JSON-XS/XS.xs

#include "EXTERN.h"
#include "perl.h"
#include "XSUB.h"

#include "assert.h"
#include "string.h"
#include "stdlib.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_PRETTY    F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
#define F_DEFAULT   0

#define INIT_SIZE   32 // initial scalar size to be allocated

#define SB do {
#define SE } while (0)

static HV *json_stash;

// 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;

// structure used for decoding JSON
typedef struct
{
  char *cur;
  char *end;
  const char *err;
  UV flags;
} dec_t;

static UV *
SvJSON (SV *sv)
{
  if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
    croak ("object is not of type JSON::XS");

  return &SvUVX (SvRV (sv));
}

/////////////////////////////////////////////////////////////////////////////

static void
need (enc_t *enc, STRLEN len)
{
  if (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);
    }
}

static void
encode_ch (enc_t *enc, char ch)
{
  need (enc, 1);
  *enc->cur++ = ch;
}

static void
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
{
  char *end = str + len;

  need (enc, len);

  while (str < end)
    {
      unsigned char ch = *(unsigned char *)str;

      if (ch == '"')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = '"';
          ++str;
        }
      else if (ch == '\\')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = '\\';
          ++str;
        }
      else if (ch >= 0x20 && ch < 0x80) // most common case
        {
          *enc->cur++ = ch;
          ++str;
        }
      else if (ch == '\015')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = 'r';
          ++str;
        }
      else if (ch == '\012')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = 'n';
          ++str;
        }
      else
        {
          STRLEN clen;
          UV uch;

          if (is_utf8)
            {
              uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
              if (clen == (STRLEN)-1)
                croak ("malformed UTF-8 character in string, cannot convert to JSON");
            }
          else
            {
              uch = ch;
              clen = 1;
            }

          if (uch < 0x80 || enc->flags & F_ASCII)
            {
              if (uch > 0xFFFFUL)
                {
                  need (enc, len += 11);
                  sprintf (enc->cur, "\\u%04x\\u%04x",
                           (uch - 0x10000) / 0x400 + 0xD800,
                           (uch - 0x10000) % 0x400 + 0xDC00);
                  enc->cur += 12;
                }
              else
                {
                  static char hexdigit [16] = "0123456789abcdef";
                  need (enc, len += 5);
                  *enc->cur++ = '\\';
                  *enc->cur++ = 'u';
                  *enc->cur++ = hexdigit [ uch >> 12      ];
                  *enc->cur++ = hexdigit [(uch >>  8) & 15];
                  *enc->cur++ = hexdigit [(uch >>  4) & 15];
                  *enc->cur++ = hexdigit [(uch >>  0) & 15];
                }

              str += clen;
            }
          else if (is_utf8)
            {
              need (enc, len += clen);
              do
                {
                  *enc->cur++ = *str++;
                }
              while (--clen);
            }
          else
            {
              need (enc, 10); // never more than 11 bytes needed
              enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
              ++str;
            }
        }

      --len;
    }
}

#define INDENT SB \
  if (enc->flags & F_INDENT)            \
    {                                   \
      int i_;                           \
      need (enc, enc->indent);          \
      for (i_ = enc->indent * 3; i_--; )\
        encode_ch (enc, ' ');           \
    }                                   \
  SE

#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

static void encode_sv (enc_t *enc, SV *sv);

static void
encode_av (enc_t *enc, AV *av)
{
  int i, len = av_len (av);

  encode_ch (enc, '['); NL;
  ++enc->indent;

  for (i = 0; i <= len; ++i)
    {
      INDENT;
      encode_sv (enc, *av_fetch (av, i, 0));

      if (i < len)
        COMMA;
    }

  NL;

  --enc->indent;
  INDENT; encode_ch (enc, ']');
}

static void
encode_he (enc_t *enc, HE *he)
{
  encode_ch (enc, '"');

  if (HeKLEN (he) == HEf_SVKEY)
    {
      SV *sv = HeSVKEY (he);
      STRLEN len;
      char *str;
      
      SvGETMAGIC (sv);
      str = SvPV (sv, len);

      encode_str (enc, str, len, SvUTF8 (sv));
    }
  else
    encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));

  encode_ch (enc, '"');

  if (enc->flags & F_SPACE_BEFORE) SPACE;
  encode_ch (enc, ':');
  if (enc->flags & F_SPACE_AFTER ) SPACE;
  encode_sv (enc, HeVAL (he));
}

// compare hash entries, used when all keys are bytestrings
static int
he_cmp_fast (const void *a_, const void *b_)
{
  int cmp;

  HE *a = *(HE **)a_;
  HE *b = *(HE **)b_;

  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;

  return cmp;
}

// compare hash entries, used when some keys are sv's or utf-x
static int
he_cmp_slow (const void *a, const void *b)
{
  return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));
}

static void
encode_hv (enc_t *enc, HV *hv)
{
  int count, i;

  encode_ch (enc, '{'); NL; ++enc->indent;

  if ((count = hv_iterinit (hv)))
    {
      // for canonical output we have to sort by keys first
      // actually, this is mostly due to the stupid so-called
      // security workaround added somewhere in 5.8.x.
      // that randomises hash orderings
      if (enc->flags & F_CANONICAL)
        {
          HE *he, *hes [count];
          int fast = 1;

          i = 0;
          while ((he = hv_iternext (hv)))
            {
              hes [i++] = he;
              if (HeKLEN (he) < 0 || HeKUTF8 (he))
                fast = 0;
            }

          assert (i == count);

          if (fast)
            qsort (hes, count, sizeof (HE *), he_cmp_fast);
          else
            {
              // hack to disable "use bytes"
              COP *oldcop = PL_curcop, cop;
              cop.op_private = 0;
              PL_curcop = &cop;

              SAVETMPS;
              qsort (hes, count, sizeof (HE *), he_cmp_slow);
              FREETMPS;

              PL_curcop = oldcop;
            }

          for (i = 0; i < count; ++i)
            {
              INDENT;
              encode_he (enc, hes [i]);

              if (i < count - 1)
                COMMA;
            }

          NL;
        }
      else
        {
          SV *sv;
          HE *he = hv_iternext (hv);

          for (;;)
            {
              INDENT;
              encode_he (enc, he);

              if (!(he = hv_iternext (hv)))
                break;

              COMMA;
            }

          NL;
        }
    }

  --enc->indent; INDENT; encode_ch (enc, '}');
}

static void
encode_sv (enc_t *enc, SV *sv)
{
  SvGETMAGIC (sv);

  if (SvPOKp (sv))
    {
      STRLEN len;
      char *str = SvPV (sv, len);
      encode_ch (enc, '"');
      encode_str (enc, str, len, SvUTF8 (sv));
      encode_ch (enc, '"');
    }
  else if (SvNOKp (sv))
    {
      need (enc, NV_DIG + 32);
      Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
      enc->cur += strlen (enc->cur);
    }
  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))
        {
          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");
        }
    }
  else if (!SvOK (sv))
    encode_str (enc, "null", 4, 0);
  else
    croak ("encountered perl type that JSON cannot handle");
}

static SV *
encode_json (SV *scalar, UV flags)
{
  if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
    croak ("hash- or arraref required (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;

  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));
  return enc.sv;
}

/////////////////////////////////////////////////////////////////////////////

#define WS \
  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

static SV *decode_sv (dec_t *dec);

static signed char decode_hexdigit[256];

static UV
decode_4hex (dec_t *dec)
{
  signed char d1, d2, d3, d4;

  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");

  dec->cur += 4;

  return ((UV)d1) << 12
       | ((UV)d2) <<  8
       | ((UV)d3) <<  4
       | ((UV)d4);

fail:
  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);
  int utf8 = 0;
  char *cur = SvPVX (sv);
  char *end = SvEND (sv);

  for (;;)
    {
      unsigned char ch = *(unsigned char *)dec->cur;

      if (ch == '"')
        break;
      else if (ch == '\\')
        {
          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;

              case 'u':
                {
                  UV lo, hi;
                  ++dec->cur;

                  hi = decode_4hex (dec);
                  if (hi == (UV)-1)
                    goto fail;

                  // possibly a surrogate pair
                  if (hi >= 0xd800 && 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 >= 0xdc00 && hi < 0xe000)
                    ERR ("missing high surrogate character in surrogate pair");

                  if (hi >= 0x80)
                    {
                      utf8 = 1;

                      APPEND_GROW (4); // at most 4 bytes for 21 bits
                      cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
                    }
                  else
                    APPEND_CH (hi);
                }
                break;

              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
            {
              *cur++ = *dec->cur++;
            }
          while (--clen);

          utf8 = 1;
        }
      else if (dec->cur == dec->end)
        ERR ("unexpected end of string while parsing json string");
      else
        ERR ("invalid character encountered");
    }

  ++dec->cur;

  SvCUR_set (sv, cur - SvPVX (sv));

  SvPOK_only (sv);
  *SvEND (sv) = 0;

  if (utf8)
    SvUTF8_on (sv);

  return sv;

fail:
  SvREFCNT_dec (sv);
  return 0;
}

static SV *
decode_num (dec_t *dec)
{
  int is_nv = 0;
  char *start = dec->cur;

  // [minus]
  if (*dec->cur == '-')
    ++dec->cur;

  if (*dec->cur == '0')
    {
      ++dec->cur;
      if (*dec->cur >= '0' && *dec->cur <= '9')
         ERR ("malformed number (leading zero must not be followed by another digit)");
    }
  else if (*dec->cur < '0' || *dec->cur > '9')
    ERR ("malformed number (no digits after initial minus)");
  else
    do
      {
        ++dec->cur;
      }
    while (*dec->cur >= '0' && *dec->cur <= '9');

  // [frac]
  if (*dec->cur == '.')
    {
      ++dec->cur;

      if (*dec->cur < '0' || *dec->cur > '9')
        ERR ("malformed number (no digits after decimal point)");

      do
        {
          ++dec->cur;
        }
      while (*dec->cur >= '0' && *dec->cur <= '9');

      is_nv = 1;
    }

  // [exp]
  if (*dec->cur == 'e' || *dec->cur == 'E')
    {
      ++dec->cur;

      if (*dec->cur == '-' || *dec->cur == '+')
        ++dec->cur;

      if (*dec->cur < '0' || *dec->cur > '9')
        ERR ("malformed number (no digits after exp sign)");

      do
        {
          ++dec->cur;
        }
      while (*dec->cur >= '0' && *dec->cur <= '9');

      is_nv = 1;
    }

  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)
          {
            if (uv < (UV)IV_MIN)
              return newSViv (-(IV)uv);
          }
        else
          return newSVuv (uv);
    }

  return newSVnv (Atof (start));

fail:
  return 0;
}

static SV *
decode_av (dec_t *dec)
{
  AV *av = newAV ();

  WS;
  if (*dec->cur == ']')
    ++dec->cur;
  else
    for (;;)
      {
        SV *value;

        value = decode_sv (dec);
        if (!value)
          goto fail;

        av_push (av, value);

        WS;

        if (*dec->cur == ']')
          {
            ++dec->cur;
            break;
          }
        
        if (*dec->cur != ',')
          ERR (", or ] expected while parsing array");

        ++dec->cur;
      }

  return newRV_noinc ((SV *)av);

fail:
  SvREFCNT_dec (av);
  return 0;
}

static SV *
decode_hv (dec_t *dec)
{
  HV *hv = newHV ();

  WS;
  if (*dec->cur == '}')
    ++dec->cur;
  else
    for (;;)
      {
        SV *key, *value;

        WS; EXPECT_CH ('"');

        key = decode_str (dec);
        if (!key)
          goto fail;

        WS; EXPECT_CH (':');

        value = decode_sv (dec);
        if (!value)
          {
            SvREFCNT_dec (key);
            goto fail;
          }

        //TODO: optimise
        hv_store_ent (hv, key, value, 0);

        WS;

        if (*dec->cur == '}')
          {
            ++dec->cur;
            break;
          }

        if (*dec->cur != ',')
          ERR (", or } expected while parsing object/hash");

        ++dec->cur;
      }

  return newRV_noinc ((SV *)hv);

fail:
  SvREFCNT_dec (hv);
  return 0;
}

static SV *
decode_sv (dec_t *dec)
{
  WS;
  switch (*dec->cur)
    {
      case '"': ++dec->cur; return decode_str (dec); 
      case '[': ++dec->cur; return decode_av (dec); 
      case '{': ++dec->cur; return decode_hv (dec);

      case '-':
      case '0': case '1': case '2': case '3': case '4':
      case '5': case '6': case '7': case '8': case '9':
        return decode_num (dec);

      case 't':
        if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
          {
            dec->cur += 4;
            return newSViv (1);
          }
        else
          ERR ("'true' expected");

        break;

      case 'f':
        if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
          {
            dec->cur += 5;
            return newSViv (0);
          }
        else
          ERR ("'false' expected");

        break;

      case 'n':
        if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
          {
            dec->cur += 4;
            return newSVsv (&PL_sv_undef);
          }
        else
          ERR ("'null' expected");

        break;

      default:
        ERR ("malformed json string");
        break;
    }

fail:
  return 0;
}

static SV *
decode_json (SV *string, UV flags)
{
  SV *sv;

  if (flags & F_UTF8)
    sv_utf8_downgrade (string, 0);
  else
    sv_utf8_upgrade (string);

  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;

  sv = decode_sv (&dec);

  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_warnings = pWARN_NONE;
      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)",
             dec.err,
             (int)offset,
             dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
    }

  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)");

  return sv;
}

MODULE = JSON::XS               PACKAGE = JSON::XS

BOOT:
{
        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;
          }

        json_stash = gv_stashpv ("JSON::XS", 1);
}

PROTOTYPES: DISABLE

SV *new (char *dummy)
        CODE:
        RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
        OUTPUT:
        RETVAL

SV *ascii (SV *self, int enable)
        ALIAS:
        ascii        = F_ASCII
        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
        CODE:
{
        UV *uv = SvJSON (self);
        if (enable)
          *uv |=  ix;
        else
          *uv &= ~ix;

        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)));

PROTOTYPES: ENABLE

void to_json (SV *scalar)
        PPCODE:
        XPUSHs (encode_json (scalar, F_UTF8));

void from_json (SV *jsonstr)
        PPCODE:
        XPUSHs (decode_json (jsonstr, F_UTF8));

Revision:	1.5
Committed:	Thu Mar 22 23:24:18 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
CVS Tags:	rel-0_2
Changes since 1.4:	+111 -68 lines
Log Message:	* empty log message *
#	User	Rev	Content
1	root	1.1	#include "EXTERN.h"
2			#include "perl.h"
3			#include "XSUB.h"
4
5			#include "assert.h"
6			#include "string.h"
7			#include "stdlib.h"
8
9			#define F_ASCII 0x00000001
10			#define F_UTF8 0x00000002
11			#define F_INDENT 0x00000004
12			#define F_CANONICAL 0x00000008
13			#define F_SPACE_BEFORE 0x00000010
14			#define F_SPACE_AFTER 0x00000020
15			#define F_JSON_RPC 0x00000040
16	root	1.3	#define F_ALLOW_NONREF 0x00000080
17	root	1.1
18	root	1.2	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
19	root	1.1	#define F_DEFAULT 0
20
21			#define INIT_SIZE 32 // initial scalar size to be allocated
22
23			#define SB do {
24			#define SE } while (0)
25
26			static HV *json_stash;
27
28			// structure used for encoding JSON
29			typedef struct
30			{
31			char *cur;
32			STRLEN len; // SvLEN (sv)
33			char *end; // SvEND (sv)
34			SV *sv;
35			UV flags;
36			int max_recurse;
37			int indent;
38			} enc_t;
39
40			// structure used for decoding JSON
41			typedef struct
42			{
43			char *cur;
44			char *end;
45	root	1.4	const char *err;
46	root	1.1	UV flags;
47			} dec_t;
48
49			static UV *
50			SvJSON (SV *sv)
51			{
52			if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53			croak ("object is not of type JSON::XS");
54
55			return &SvUVX (SvRV (sv));
56			}
57
58			/////////////////////////////////////////////////////////////////////////////
59
60			static void
61			need (enc_t *enc, STRLEN len)
62			{
63			if (enc->cur + len >= enc->end)
64			{
65			STRLEN cur = enc->cur - SvPVX (enc->sv);
66			SvGROW (enc->sv, cur + len + 1);
67			enc->cur = SvPVX (enc->sv) + cur;
68	root	1.4	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);
69	root	1.1	}
70			}
71
72			static void
73			encode_ch (enc_t *enc, char ch)
74			{
75			need (enc, 1);
76			*enc->cur++ = ch;
77			}
78
79			static void
80			encode_str (enc_t enc, char str, STRLEN len, int is_utf8)
81			{
82			char *end = str + len;
83
84	root	1.4	need (enc, len);
85
86	root	1.1	while (str < end)
87			{
88			unsigned char ch = (unsigned char )str;
89	root	1.4
90			if (ch == '"')
91			{
92			need (enc, len += 1);
93			*enc->cur++ = '\\';
94			*enc->cur++ = '"';
95			++str;
96			}
97			else if (ch == '\\')
98			{
99			need (enc, len += 1);
100			*enc->cur++ = '\\';
101			*enc->cur++ = '\\';
102			++str;
103			}
104			else if (ch >= 0x20 && ch < 0x80) // most common case
105	root	1.1	{
106			*enc->cur++ = ch;
107	root	1.4	++str;
108			}
109			else if (ch == '\015')
110			{
111			need (enc, len += 1);
112			*enc->cur++ = '\\';
113			*enc->cur++ = 'r';
114			++str;
115			}
116			else if (ch == '\012')
117			{
118			need (enc, len += 1);
119			*enc->cur++ = '\\';
120			*enc->cur++ = 'n';
121			++str;
122	root	1.1	}
123			else
124			{
125			STRLEN clen;
126			UV uch;
127
128			if (is_utf8)
129			{
130			uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
131	root	1.5	if (clen == (STRLEN)-1)
132	root	1.1	croak ("malformed UTF-8 character in string, cannot convert to JSON");
133			}
134			else
135			{
136			uch = ch;
137			clen = 1;
138			}
139
140	root	1.4	if (uch < 0x80 \|\| enc->flags & F_ASCII)
141	root	1.1	{
142			if (uch > 0xFFFFUL)
143			{
144	root	1.4	need (enc, len += 11);
145	root	1.1	sprintf (enc->cur, "\\u%04x\\u%04x",
146			(uch - 0x10000) / 0x400 + 0xD800,
147			(uch - 0x10000) % 0x400 + 0xDC00);
148			enc->cur += 12;
149			}
150			else
151			{
152	root	1.4	static char hexdigit [16] = "0123456789abcdef";
153			need (enc, len += 5);
154			*enc->cur++ = '\\';
155			*enc->cur++ = 'u';
156			*enc->cur++ = hexdigit [ uch >> 12 ];
157			*enc->cur++ = hexdigit [(uch >> 8) & 15];
158			*enc->cur++ = hexdigit [(uch >> 4) & 15];
159			*enc->cur++ = hexdigit [(uch >> 0) & 15];
160	root	1.1	}
161	root	1.4
162			str += clen;
163	root	1.1	}
164			else if (is_utf8)
165			{
166	root	1.4	need (enc, len += clen);
167	root	1.5	do
168			{
169			enc->cur++ = str++;
170			}
171			while (--clen);
172	root	1.1	}
173			else
174	root	1.4	{
175			need (enc, 10); // never more than 11 bytes needed
176			enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
177			++str;
178			}
179	root	1.1	}
180
181			--len;
182			}
183			}
184
185			#define INDENT SB \
186			if (enc->flags & F_INDENT) \
187			{ \
188			int i_; \
189			need (enc, enc->indent); \
190			for (i_ = enc->indent * 3; i_--; )\
191			encode_ch (enc, ' '); \
192			} \
193			SE
194
195	root	1.2	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE
196	root	1.1	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
197			#define COMMA SB \
198			encode_ch (enc, ','); \
199			if (enc->flags & F_INDENT) \
200			NL; \
201			else if (enc->flags & F_SPACE_AFTER) \
202			SPACE; \
203			SE
204
205			static void encode_sv (enc_t enc, SV sv);
206
207			static void
208			encode_av (enc_t enc, AV av)
209			{
210			int i, len = av_len (av);
211
212			encode_ch (enc, '['); NL;
213			++enc->indent;
214
215			for (i = 0; i <= len; ++i)
216			{
217			INDENT;
218			encode_sv (enc, *av_fetch (av, i, 0));
219
220			if (i < len)
221			COMMA;
222			}
223
224			NL;
225
226			--enc->indent;
227			INDENT; encode_ch (enc, ']');
228			}
229
230			static void
231			encode_he (enc_t enc, HE he)
232			{
233			encode_ch (enc, '"');
234
235			if (HeKLEN (he) == HEf_SVKEY)
236			{
237			SV *sv = HeSVKEY (he);
238			STRLEN len;
239	root	1.4	char *str;
240
241			SvGETMAGIC (sv);
242			str = SvPV (sv, len);
243	root	1.1
244			encode_str (enc, str, len, SvUTF8 (sv));
245			}
246			else
247			encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
248
249			encode_ch (enc, '"');
250
251			if (enc->flags & F_SPACE_BEFORE) SPACE;
252			encode_ch (enc, ':');
253			if (enc->flags & F_SPACE_AFTER ) SPACE;
254			encode_sv (enc, HeVAL (he));
255			}
256
257			// compare hash entries, used when all keys are bytestrings
258			static int
259			he_cmp_fast (const void a_, const void b_)
260			{
261			int cmp;
262
263			HE a = (HE **)a_;
264			HE b = (HE **)b_;
265
266			STRLEN la = HeKLEN (a);
267			STRLEN lb = HeKLEN (b);
268
269			if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
270			cmp = la < lb ? -1 : la == lb ? 0 : 1;
271
272			return cmp;
273			}
274
275			// compare hash entries, used when some keys are sv's or utf-x
276			static int
277			he_cmp_slow (const void a, const void b)
278			{
279			return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));
280			}
281
282			static void
283			encode_hv (enc_t enc, HV hv)
284			{
285			int count, i;
286
287			encode_ch (enc, '{'); NL; ++enc->indent;
288
289			if ((count = hv_iterinit (hv)))
290			{
291			// for canonical output we have to sort by keys first
292			// actually, this is mostly due to the stupid so-called
293			// security workaround added somewhere in 5.8.x.
294			// that randomises hash orderings
295			if (enc->flags & F_CANONICAL)
296			{
297			HE he, hes [count];
298			int fast = 1;
299
300			i = 0;
301			while ((he = hv_iternext (hv)))
302			{
303			hes [i++] = he;
304			if (HeKLEN (he) < 0 \|\| HeKUTF8 (he))
305			fast = 0;
306			}
307
308			assert (i == count);
309
310			if (fast)
311			qsort (hes, count, sizeof (HE *), he_cmp_fast);
312			else
313			{
314			// hack to disable "use bytes"
315			COP *oldcop = PL_curcop, cop;
316			cop.op_private = 0;
317			PL_curcop = &cop;
318
319			SAVETMPS;
320			qsort (hes, count, sizeof (HE *), he_cmp_slow);
321			FREETMPS;
322
323			PL_curcop = oldcop;
324			}
325
326			for (i = 0; i < count; ++i)
327			{
328			INDENT;
329			encode_he (enc, hes [i]);
330
331			if (i < count - 1)
332			COMMA;
333			}
334
335			NL;
336			}
337			else
338			{
339			SV *sv;
340			HE *he = hv_iternext (hv);
341
342			for (;;)
343			{
344			INDENT;
345			encode_he (enc, he);
346
347			if (!(he = hv_iternext (hv)))
348			break;
349
350			COMMA;
351			}
352
353			NL;
354			}
355			}
356
357			--enc->indent; INDENT; encode_ch (enc, '}');
358			}
359
360			static void
361			encode_sv (enc_t enc, SV sv)
362			{
363	root	1.4	SvGETMAGIC (sv);
364
365	root	1.1	if (SvPOKp (sv))
366			{
367			STRLEN len;
368			char *str = SvPV (sv, len);
369			encode_ch (enc, '"');
370			encode_str (enc, str, len, SvUTF8 (sv));
371			encode_ch (enc, '"');
372			}
373			else if (SvNOKp (sv))
374			{
375			need (enc, NV_DIG + 32);
376			Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
377			enc->cur += strlen (enc->cur);
378			}
379			else if (SvIOKp (sv))
380			{
381			need (enc, 64);
382			enc->cur +=
383			SvIsUV(sv)
384			? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
385			: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
386			}
387			else if (SvROK (sv))
388			{
389			if (!--enc->max_recurse)
390			croak ("data structure too deep (hit recursion limit)");
391
392			sv = SvRV (sv);
393
394			switch (SvTYPE (sv))
395			{
396			case SVt_PVAV: encode_av (enc, (AV *)sv); break;
397			case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
398
399			default:
400			croak ("JSON can only represent references to arrays or hashes");
401			}
402			}
403			else if (!SvOK (sv))
404			encode_str (enc, "null", 4, 0);
405			else
406			croak ("encountered perl type that JSON cannot handle");
407			}
408
409			static SV *
410			encode_json (SV *scalar, UV flags)
411			{
412	root	1.3	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
413			croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
414
415	root	1.1	enc_t enc;
416			enc.flags = flags;
417			enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
418			enc.cur = SvPVX (enc.sv);
419			enc.end = SvEND (enc.sv);
420			enc.max_recurse = 0;
421			enc.indent = 0;
422
423			SvPOK_only (enc.sv);
424			encode_sv (&enc, scalar);
425
426			if (!(flags & (F_ASCII \| F_UTF8)))
427			SvUTF8_on (enc.sv);
428
429			SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
430			return enc.sv;
431			}
432
433			/////////////////////////////////////////////////////////////////////////////
434
435			#define WS \
436			for (;;) \
437			{ \
438			char ch = *dec->cur; \
439			if (ch > 0x20 \
440			\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
441			break; \
442			++dec->cur; \
443			}
444
445			#define ERR(reason) SB dec->err = reason; goto fail; SE
446			#define EXPECT_CH(ch) SB \
447			if (*dec->cur != ch) \
448			ERR (# ch " expected"); \
449			++dec->cur; \
450			SE
451
452			static SV decode_sv (dec_t dec);
453
454			static signed char decode_hexdigit[256];
455
456			static UV
457			decode_4hex (dec_t *dec)
458			{
459			signed char d1, d2, d3, d4;
460
461			d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
462			if (d1 < 0) ERR ("four hexadecimal digits expected");
463			d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
464			if (d2 < 0) ERR ("four hexadecimal digits expected");
465			d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
466			if (d3 < 0) ERR ("four hexadecimal digits expected");
467			d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
468			if (d4 < 0) ERR ("four hexadecimal digits expected");
469
470			dec->cur += 4;
471
472			return ((UV)d1) << 12
473			\| ((UV)d2) << 8
474			\| ((UV)d3) << 4
475			\| ((UV)d4);
476
477			fail:
478			return (UV)-1;
479			}
480
481	root	1.4	#define APPEND_GROW(n) SB \
482			if (cur + (n) >= end) \
483			{ \
484			STRLEN ofs = cur - SvPVX (sv); \
485			SvGROW (sv, ofs + (n) + 1); \
486			cur = SvPVX (sv) + ofs; \
487			end = SvEND (sv); \
488			} \
489			SE
490
491			#define APPEND_CH(ch) SB \
492			APPEND_GROW (1); \
493			*cur++ = (ch); \
494			SE
495
496	root	1.1	static SV *
497			decode_str (dec_t *dec)
498			{
499			SV *sv = NEWSV (0,2);
500			int utf8 = 0;
501	root	1.4	char *cur = SvPVX (sv);
502			char *end = SvEND (sv);
503	root	1.1
504			for (;;)
505			{
506			unsigned char ch = (unsigned char )dec->cur;
507
508			if (ch == '"')
509			break;
510			else if (ch == '\\')
511			{
512			switch (*++dec->cur)
513			{
514			case '\\':
515			case '/':
516			case '"': APPEND_CH (*dec->cur++); break;
517
518			case 'b': APPEND_CH ('\010'); ++dec->cur; break;
519			case 't': APPEND_CH ('\011'); ++dec->cur; break;
520			case 'n': APPEND_CH ('\012'); ++dec->cur; break;
521			case 'f': APPEND_CH ('\014'); ++dec->cur; break;
522			case 'r': APPEND_CH ('\015'); ++dec->cur; break;
523
524			case 'u':
525			{
526			UV lo, hi;
527			++dec->cur;
528
529			hi = decode_4hex (dec);
530			if (hi == (UV)-1)
531			goto fail;
532
533			// possibly a surrogate pair
534			if (hi >= 0xd800 && hi < 0xdc00)
535			{
536			if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
537	root	1.5	ERR ("missing low surrogate character in surrogate pair");
538	root	1.1
539			dec->cur += 2;
540
541			lo = decode_4hex (dec);
542			if (lo == (UV)-1)
543			goto fail;
544
545			if (lo < 0xdc00 \|\| lo >= 0xe000)
546			ERR ("surrogate pair expected");
547
548			hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
549			}
550	root	1.5	else if (hi >= 0xdc00 && hi < 0xe000)
551			ERR ("missing high surrogate character in surrogate pair");
552	root	1.1
553			if (hi >= 0x80)
554			{
555			utf8 = 1;
556
557	root	1.4	APPEND_GROW (4); // at most 4 bytes for 21 bits
558			cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
559	root	1.1	}
560			else
561			APPEND_CH (hi);
562			}
563			break;
564	root	1.5
565			default:
566			--dec->cur;
567			ERR ("illegal backslash escape sequence in string");
568	root	1.1	}
569			}
570			else if (ch >= 0x20 && ch <= 0x7f)
571			APPEND_CH (*dec->cur++);
572			else if (ch >= 0x80)
573			{
574			STRLEN clen;
575			UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
576	root	1.5	if (clen == (STRLEN)-1)
577	root	1.1	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
578
579	root	1.4	APPEND_GROW (clen);
580	root	1.5	do
581			{
582			cur++ = dec->cur++;
583			}
584			while (--clen);
585
586			utf8 = 1;
587	root	1.1	}
588	root	1.5	else if (dec->cur == dec->end)
589			ERR ("unexpected end of string while parsing json string");
590	root	1.1	else
591			ERR ("invalid character encountered");
592			}
593
594			++dec->cur;
595
596	root	1.4	SvCUR_set (sv, cur - SvPVX (sv));
597
598	root	1.1	SvPOK_only (sv);
599			*SvEND (sv) = 0;
600
601			if (utf8)
602			SvUTF8_on (sv);
603
604			return sv;
605
606			fail:
607			SvREFCNT_dec (sv);
608			return 0;
609			}
610
611			static SV *
612			decode_num (dec_t *dec)
613			{
614			int is_nv = 0;
615			char *start = dec->cur;
616
617			// [minus]
618			if (*dec->cur == '-')
619			++dec->cur;
620
621			if (*dec->cur == '0')
622			{
623			++dec->cur;
624			if (dec->cur >= '0' && dec->cur <= '9')
625			ERR ("malformed number (leading zero must not be followed by another digit)");
626			}
627	root	1.5	else if (dec->cur < '0' \|\| dec->cur > '9')
628			ERR ("malformed number (no digits after initial minus)");
629			else
630			do
631			{
632			++dec->cur;
633			}
634			while (dec->cur >= '0' && dec->cur <= '9');
635	root	1.1
636			// [frac]
637			if (*dec->cur == '.')
638			{
639	root	1.5	++dec->cur;
640
641			if (dec->cur < '0' \|\| dec->cur > '9')
642			ERR ("malformed number (no digits after decimal point)");
643	root	1.1
644			do
645			{
646			++dec->cur;
647			}
648			while (dec->cur >= '0' && dec->cur <= '9');
649	root	1.5
650			is_nv = 1;
651	root	1.1	}
652
653			// [exp]
654			if (dec->cur == 'e' \|\| dec->cur == 'E')
655			{
656	root	1.5	++dec->cur;
657	root	1.1
658			if (dec->cur == '-' \|\| dec->cur == '+')
659			++dec->cur;
660
661	root	1.5	if (dec->cur < '0' \|\| dec->cur > '9')
662			ERR ("malformed number (no digits after exp sign)");
663
664			do
665			{
666			++dec->cur;
667			}
668			while (dec->cur >= '0' && dec->cur <= '9');
669
670			is_nv = 1;
671	root	1.1	}
672
673			if (!is_nv)
674			{
675			UV uv;
676			int numtype = grok_number (start, dec->cur - start, &uv);
677			if (numtype & IS_NUMBER_IN_UV)
678			if (numtype & IS_NUMBER_NEG)
679			{
680			if (uv < (UV)IV_MIN)
681			return newSViv (-(IV)uv);
682			}
683			else
684			return newSVuv (uv);
685			}
686
687			return newSVnv (Atof (start));
688
689			fail:
690			return 0;
691			}
692
693			static SV *
694			decode_av (dec_t *dec)
695			{
696			AV *av = newAV ();
697
698	root	1.5	WS;
699			if (*dec->cur == ']')
700			++dec->cur;
701			else
702			for (;;)
703			{
704			SV *value;
705	root	1.1
706	root	1.5	value = decode_sv (dec);
707			if (!value)
708			goto fail;
709	root	1.1
710	root	1.5	av_push (av, value);
711	root	1.1
712	root	1.5	WS;
713	root	1.1
714	root	1.5	if (*dec->cur == ']')
715			{
716			++dec->cur;
717			break;
718			}
719
720			if (*dec->cur != ',')
721			ERR (", or ] expected while parsing array");
722	root	1.1
723	root	1.5	++dec->cur;
724			}
725	root	1.1
726			return newRV_noinc ((SV *)av);
727
728			fail:
729			SvREFCNT_dec (av);
730			return 0;
731			}
732
733			static SV *
734			decode_hv (dec_t *dec)
735			{
736			HV *hv = newHV ();
737
738	root	1.5	WS;
739			if (*dec->cur == '}')
740			++dec->cur;
741			else
742			for (;;)
743			{
744			SV key, value;
745	root	1.1
746	root	1.5	WS; EXPECT_CH ('"');
747	root	1.1
748	root	1.5	key = decode_str (dec);
749			if (!key)
750			goto fail;
751	root	1.1
752	root	1.5	WS; EXPECT_CH (':');
753	root	1.1
754	root	1.5	value = decode_sv (dec);
755			if (!value)
756			{
757			SvREFCNT_dec (key);
758			goto fail;
759			}
760	root	1.1
761	root	1.5	//TODO: optimise
762			hv_store_ent (hv, key, value, 0);
763	root	1.1
764	root	1.5	WS;
765	root	1.1
766	root	1.5	if (*dec->cur == '}')
767			{
768			++dec->cur;
769			break;
770			}
771	root	1.1
772	root	1.5	if (*dec->cur != ',')
773			ERR (", or } expected while parsing object/hash");
774	root	1.1
775	root	1.5	++dec->cur;
776			}
777	root	1.1
778			return newRV_noinc ((SV *)hv);
779
780			fail:
781			SvREFCNT_dec (hv);
782			return 0;
783			}
784
785			static SV *
786			decode_sv (dec_t *dec)
787			{
788			WS;
789			switch (*dec->cur)
790			{
791			case '"': ++dec->cur; return decode_str (dec);
792			case '[': ++dec->cur; return decode_av (dec);
793			case '{': ++dec->cur; return decode_hv (dec);
794
795			case '-':
796			case '0': case '1': case '2': case '3': case '4':
797			case '5': case '6': case '7': case '8': case '9':
798			return decode_num (dec);
799
800			case 't':
801			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
802			{
803			dec->cur += 4;
804			return newSViv (1);
805			}
806			else
807			ERR ("'true' expected");
808
809			break;
810
811			case 'f':
812			if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
813			{
814			dec->cur += 5;
815			return newSViv (0);
816			}
817			else
818			ERR ("'false' expected");
819
820			break;
821
822			case 'n':
823			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
824			{
825			dec->cur += 4;
826	root	1.5	return newSVsv (&PL_sv_undef);
827	root	1.1	}
828			else
829			ERR ("'null' expected");
830
831			break;
832
833			default:
834			ERR ("malformed json string");
835			break;
836			}
837
838			fail:
839			return 0;
840			}
841
842			static SV *
843			decode_json (SV *string, UV flags)
844			{
845			SV *sv;
846
847	root	1.5	if (flags & F_UTF8)
848			sv_utf8_downgrade (string, 0);
849			else
850	root	1.1	sv_utf8_upgrade (string);
851
852			SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
853
854			dec_t dec;
855			dec.flags = flags;
856			dec.cur = SvPVX (string);
857			dec.end = SvEND (string);
858			dec.err = 0;
859
860			sv = decode_sv (&dec);
861
862			if (!sv)
863			{
864			IV offset = utf8_distance (dec.cur, SvPVX (string));
865			SV *uni = sv_newmortal ();
866	root	1.5	// horrible hack to silence warning inside pv_uni_display
867			COP cop;
868			memset (&cop, 0, sizeof (cop));
869			cop.cop_warnings = pWARN_NONE;
870			SAVEVPTR (PL_curcop);
871			PL_curcop = &cop;
872	root	1.1
873			pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
874	root	1.5	croak ("%s, at character offset %d (%s)",
875	root	1.1	dec.err,
876			(int)offset,
877			dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
878			}
879
880	root	1.3	sv = sv_2mortal (sv);
881
882			if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
883			croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");
884
885			return sv;
886	root	1.1	}
887
888			MODULE = JSON::XS PACKAGE = JSON::XS
889
890			BOOT:
891			{
892			int i;
893
894			memset (decode_hexdigit, 0xff, 256);
895			for (i = 10; i--; )
896			decode_hexdigit ['0' + i] = i;
897
898	root	1.4	for (i = 7; i--; )
899	root	1.1	{
900			decode_hexdigit ['a' + i] = 10 + i;
901			decode_hexdigit ['A' + i] = 10 + i;
902			}
903
904			json_stash = gv_stashpv ("JSON::XS", 1);
905			}
906
907	root	1.4	PROTOTYPES: DISABLE
908
909	root	1.1	SV new (char dummy)
910			CODE:
911			RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
912			OUTPUT:
913			RETVAL
914
915			SV ascii (SV self, int enable)
916			ALIAS:
917			ascii = F_ASCII
918			utf8 = F_UTF8
919			indent = F_INDENT
920			canonical = F_CANONICAL
921			space_before = F_SPACE_BEFORE
922			space_after = F_SPACE_AFTER
923			json_rpc = F_JSON_RPC
924	root	1.2	pretty = F_PRETTY
925	root	1.3	allow_nonref = F_ALLOW_NONREF
926	root	1.1	CODE:
927			{
928			UV *uv = SvJSON (self);
929			if (enable)
930			*uv \|= ix;
931			else
932			*uv &= ~ix;
933
934			RETVAL = newSVsv (self);
935			}
936			OUTPUT:
937			RETVAL
938
939			void encode (SV self, SV scalar)
940			PPCODE:
941			XPUSHs (encode_json (scalar, *SvJSON (self)));
942
943	root	1.2	void decode (SV self, SV jsonstr)
944	root	1.1	PPCODE:
945	root	1.2	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
946
947	root	1.4	PROTOTYPES: ENABLE
948
949	root	1.2	void to_json (SV *scalar)
950			PPCODE:
951			XPUSHs (encode_json (scalar, F_UTF8));
952
953			void from_json (SV *jsonstr)
954			PPCODE:
955			XPUSHs (decode_json (jsonstr, F_UTF8));
956	root	1.1