--- JSON-XS/XS.xs 2007/03/23 15:10:55 1.6 +++ JSON-XS/XS.xs 2007/06/06 14:52:49 1.35 @@ -5,47 +5,55 @@ #include "assert.h" #include "string.h" #include "stdlib.h" +#include "stdio.h" -#define F_ASCII 0x00000001 -#define F_UTF8 0x00000002 -#define F_INDENT 0x00000004 -#define F_CANONICAL 0x00000008 -#define F_SPACE_BEFORE 0x00000010 -#define F_SPACE_AFTER 0x00000020 -#define F_JSON_RPC 0x00000040 -#define F_ALLOW_NONREF 0x00000080 -#define F_SHRINK 0x00000100 +#if defined(__BORLANDC__) || defined(_MSC_VER) +# define snprintf _snprintf // C compilers have this in stdio.h +#endif + +#define F_ASCII 0x00000001UL +#define F_LATIN1 0x00000002UL +#define F_UTF8 0x00000004UL +#define F_INDENT 0x00000008UL +#define F_CANONICAL 0x00000010UL +#define F_SPACE_BEFORE 0x00000020UL +#define F_SPACE_AFTER 0x00000040UL +#define F_ALLOW_NONREF 0x00000100UL +#define F_SHRINK 0x00000200UL +#define F_MAXDEPTH 0xf8000000UL +#define S_MAXDEPTH 27 + +#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) + +// F_SELFCONVERT? <=> to_json/toJson +// F_BLESSED? <=> { $__class__$ => } #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER -#define F_DEFAULT 0 +#define F_DEFAULT (9UL << S_MAXDEPTH) #define INIT_SIZE 32 // initial scalar size to be allocated +#define INDENT_STEP 3 // spaces per indentation level + +#define SHORT_STRING_LEN 512 // special-case strings of up to this size #define SB do { #define SE } while (0) -static HV *json_stash; +#if __GNUC__ >= 3 +# define expect(expr,value) __builtin_expect ((expr),(value)) +# define inline inline +#else +# define expect(expr,value) (expr) +# define inline static +#endif -// structure used for encoding JSON -typedef struct -{ - char *cur; - STRLEN len; // SvLEN (sv) - char *end; // SvEND (sv) - SV *sv; - UV flags; - int max_recurse; - int indent; -} enc_t; +#define expect_false(expr) expect ((expr) != 0, 0) +#define expect_true(expr) expect ((expr) != 0, 1) -// structure used for decoding JSON -typedef struct -{ - char *cur; - char *end; - const char *err; - UV flags; -} dec_t; +static HV *json_stash; // JSON::XS:: + +///////////////////////////////////////////////////////////////////////////// +// utility functions static UV * SvJSON (SV *sv) @@ -56,21 +64,69 @@ return &SvUVX (SvRV (sv)); } +static void +shrink (SV *sv) +{ + sv_utf8_downgrade (sv, 1); + if (SvLEN (sv) > SvCUR (sv) + 1) + { +#ifdef SvPV_shrink_to_cur + SvPV_shrink_to_cur (sv); +#elif defined (SvPV_renew) + SvPV_renew (sv, SvCUR (sv) + 1); +#endif + } +} + +// decode an utf-8 character and return it, or (UV)-1 in +// case of an error. +// we special-case "safe" characters from U+80 .. U+7FF, +// but use the very good perl function to parse anything else. +// note that we never call this function for a ascii codepoints +inline UV +decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) +{ + if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) + return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); + else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) + { + *clen = 2; + return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); + } + else + { + *clen = (STRLEN)-1; + return (UV)-1; + } +} + ///////////////////////////////////////////////////////////////////////////// +// encoder -static void +// structure used for encoding JSON +typedef struct +{ + char *cur; // SvPVX (sv) + current output position + char *end; // SvEND (sv) + SV *sv; // result scalar + U32 flags; // F_* + U32 indent; // indentation level + U32 maxdepth; // max. indentation/recursion level +} enc_t; + +inline void need (enc_t *enc, STRLEN len) { - if (enc->cur + len >= enc->end) + if (expect_false (enc->cur + len >= enc->end)) { STRLEN cur = enc->cur - SvPVX (enc->sv); SvGROW (enc->sv, cur + len + 1); enc->cur = SvPVX (enc->sv) + cur; - enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); + enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; } } -static void +inline void encode_ch (enc_t *enc, char ch) { need (enc, 1); @@ -88,15 +144,15 @@ { unsigned char ch = *(unsigned char *)str; - if (ch >= 0x20 && ch < 0x80) // most common case + if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case { - if (ch == '"') // but with slow exceptions + if (expect_false (ch == '"')) // but with slow exceptions { need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = '"'; } - else if (ch == '\\') + else if (expect_false (ch == '\\')) { need (enc, len += 1); *enc->cur++ = '\\'; @@ -124,9 +180,10 @@ if (is_utf8) { - uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); + //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); + uch = decode_utf8 (str, end - str, &clen); if (clen == (STRLEN)-1) - croak ("malformed UTF-8 character in string, cannot convert to JSON"); + croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); } else { @@ -134,14 +191,17 @@ clen = 1; } - if (uch < 0x80 || enc->flags & F_ASCII) + if (uch > 0x10FFFFUL) + croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); + + if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF)) { if (uch > 0xFFFFUL) { need (enc, len += 11); sprintf (enc->cur, "\\u%04x\\u%04x", - (uch - 0x10000) / 0x400 + 0xD800, - (uch - 0x10000) % 0x400 + 0xDC00); + (int)((uch - 0x10000) / 0x400 + 0xD800), + (int)((uch - 0x10000) % 0x400 + 0xDC00)); enc->cur += 12; } else @@ -158,6 +218,11 @@ str += clen; } + else if (enc->flags & F_LATIN1) + { + *enc->cur++ = uch; + str += clen; + } else if (is_utf8) { need (enc, len += clen); @@ -169,7 +234,7 @@ } else { - need (enc, 10); // never more than 11 bytes needed + need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); ++str; } @@ -181,25 +246,46 @@ } } -#define INDENT SB \ - if (enc->flags & F_INDENT) \ - { \ - int i_; \ - need (enc, enc->indent); \ - for (i_ = enc->indent * 3; i_--; )\ - encode_ch (enc, ' '); \ - } \ - SE +inline void +encode_indent (enc_t *enc) +{ + if (enc->flags & F_INDENT) + { + int spaces = enc->indent * INDENT_STEP; -#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE -#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE -#define COMMA SB \ - encode_ch (enc, ','); \ - if (enc->flags & F_INDENT) \ - NL; \ - else if (enc->flags & F_SPACE_AFTER) \ - SPACE; \ - SE + need (enc, spaces); + memset (enc->cur, ' ', spaces); + enc->cur += spaces; + } +} + +inline void +encode_space (enc_t *enc) +{ + need (enc, 1); + encode_ch (enc, ' '); +} + +inline void +encode_nl (enc_t *enc) +{ + if (enc->flags & F_INDENT) + { + need (enc, 1); + encode_ch (enc, '\n'); + } +} + +inline void +encode_comma (enc_t *enc) +{ + encode_ch (enc, ','); + + if (enc->flags & F_INDENT) + encode_nl (enc); + else if (enc->flags & F_SPACE_AFTER) + encode_space (enc); +} static void encode_sv (enc_t *enc, SV *sv); @@ -208,22 +294,25 @@ { int i, len = av_len (av); - encode_ch (enc, '['); NL; + if (enc->indent >= enc->maxdepth) + croak ("data structure too deep (hit recursion limit)"); + + encode_ch (enc, '['); encode_nl (enc); ++enc->indent; for (i = 0; i <= len; ++i) { - INDENT; + encode_indent (enc); encode_sv (enc, *av_fetch (av, i, 0)); if (i < len) - COMMA; + encode_comma (enc); } - NL; + encode_nl (enc); --enc->indent; - INDENT; encode_ch (enc, ']'); + encode_indent (enc); encode_ch (enc, ']'); } static void @@ -247,9 +336,9 @@ encode_ch (enc, '"'); - if (enc->flags & F_SPACE_BEFORE) SPACE; + if (enc->flags & F_SPACE_BEFORE) encode_space (enc); encode_ch (enc, ':'); - if (enc->flags & F_SPACE_AFTER ) SPACE; + if (enc->flags & F_SPACE_AFTER ) encode_space (enc); encode_sv (enc, HeVAL (he)); } @@ -265,8 +354,8 @@ STRLEN la = HeKLEN (a); STRLEN lb = HeKLEN (b); - if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) - cmp = la < lb ? -1 : la == lb ? 0 : 1; + if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) + cmp = la - lb; return cmp; } @@ -283,7 +372,10 @@ { int count, i; - encode_ch (enc, '{'); NL; ++enc->indent; + if (enc->indent >= enc->maxdepth) + croak ("data structure too deep (hit recursion limit)"); + + encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; if ((count = hv_iterinit (hv))) { @@ -293,8 +385,13 @@ // that randomises hash orderings if (enc->flags & F_CANONICAL) { - HE *he, *hes [count]; int fast = 1; + HE *he; +#if defined(__BORLANDC__) || defined(_MSC_VER) + HE **hes = _alloca (count * sizeof (HE)); +#else + HE *hes [count]; // if your compiler dies here, you need to enable C99 mode +#endif i = 0; while ((he = hv_iternext (hv))) @@ -310,50 +407,81 @@ qsort (hes, count, sizeof (HE *), he_cmp_fast); else { - // hack to disable "use bytes" - COP *oldcop = PL_curcop, cop; + // hack to forcefully disable "use bytes" + COP cop = *PL_curcop; cop.op_private = 0; - PL_curcop = &cop; + ENTER; SAVETMPS; + + SAVEVPTR (PL_curcop); + PL_curcop = &cop; + qsort (hes, count, sizeof (HE *), he_cmp_slow); - FREETMPS; - PL_curcop = oldcop; + FREETMPS; + LEAVE; } for (i = 0; i < count; ++i) { - INDENT; + encode_indent (enc); encode_he (enc, hes [i]); if (i < count - 1) - COMMA; + encode_comma (enc); } - NL; + encode_nl (enc); } else { - SV *sv; HE *he = hv_iternext (hv); for (;;) { - INDENT; + encode_indent (enc); encode_he (enc, he); if (!(he = hv_iternext (hv))) break; - COMMA; + encode_comma (enc); } - NL; + encode_nl (enc); } } - --enc->indent; INDENT; encode_ch (enc, '}'); + --enc->indent; encode_indent (enc); encode_ch (enc, '}'); +} + +// encode objects, arrays and special \0=false and \1=true values. +static void +encode_rv (enc_t *enc, SV *sv) +{ + svtype svt; + + SvGETMAGIC (sv); + svt = SvTYPE (sv); + + if (svt == SVt_PVHV) + encode_hv (enc, (HV *)sv); + else if (svt == SVt_PVAV) + encode_av (enc, (AV *)sv); + else if (svt < SVt_PVAV) + { + if (SvNIOK (sv) && SvIV (sv) == 0) + encode_str (enc, "false", 5, 0); + else if (SvNIOK (sv) && SvIV (sv) == 1) + encode_str (enc, "true", 4, 0); + else + croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", + SvPV_nolen (sv_2mortal (newRV_inc (sv)))); + } + else + croak ("encountered %s, but JSON can only represent references to arrays or hashes", + SvPV_nolen (sv_2mortal (newRV_inc (sv)))); } static void @@ -377,82 +505,118 @@ } else if (SvIOKp (sv)) { - need (enc, 64); - enc->cur += - SvIsUV(sv) - ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) - : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); - } - else if (SvROK (sv)) - { - if (!--enc->max_recurse) - croak ("data structure too deep (hit recursion limit)"); - - sv = SvRV (sv); - - switch (SvTYPE (sv)) + // we assume we can always read an IV as a UV + if (SvUV (sv) & ~(UV)0x7fff) { - case SVt_PVAV: encode_av (enc, (AV *)sv); break; - case SVt_PVHV: encode_hv (enc, (HV *)sv); break; - - default: - croak ("JSON can only represent references to arrays or hashes"); + need (enc, 32); + enc->cur += + SvIsUV(sv) + ? snprintf (enc->cur, 32, "%"UVuf, (UV)SvUVX (sv)) + : snprintf (enc->cur, 32, "%"IVdf, (IV)SvIVX (sv)); + } + else + { + // optimise the "small number case" + // code will likely be branchless and use only a single multiplication + I32 i = SvIV (sv); + U32 u; + + need (enc, 6); + + *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; + u = i < 0 ? -i : i; + + // convert to 4.28 fixed-point representation + u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits + + char digit, nz = 0; + + digit = u >> 28; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0xfffffff) * 5; + digit = u >> 27; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x7ffffff) * 5; + digit = u >> 26; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x3ffffff) * 5; + digit = u >> 25; *enc->cur = digit + '0'; nz |= digit; enc->cur += nz ? 1 : 0; u = (u & 0x1ffffff) * 5; + digit = u >> 24; *enc->cur = digit + '0'; nz |= digit; enc->cur += 1; } } + else if (SvROK (sv)) + encode_rv (enc, SvRV (sv)); else if (!SvOK (sv)) encode_str (enc, "null", 4, 0); else - croak ("encountered perl type that JSON cannot handle"); + croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", + SvPV_nolen (sv), SvFLAGS (sv)); } static SV * -encode_json (SV *scalar, UV flags) +encode_json (SV *scalar, U32 flags) { + enc_t enc; + if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) - croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)"); + croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); - enc_t enc; - enc.flags = flags; - enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); - enc.cur = SvPVX (enc.sv); - enc.end = SvEND (enc.sv); - enc.max_recurse = 0; - enc.indent = 0; + enc.flags = flags; + enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); + enc.cur = SvPVX (enc.sv); + enc.end = SvEND (enc.sv); + enc.indent = 0; + enc.maxdepth = DEC_DEPTH (flags); SvPOK_only (enc.sv); encode_sv (&enc, scalar); - if (!(flags & (F_ASCII | F_UTF8))) - SvUTF8_on (enc.sv); - SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); + *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings + + if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8))) + SvUTF8_on (enc.sv); -#ifdef SvPV_shrink_to_cur if (enc.flags & F_SHRINK) - SvPV_shrink_to_cur (enc.sv); -#endif + shrink (enc.sv); + return enc.sv; } ///////////////////////////////////////////////////////////////////////////// +// decoder + +// structure used for decoding JSON +typedef struct +{ + char *cur; // current parser pointer + char *end; // end of input string + const char *err; // parse error, if != 0 + U32 flags; // F_* + U32 depth; // recursion depth + U32 maxdepth; // recursion depth limit +} dec_t; -#define WS \ - for (;;) \ - { \ - char ch = *dec->cur; \ - if (ch > 0x20 \ - || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ - break; \ - ++dec->cur; \ +inline void +decode_ws (dec_t *dec) +{ + for (;;) + { + char ch = *dec->cur; + + if (ch > 0x20 + || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) + break; + + ++dec->cur; } +} #define ERR(reason) SB dec->err = reason; goto fail; SE + #define EXPECT_CH(ch) SB \ if (*dec->cur != ch) \ ERR (# ch " expected"); \ ++dec->cur; \ SE +#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") +#define DEC_DEC_DEPTH --dec->depth + static SV *decode_sv (dec_t *dec); static signed char decode_hexdigit[256]; @@ -461,15 +625,12 @@ decode_4hex (dec_t *dec) { signed char d1, d2, d3, d4; + unsigned char *cur = (unsigned char *)dec->cur; - d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; - if (d1 < 0) ERR ("four hexadecimal digits expected"); - d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; - if (d2 < 0) ERR ("four hexadecimal digits expected"); - d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]]; - if (d3 < 0) ERR ("four hexadecimal digits expected"); - d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]]; - if (d4 < 0) ERR ("four hexadecimal digits expected"); + d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected"); + d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected"); + d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected"); + d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected"); dec->cur += 4; @@ -482,138 +643,148 @@ return (UV)-1; } -#define APPEND_GROW(n) SB \ - if (cur + (n) >= end) \ - { \ - STRLEN ofs = cur - SvPVX (sv); \ - SvGROW (sv, ofs + (n) + 1); \ - cur = SvPVX (sv) + ofs; \ - end = SvEND (sv); \ - } \ - SE - -#define APPEND_CH(ch) SB \ - APPEND_GROW (1); \ - *cur++ = (ch); \ - SE - static SV * decode_str (dec_t *dec) { - SV *sv = NEWSV (0,2); + SV *sv = 0; int utf8 = 0; - char *cur = SvPVX (sv); - char *end = SvEND (sv); - for (;;) + do { - unsigned char ch = *(unsigned char *)dec->cur; + char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; + char *cur = buf; - if (ch == '"') - break; - else if (ch == '\\') + do { - switch (*++dec->cur) - { - case '\\': - case '/': - case '"': APPEND_CH (*dec->cur++); break; - - case 'b': APPEND_CH ('\010'); ++dec->cur; break; - case 't': APPEND_CH ('\011'); ++dec->cur; break; - case 'n': APPEND_CH ('\012'); ++dec->cur; break; - case 'f': APPEND_CH ('\014'); ++dec->cur; break; - case 'r': APPEND_CH ('\015'); ++dec->cur; break; + unsigned char ch = *(unsigned char *)dec->cur++; - case 'u': + if (expect_false (ch == '"')) + { + --dec->cur; + break; + } + else if (expect_false (ch == '\\')) + { + switch (*dec->cur) { - UV lo, hi; - ++dec->cur; + case '\\': + case '/': + case '"': *cur++ = *dec->cur++; break; + + case 'b': ++dec->cur; *cur++ = '\010'; break; + case 't': ++dec->cur; *cur++ = '\011'; break; + case 'n': ++dec->cur; *cur++ = '\012'; break; + case 'f': ++dec->cur; *cur++ = '\014'; break; + case 'r': ++dec->cur; *cur++ = '\015'; break; - hi = decode_4hex (dec); - if (hi == (UV)-1) - goto fail; - - // possibly a surrogate pair - if (hi >= 0xd800 && hi < 0xdc00) + case 'u': { - if (dec->cur [0] != '\\' || dec->cur [1] != 'u') - ERR ("missing low surrogate character in surrogate pair"); - - dec->cur += 2; + UV lo, hi; + ++dec->cur; - lo = decode_4hex (dec); - if (lo == (UV)-1) + hi = decode_4hex (dec); + if (hi == (UV)-1) goto fail; - if (lo < 0xdc00 || lo >= 0xe000) - ERR ("surrogate pair expected"); - - hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; - } - else if (hi >= 0xdc00 && hi < 0xe000) - ERR ("missing high surrogate character in surrogate pair"); + // possibly a surrogate pair + if (hi >= 0xd800) + if (hi < 0xdc00) + { + if (dec->cur [0] != '\\' || dec->cur [1] != 'u') + ERR ("missing low surrogate character in surrogate pair"); + + dec->cur += 2; + + lo = decode_4hex (dec); + if (lo == (UV)-1) + goto fail; + + if (lo < 0xdc00 || lo >= 0xe000) + ERR ("surrogate pair expected"); + + hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; + } + else if (hi < 0xe000) + ERR ("missing high surrogate character in surrogate pair"); - if (hi >= 0x80) - { - utf8 = 1; + if (hi >= 0x80) + { + utf8 = 1; - APPEND_GROW (4); // at most 4 bytes for 21 bits - cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); + cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); + } + else + *cur++ = hi; } - else - APPEND_CH (hi); - } - break; + break; - default: - --dec->cur; - ERR ("illegal backslash escape sequence in string"); + default: + --dec->cur; + ERR ("illegal backslash escape sequence in string"); + } } - } - else if (ch >= 0x20 && ch <= 0x7f) - APPEND_CH (*dec->cur++); - else if (ch >= 0x80) - { - STRLEN clen; - UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); - if (clen == (STRLEN)-1) - ERR ("malformed UTF-8 character in string, cannot convert to JSON"); - - APPEND_GROW (clen); - do + else if (expect_true (ch >= 0x20 && ch <= 0x7f)) + *cur++ = ch; + else if (ch >= 0x80) { - *cur++ = *dec->cur++; + STRLEN clen; + UV uch; + + --dec->cur; + + uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); + if (clen == (STRLEN)-1) + ERR ("malformed UTF-8 character in JSON string"); + + do + *cur++ = *dec->cur++; + while (--clen); + + utf8 = 1; } - while (--clen); + else + { + --dec->cur; - utf8 = 1; + if (!ch) + ERR ("unexpected end of string while parsing JSON string"); + else + ERR ("invalid character encountered while parsing JSON string"); + } } - else if (dec->cur == dec->end) - ERR ("unexpected end of string while parsing json string"); - else - ERR ("invalid character encountered"); - } + while (cur < buf + SHORT_STRING_LEN); - ++dec->cur; + { + STRLEN len = cur - buf; - SvCUR_set (sv, cur - SvPVX (sv)); + if (sv) + { + SvGROW (sv, SvCUR (sv) + len + 1); + memcpy (SvPVX (sv) + SvCUR (sv), buf, len); + SvCUR_set (sv, SvCUR (sv) + len); + } + else + sv = newSVpvn (buf, len); + } + } + while (*dec->cur != '"'); - SvPOK_only (sv); - *SvEND (sv) = 0; + ++dec->cur; - if (utf8) - SvUTF8_on (sv); + if (sv) + { + SvPOK_only (sv); + *SvEND (sv) = 0; -#ifdef SvPV_shrink_to_cur - if (dec->flags & F_SHRINK) - SvPV_shrink_to_cur (sv); -#endif + if (utf8) + SvUTF8_on (sv); + } + else + sv = newSVpvn ("", 0); return sv; fail: - SvREFCNT_dec (sv); return 0; } @@ -681,16 +852,36 @@ if (!is_nv) { - UV uv; - int numtype = grok_number (start, dec->cur - start, &uv); - if (numtype & IS_NUMBER_IN_UV) - if (numtype & IS_NUMBER_NEG) + // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so + if (*start == '-') + switch (dec->cur - start) { - if (uv < (UV)IV_MIN) - return newSViv (-(IV)uv); + case 2: return newSViv (-( start [1] - '0' )); + case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); + case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); + case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); } - else - return newSVuv (uv); + else + switch (dec->cur - start) + { + case 1: return newSViv ( start [0] - '0' ); + case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); + case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); + case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); + } + + { + UV uv; + int numtype = grok_number (start, dec->cur - start, &uv); + if (numtype & IS_NUMBER_IN_UV) + if (numtype & IS_NUMBER_NEG) + { + if (uv < (UV)IV_MIN) + return newSViv (-(IV)uv); + } + else + return newSVuv (uv); + } } return newSVnv (Atof (start)); @@ -704,7 +895,9 @@ { AV *av = newAV (); - WS; + DEC_INC_DEPTH; + decode_ws (dec); + if (*dec->cur == ']') ++dec->cur; else @@ -718,7 +911,7 @@ av_push (av, value); - WS; + decode_ws (dec); if (*dec->cur == ']') { @@ -732,10 +925,12 @@ ++dec->cur; } + DEC_DEC_DEPTH; return newRV_noinc ((SV *)av); fail: SvREFCNT_dec (av); + DEC_DEC_DEPTH; return 0; } @@ -744,7 +939,9 @@ { HV *hv = newHV (); - WS; + DEC_INC_DEPTH; + decode_ws (dec); + if (*dec->cur == '}') ++dec->cur; else @@ -752,13 +949,13 @@ { SV *key, *value; - WS; EXPECT_CH ('"'); + decode_ws (dec); EXPECT_CH ('"'); key = decode_str (dec); if (!key) goto fail; - WS; EXPECT_CH (':'); + decode_ws (dec); EXPECT_CH (':'); value = decode_sv (dec); if (!value) @@ -767,10 +964,10 @@ goto fail; } - //TODO: optimise hv_store_ent (hv, key, value, 0); + SvREFCNT_dec (key); - WS; + decode_ws (dec); if (*dec->cur == '}') { @@ -784,17 +981,19 @@ ++dec->cur; } + DEC_DEC_DEPTH; return newRV_noinc ((SV *)hv); fail: SvREFCNT_dec (hv); + DEC_DEC_DEPTH; return 0; } static SV * decode_sv (dec_t *dec) { - WS; + decode_ws (dec); switch (*dec->cur) { case '"': ++dec->cur; return decode_str (dec); @@ -840,7 +1039,7 @@ break; default: - ERR ("malformed json string"); + ERR ("malformed JSON string, neither array, object, number, string or atom"); break; } @@ -849,10 +1048,15 @@ } static SV * -decode_json (SV *string, UV flags) +decode_json (SV *string, U32 flags, UV *offset_return) { + dec_t dec; + UV offset; SV *sv; + SvGETMAGIC (string); + SvUPGRADE (string, SVt_PV); + if (flags & F_UTF8) sv_utf8_downgrade (string, 0); else @@ -860,27 +1064,53 @@ SvGROW (string, SvCUR (string) + 1); // should basically be a NOP - dec_t dec; - dec.flags = flags; - dec.cur = SvPVX (string); - dec.end = SvEND (string); - dec.err = 0; + dec.flags = flags; + dec.cur = SvPVX (string); + dec.end = SvEND (string); + dec.err = 0; + dec.depth = 0; + dec.maxdepth = DEC_DEPTH (dec.flags); + *dec.end = 0; // this should basically be a nop, too, but make sure it's there sv = decode_sv (&dec); + if (!(offset_return || !sv)) + { + // check for trailing garbage + decode_ws (&dec); + + if (*dec.cur) + { + dec.err = "garbage after JSON object"; + SvREFCNT_dec (sv); + sv = 0; + } + } + + if (offset_return || !sv) + { + offset = dec.flags & F_UTF8 + ? dec.cur - SvPVX (string) + : utf8_distance (dec.cur, SvPVX (string)); + + if (offset_return) + *offset_return = offset; + } + if (!sv) { - IV offset = utf8_distance (dec.cur, SvPVX (string)); SV *uni = sv_newmortal (); + // horrible hack to silence warning inside pv_uni_display - COP cop; - memset (&cop, 0, sizeof (cop)); + COP cop = *PL_curcop; cop.cop_warnings = pWARN_NONE; + ENTER; SAVEVPTR (PL_curcop); PL_curcop = &cop; - pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); - croak ("%s, at character offset %d (%s)", + LEAVE; + + croak ("%s, at character offset %d [\"%s\"]", dec.err, (int)offset, dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); @@ -889,11 +1119,14 @@ sv = sv_2mortal (sv); if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) - croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); + croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); return sv; } +///////////////////////////////////////////////////////////////////////////// +// XS interface functions + MODULE = JSON::XS PACKAGE = JSON::XS BOOT: @@ -901,14 +1134,13 @@ int i; memset (decode_hexdigit, 0xff, 256); - for (i = 10; i--; ) - decode_hexdigit ['0' + i] = i; - for (i = 7; i--; ) - { - decode_hexdigit ['a' + i] = 10 + i; - decode_hexdigit ['A' + i] = 10 + i; - } + for (i = 0; i < 256; ++i) + decode_hexdigit [i] = + i >= '0' && i <= '9' ? i - '0' + : i >= 'a' && i <= 'f' ? i - 'a' + 10 + : i >= 'A' && i <= 'F' ? i - 'A' + 10 + : -1; json_stash = gv_stashpv ("JSON::XS", 1); } @@ -924,12 +1156,12 @@ SV *ascii (SV *self, int enable = 1) ALIAS: ascii = F_ASCII + latin1 = F_LATIN1 utf8 = F_UTF8 indent = F_INDENT canonical = F_CANONICAL space_before = F_SPACE_BEFORE space_after = F_SPACE_AFTER - json_rpc = F_JSON_RPC pretty = F_PRETTY allow_nonref = F_ALLOW_NONREF shrink = F_SHRINK @@ -946,21 +1178,52 @@ OUTPUT: RETVAL +SV *max_depth (SV *self, UV max_depth = 0x80000000UL) + CODE: +{ + UV *uv = SvJSON (self); + UV log2 = 0; + + if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; + + while ((1UL << log2) < max_depth) + ++log2; + + *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); + + RETVAL = newSVsv (self); +} + OUTPUT: + RETVAL + void encode (SV *self, SV *scalar) PPCODE: XPUSHs (encode_json (scalar, *SvJSON (self))); void decode (SV *self, SV *jsonstr) PPCODE: - XPUSHs (decode_json (jsonstr, *SvJSON (self))); + XPUSHs (decode_json (jsonstr, *SvJSON (self), 0)); + +void decode_prefix (SV *self, SV *jsonstr) + PPCODE: +{ + UV offset; + EXTEND (SP, 2); + PUSHs (decode_json (jsonstr, *SvJSON (self), &offset)); + PUSHs (sv_2mortal (newSVuv (offset))); +} PROTOTYPES: ENABLE void to_json (SV *scalar) + ALIAS: + objToJson = 0 PPCODE: - XPUSHs (encode_json (scalar, F_UTF8)); + XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); void from_json (SV *jsonstr) + ALIAS: + jsonToObj = 0 PPCODE: - XPUSHs (decode_json (jsonstr, F_UTF8)); + XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));