… | |
… | |
27 | #define F_SPACE_AFTER 0x00000040UL |
27 | #define F_SPACE_AFTER 0x00000040UL |
28 | #define F_ALLOW_NONREF 0x00000100UL |
28 | #define F_ALLOW_NONREF 0x00000100UL |
29 | #define F_SHRINK 0x00000200UL |
29 | #define F_SHRINK 0x00000200UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
|
|
32 | #define F_RELAXED 0x00001000UL |
|
|
33 | |
32 | #define F_MAXDEPTH 0xf8000000UL |
34 | #define F_MAXDEPTH 0xf8000000UL |
33 | #define S_MAXDEPTH 27 |
35 | #define S_MAXDEPTH 27 |
34 | #define F_MAXSIZE 0x01f00000UL |
36 | #define F_MAXSIZE 0x01f00000UL |
35 | #define S_MAXSIZE 20 |
37 | #define S_MAXSIZE 20 |
36 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
… | |
… | |
335 | --enc->indent; |
337 | --enc->indent; |
336 | encode_indent (enc); encode_ch (enc, ']'); |
338 | encode_indent (enc); encode_ch (enc, ']'); |
337 | } |
339 | } |
338 | |
340 | |
339 | static void |
341 | static void |
340 | encode_he (enc_t *enc, HE *he) |
342 | encode_hk (enc_t *enc, HE *he) |
341 | { |
343 | { |
342 | encode_ch (enc, '"'); |
344 | encode_ch (enc, '"'); |
343 | |
345 | |
344 | if (HeKLEN (he) == HEf_SVKEY) |
346 | if (HeKLEN (he) == HEf_SVKEY) |
345 | { |
347 | { |
… | |
… | |
358 | encode_ch (enc, '"'); |
360 | encode_ch (enc, '"'); |
359 | |
361 | |
360 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
362 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
361 | encode_ch (enc, ':'); |
363 | encode_ch (enc, ':'); |
362 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
364 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
363 | encode_sv (enc, HeVAL (he)); |
|
|
364 | } |
365 | } |
365 | |
366 | |
366 | // compare hash entries, used when all keys are bytestrings |
367 | // compare hash entries, used when all keys are bytestrings |
367 | static int |
368 | static int |
368 | he_cmp_fast (const void *a_, const void *b_) |
369 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
373 | HE *b = *(HE **)b_; |
374 | HE *b = *(HE **)b_; |
374 | |
375 | |
375 | STRLEN la = HeKLEN (a); |
376 | STRLEN la = HeKLEN (a); |
376 | STRLEN lb = HeKLEN (b); |
377 | STRLEN lb = HeKLEN (b); |
377 | |
378 | |
378 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
379 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
379 | cmp = la - lb; |
380 | cmp = lb - la; |
380 | |
381 | |
381 | return cmp; |
382 | return cmp; |
382 | } |
383 | } |
383 | |
384 | |
384 | // compare hash entries, used when some keys are sv's or utf-x |
385 | // compare hash entries, used when some keys are sv's or utf-x |
385 | static int |
386 | static int |
386 | he_cmp_slow (const void *a, const void *b) |
387 | he_cmp_slow (const void *a, const void *b) |
387 | { |
388 | { |
388 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
389 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
389 | } |
390 | } |
390 | |
391 | |
391 | static void |
392 | static void |
392 | encode_hv (enc_t *enc, HV *hv) |
393 | encode_hv (enc_t *enc, HV *hv) |
393 | { |
394 | { |
|
|
395 | HE *he; |
394 | int count, i; |
396 | int count; |
395 | |
397 | |
396 | if (enc->indent >= enc->maxdepth) |
398 | if (enc->indent >= enc->maxdepth) |
397 | croak ("data structure too deep (hit recursion limit)"); |
399 | croak ("data structure too deep (hit recursion limit)"); |
398 | |
400 | |
399 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
401 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
400 | |
402 | |
401 | if ((count = hv_iterinit (hv))) |
|
|
402 | { |
|
|
403 | // for canonical output we have to sort by keys first |
403 | // for canonical output we have to sort by keys first |
404 | // actually, this is mostly due to the stupid so-called |
404 | // actually, this is mostly due to the stupid so-called |
405 | // security workaround added somewhere in 5.8.x. |
405 | // security workaround added somewhere in 5.8.x. |
406 | // that randomises hash orderings |
406 | // that randomises hash orderings |
407 | if (enc->json.flags & F_CANONICAL) |
407 | if (enc->json.flags & F_CANONICAL) |
|
|
408 | { |
|
|
409 | int count = hv_iterinit (hv); |
|
|
410 | |
|
|
411 | if (SvMAGICAL (hv)) |
408 | { |
412 | { |
|
|
413 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
414 | // but I don't care for the speed of this special case. |
|
|
415 | // note also that we will run into undefined behaviour when the two iterations |
|
|
416 | // do not result in the same count, something I might care for in some later release. |
|
|
417 | |
|
|
418 | count = 0; |
|
|
419 | while (hv_iternext (hv)) |
|
|
420 | ++count; |
|
|
421 | |
|
|
422 | hv_iterinit (hv); |
|
|
423 | } |
|
|
424 | |
|
|
425 | if (count) |
|
|
426 | { |
409 | int fast = 1; |
427 | int i, fast = 1; |
410 | HE *he; |
|
|
411 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
428 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
412 | HE **hes = _alloca (count * sizeof (HE)); |
429 | HE **hes = _alloca (count * sizeof (HE)); |
413 | #else |
430 | #else |
414 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
431 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
415 | #endif |
432 | #endif |
… | |
… | |
442 | |
459 | |
443 | FREETMPS; |
460 | FREETMPS; |
444 | LEAVE; |
461 | LEAVE; |
445 | } |
462 | } |
446 | |
463 | |
447 | for (i = 0; i < count; ++i) |
464 | while (count--) |
448 | { |
465 | { |
449 | encode_indent (enc); |
466 | encode_indent (enc); |
|
|
467 | he = hes [count]; |
450 | encode_he (enc, hes [i]); |
468 | encode_hk (enc, he); |
|
|
469 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
451 | |
470 | |
452 | if (i < count - 1) |
471 | if (count) |
453 | encode_comma (enc); |
472 | encode_comma (enc); |
454 | } |
473 | } |
455 | |
|
|
456 | encode_nl (enc); |
|
|
457 | } |
474 | } |
|
|
475 | } |
458 | else |
476 | else |
459 | { |
477 | { |
|
|
478 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
460 | HE *he = hv_iternext (hv); |
479 | if ((he = hv_iternext (hv))) |
461 | |
|
|
462 | for (;;) |
480 | for (;;) |
463 | { |
481 | { |
464 | encode_indent (enc); |
482 | encode_indent (enc); |
465 | encode_he (enc, he); |
483 | encode_hk (enc, he); |
|
|
484 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
466 | |
485 | |
467 | if (!(he = hv_iternext (hv))) |
486 | if (!(he = hv_iternext (hv))) |
468 | break; |
487 | break; |
469 | |
488 | |
470 | encode_comma (enc); |
489 | encode_comma (enc); |
471 | } |
490 | } |
|
|
491 | } |
472 | |
492 | |
473 | encode_nl (enc); |
493 | encode_nl (enc); |
474 | } |
|
|
475 | } |
|
|
476 | |
494 | |
477 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
495 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
478 | } |
496 | } |
479 | |
497 | |
480 | // encode objects, arrays and special \0=false and \1=true values. |
498 | // encode objects, arrays and special \0=false and \1=true values. |
… | |
… | |
692 | |
710 | |
693 | if (ch > 0x20 |
711 | if (ch > 0x20 |
694 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
712 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
695 | break; |
713 | break; |
696 | |
714 | |
|
|
715 | if (ch == '#' && dec->json.flags & F_RELAXED) |
|
|
716 | ++dec->cur; |
|
|
717 | |
697 | ++dec->cur; |
718 | ++dec->cur; |
698 | } |
719 | } |
699 | } |
720 | } |
700 | |
721 | |
701 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
722 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
… | |
… | |
1037 | |
1058 | |
1038 | if (*dec->cur != ',') |
1059 | if (*dec->cur != ',') |
1039 | ERR (", or ] expected while parsing array"); |
1060 | ERR (", or ] expected while parsing array"); |
1040 | |
1061 | |
1041 | ++dec->cur; |
1062 | ++dec->cur; |
|
|
1063 | |
|
|
1064 | decode_ws (dec); |
|
|
1065 | |
|
|
1066 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1067 | { |
|
|
1068 | ++dec->cur; |
|
|
1069 | break; |
|
|
1070 | } |
1042 | } |
1071 | } |
1043 | |
1072 | |
1044 | DEC_DEC_DEPTH; |
1073 | DEC_DEC_DEPTH; |
1045 | return newRV_noinc ((SV *)av); |
1074 | return newRV_noinc ((SV *)av); |
1046 | |
1075 | |
… | |
… | |
1062 | if (*dec->cur == '}') |
1091 | if (*dec->cur == '}') |
1063 | ++dec->cur; |
1092 | ++dec->cur; |
1064 | else |
1093 | else |
1065 | for (;;) |
1094 | for (;;) |
1066 | { |
1095 | { |
1067 | decode_ws (dec); EXPECT_CH ('"'); |
1096 | EXPECT_CH ('"'); |
1068 | |
1097 | |
1069 | // heuristic: assume that |
1098 | // heuristic: assume that |
1070 | // a) decode_str + hv_store_ent are abysmally slow. |
1099 | // a) decode_str + hv_store_ent are abysmally slow. |
1071 | // b) most hash keys are short, simple ascii text. |
1100 | // b) most hash keys are short, simple ascii text. |
1072 | // => try to "fast-match" such strings to avoid |
1101 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1086 | if (!key) |
1115 | if (!key) |
1087 | goto fail; |
1116 | goto fail; |
1088 | |
1117 | |
1089 | decode_ws (dec); EXPECT_CH (':'); |
1118 | decode_ws (dec); EXPECT_CH (':'); |
1090 | |
1119 | |
|
|
1120 | decode_ws (dec); |
1091 | value = decode_sv (dec); |
1121 | value = decode_sv (dec); |
1092 | if (!value) |
1122 | if (!value) |
1093 | { |
1123 | { |
1094 | SvREFCNT_dec (key); |
1124 | SvREFCNT_dec (key); |
1095 | goto fail; |
1125 | goto fail; |
… | |
… | |
1107 | int len = p - key; |
1137 | int len = p - key; |
1108 | dec->cur = p + 1; |
1138 | dec->cur = p + 1; |
1109 | |
1139 | |
1110 | decode_ws (dec); EXPECT_CH (':'); |
1140 | decode_ws (dec); EXPECT_CH (':'); |
1111 | |
1141 | |
|
|
1142 | decode_ws (dec); |
1112 | value = decode_sv (dec); |
1143 | value = decode_sv (dec); |
1113 | if (!value) |
1144 | if (!value) |
1114 | goto fail; |
1145 | goto fail; |
1115 | |
1146 | |
1116 | hv_store (hv, key, len, value, 0); |
1147 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1132 | |
1163 | |
1133 | if (*dec->cur != ',') |
1164 | if (*dec->cur != ',') |
1134 | ERR (", or } expected while parsing object/hash"); |
1165 | ERR (", or } expected while parsing object/hash"); |
1135 | |
1166 | |
1136 | ++dec->cur; |
1167 | ++dec->cur; |
|
|
1168 | |
|
|
1169 | decode_ws (dec); |
|
|
1170 | |
|
|
1171 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1172 | { |
|
|
1173 | ++dec->cur; |
|
|
1174 | break; |
|
|
1175 | } |
1137 | } |
1176 | } |
1138 | |
1177 | |
1139 | DEC_DEC_DEPTH; |
1178 | DEC_DEC_DEPTH; |
1140 | sv = newRV_noinc ((SV *)hv); |
1179 | sv = newRV_noinc ((SV *)hv); |
1141 | |
1180 | |
… | |
… | |
1206 | } |
1245 | } |
1207 | |
1246 | |
1208 | static SV * |
1247 | static SV * |
1209 | decode_sv (dec_t *dec) |
1248 | decode_sv (dec_t *dec) |
1210 | { |
1249 | { |
1211 | decode_ws (dec); |
|
|
1212 | |
|
|
1213 | // the beauty of JSON: you need exactly one character lookahead |
1250 | // the beauty of JSON: you need exactly one character lookahead |
1214 | // to parse anything. |
1251 | // to parse anything. |
1215 | switch (*dec->cur) |
1252 | switch (*dec->cur) |
1216 | { |
1253 | { |
1217 | case '"': ++dec->cur; return decode_str (dec); |
1254 | case '"': ++dec->cur; return decode_str (dec); |
… | |
… | |
1301 | |
1338 | |
1302 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1339 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1303 | dec.json.flags |= F_HOOK; |
1340 | dec.json.flags |= F_HOOK; |
1304 | |
1341 | |
1305 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1342 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1343 | |
|
|
1344 | decode_ws (&dec); |
1306 | sv = decode_sv (&dec); |
1345 | sv = decode_sv (&dec); |
1307 | |
1346 | |
1308 | if (!(offset_return || !sv)) |
1347 | if (!(offset_return || !sv)) |
1309 | { |
1348 | { |
1310 | // check for trailing garbage |
1349 | // check for trailing garbage |
… | |
… | |
1407 | pretty = F_PRETTY |
1446 | pretty = F_PRETTY |
1408 | allow_nonref = F_ALLOW_NONREF |
1447 | allow_nonref = F_ALLOW_NONREF |
1409 | shrink = F_SHRINK |
1448 | shrink = F_SHRINK |
1410 | allow_blessed = F_ALLOW_BLESSED |
1449 | allow_blessed = F_ALLOW_BLESSED |
1411 | convert_blessed = F_CONV_BLESSED |
1450 | convert_blessed = F_CONV_BLESSED |
|
|
1451 | relaxed = F_RELAXED |
1412 | PPCODE: |
1452 | PPCODE: |
1413 | { |
1453 | { |
1414 | if (enable) |
1454 | if (enable) |
1415 | self->flags |= ix; |
1455 | self->flags |= ix; |
1416 | else |
1456 | else |