> \x7E |0 # TILDE |
1033 | <U007F> \x7F |0 # <control> |
1034 | END CHARMAP |
1035 | |
1036 | =over 4 |
1037 | |
1038 | =item * |
1039 | |
1040 | Anything that follows C<#> is treated as comments. |
1041 | |
1042 | =item * |
1043 | |
1044 | The header section continues until CHARMAP. This section Has a form of |
1045 | I<E<lt>keywordE<gt> value>, one at a line. For a value, strings must |
1046 | be quoted. Barewords are treated as numbers. I<\xXX> represents a |
1047 | byte. |
1048 | |
1049 | Most of the keywords are self-explanatory. I<subchar> means |
1050 | substitution character, not subcharacter. When you decode a Unicode |
1051 | sequence to this encoding but no matching character is found, the byte |
1052 | sequence defined here will be used. For most cases, the value here is |
1053 | \x3F, in ASCII this is a question mark. |
1054 | |
1055 | =item * |
1056 | |
1057 | CHARMAP starts the character map section. Each line has a form as |
1058 | follows; |
1059 | |
1060 | <UXXXX> \xXX.. |0 # comment |
1061 | ^ ^ ^ |
1062 | | | +- Fallback flag |
1063 | | +-------- Encoded byte sequence |
1064 | +-------------- Unicode Character ID in hex |
1065 | |
1066 | The format is roughly the same as a header section except for fallback |
1067 | flag. It is | followed by 0..3. And their meaning as follows |
1068 | |
1069 | =over 2 |
1070 | |
1071 | =item |0 |
1072 | |
1073 | Round trip safe. A character decoded to Unicode encodes back to the |
1074 | same byte sequence. most character belong to this. |
1075 | |
1076 | =item |1 |
1077 | |
1078 | Fallback for unicode -> encoding. When seen, enc2xs adds this |
1079 | character for encode map only |
1080 | |
1081 | =item |2 |
1082 | |
1083 | Skip sub-char mapping should there be no code point. |
1084 | |
1085 | =item |3 |
1086 | |
1087 | Fallback for encoding -> unicode. When seen, enc2xs adds this |
1088 | character for decode map only |
1089 | |
1090 | =back |
1091 | |
1092 | =item * |
1093 | |
1094 | And finally, END OF CHARMAP ends the section. |
1095 | |
1096 | =back |
1097 | |
1098 | Needless to say, if you are manually creating a UCM file, you should |
1099 | copy ascii.ucm or existing encoding which is close to yours than write |
1100 | your own from scratch. |
1101 | |
1102 | When you do so, make sure you leave at least B<U0000> to B<U0020> as |
1103 | is, unless your environment is on EBCDIC. |
1104 | |
1105 | B<CAVEAT>: not all features in UCM are implemented. For example, |
1106 | icu:state is not used. Because of that, you need to write a perl |
1107 | module if you want to support algorithmical encodings, notablly |
1108 | ISO-2022 series. Such modules include L<Encode::JP::2022_JP>, |
1109 | L<Encode::KR::2022_KR>, and L<Encode::TW::HZ>. |
1110 | |
1111 | =head1 Bookmarks |
1112 | |
1113 | ICU Home Page |
1114 | L<http://oss.software.ibm.com/icu/> |
1115 | |
1116 | ICU Character Mapping Tables |
1117 | L<http://oss.software.ibm.com/icu/charset/> |
1118 | |
1119 | ICU:Conversion Data |
1120 | L<http://oss.software.ibm.com/icu/userguide/conversion-data.html> |
1121 | |
1122 | =head1 SEE ALSO |
1123 | |
1124 | L<Encode>, |
1125 | L<perlmod>, |
1126 | L<perlpod> |
1127 | |
1128 | =cut |
1129 | |
1130 | # -Q to disable the duplicate codepoint test |
1131 | # -S make mapping errors fatal |
1132 | # -q to remove comments written to output files |
1133 | # -O to enable the (brute force) substring optimiser |
1134 | # -o <output> to specify the output file name (else it's the first arg) |
1135 | # -f <inlist> to give a file with a list of input files (else use the args) |
1136 | # -n <name> to name the encoding (else use the basename of the input file. |
1137 | |
1138 | With %seen holding array refs: |
1139 | |
1140 | 865.66 real 28.80 user 8.79 sys |
1141 | 7904 maximum resident set size |
1142 | 1356 average shared memory size |
1143 | 18566 average unshared data size |
1144 | 229 average unshared stack size |
1145 | 46080 page reclaims |
1146 | 33373 page faults |
1147 | |
1148 | With %seen holding simple scalars: |
1149 | |
1150 | 342.16 real 27.11 user 3.54 sys |
1151 | 8388 maximum resident set size |
1152 | 1394 average shared memory size |
1153 | 14969 average unshared data size |
1154 | 236 average unshared stack size |
1155 | 28159 page reclaims |
1156 | 9839 page faults |
1157 | |
1158 | Yes, 5 minutes is faster than 15. Above is for CP936 in CN. Only difference is |
1159 | how %seen is storing things its seen. So it is pathalogically bad on a 16M |
1160 | RAM machine, but it's going to help even on modern machines. |
1161 | Swapping is bad, m'kay :-) |