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_SHRINK       0x00000100

#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 >= 0x20 && ch < 0x80) // most common case
        {
          if (ch == '"') // but with slow exceptions
            {
              need (enc, len += 1);
              *enc->cur++ = '\\';
              *enc->cur++ = '"';
            }
          else if (ch == '\\')
            {
              need (enc, len += 1);
              *enc->cur++ = '\\';
              *enc->cur++ = '\\';
            }
          else
            *enc->cur++ = ch;

          ++str;
        }
      else
        {
          switch (ch)
            {
              case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
              case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
              case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
              case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
              case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;

              default:
                {
                  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));

#ifdef SvPV_shrink_to_cur
  if (enc.flags & F_SHRINK)
    SvPV_shrink_to_cur (enc.sv);
#endif
  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);

#ifdef SvPV_shrink_to_cur
  if (dec->flags & F_SHRINK)
    SvPV_shrink_to_cur (sv);
#endif

  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 = 1)
        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
        shrink       = F_SHRINK
        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:	1.6
Committed:	Fri Mar 23 15:10:55 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
Changes since 1.5:	+89 -79 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	1.6	#define F_SHRINK 0x00000100
18	root	1.1
19	root	1.2	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
20	root	1.1	#define F_DEFAULT 0
21
22			#define INIT_SIZE 32 // initial scalar size to be allocated
23
24			#define SB do {
25			#define SE } while (0)
26
27			static HV *json_stash;
28
29			// structure used for encoding JSON
30			typedef struct
31			{
32			char *cur;
33			STRLEN len; // SvLEN (sv)
34			char *end; // SvEND (sv)
35			SV *sv;
36			UV flags;
37			int max_recurse;
38			int indent;
39			} enc_t;
40
41			// structure used for decoding JSON
42			typedef struct
43			{
44			char *cur;
45			char *end;
46	root	1.4	const char *err;
47	root	1.1	UV flags;
48			} dec_t;
49
50			static UV *
51			SvJSON (SV *sv)
52			{
53			if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54			croak ("object is not of type JSON::XS");
55
56			return &SvUVX (SvRV (sv));
57			}
58
59			/////////////////////////////////////////////////////////////////////////////
60
61			static void
62			need (enc_t *enc, STRLEN len)
63			{
64			if (enc->cur + len >= enc->end)
65			{
66			STRLEN cur = enc->cur - SvPVX (enc->sv);
67			SvGROW (enc->sv, cur + len + 1);
68			enc->cur = SvPVX (enc->sv) + cur;
69	root	1.4	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);
70	root	1.1	}
71			}
72
73			static void
74			encode_ch (enc_t *enc, char ch)
75			{
76			need (enc, 1);
77			*enc->cur++ = ch;
78			}
79
80			static void
81			encode_str (enc_t enc, char str, STRLEN len, int is_utf8)
82			{
83			char *end = str + len;
84
85	root	1.4	need (enc, len);
86
87	root	1.1	while (str < end)
88			{
89			unsigned char ch = (unsigned char )str;
90	root	1.4
91	root	1.6	if (ch >= 0x20 && ch < 0x80) // most common case
92	root	1.4	{
93	root	1.6	if (ch == '"') // but with slow exceptions
94			{
95			need (enc, len += 1);
96			*enc->cur++ = '\\';
97			*enc->cur++ = '"';
98			}
99			else if (ch == '\\')
100			{
101			need (enc, len += 1);
102			*enc->cur++ = '\\';
103			*enc->cur++ = '\\';
104			}
105			else
106			*enc->cur++ = ch;
107
108	root	1.4	++str;
109	root	1.1	}
110			else
111			{
112	root	1.6	switch (ch)
113	root	1.1	{
114	root	1.6	case '\010': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'b'; ++str; break;
115			case '\011': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 't'; ++str; break;
116			case '\012': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'n'; ++str; break;
117			case '\014': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'f'; ++str; break;
118			case '\015': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'r'; ++str; break;
119	root	1.1
120	root	1.6	default:
121	root	1.1	{
122	root	1.6	STRLEN clen;
123			UV uch;
124
125			if (is_utf8)
126			{
127			uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
128			if (clen == (STRLEN)-1)
129			croak ("malformed UTF-8 character in string, cannot convert to JSON");
130			}
131			else
132			{
133			uch = ch;
134			clen = 1;
135			}
136
137			if (uch < 0x80 \|\| enc->flags & F_ASCII)
138			{
139			if (uch > 0xFFFFUL)
140			{
141			need (enc, len += 11);
142			sprintf (enc->cur, "\\u%04x\\u%04x",
143			(uch - 0x10000) / 0x400 + 0xD800,
144			(uch - 0x10000) % 0x400 + 0xDC00);
145			enc->cur += 12;
146			}
147			else
148			{
149			static char hexdigit [16] = "0123456789abcdef";
150			need (enc, len += 5);
151			*enc->cur++ = '\\';
152			*enc->cur++ = 'u';
153			*enc->cur++ = hexdigit [ uch >> 12 ];
154			*enc->cur++ = hexdigit [(uch >> 8) & 15];
155			*enc->cur++ = hexdigit [(uch >> 4) & 15];
156			*enc->cur++ = hexdigit [(uch >> 0) & 15];
157			}
158	root	1.4
159	root	1.6	str += clen;
160			}
161			else if (is_utf8)
162			{
163			need (enc, len += clen);
164			do
165			{
166			enc->cur++ = str++;
167			}
168			while (--clen);
169			}
170			else
171			{
172			need (enc, 10); // never more than 11 bytes needed
173			enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174			++str;
175			}
176	root	1.5	}
177	root	1.4	}
178	root	1.1	}
179
180			--len;
181			}
182			}
183
184			#define INDENT SB \
185			if (enc->flags & F_INDENT) \
186			{ \
187			int i_; \
188			need (enc, enc->indent); \
189			for (i_ = enc->indent * 3; i_--; )\
190			encode_ch (enc, ' '); \
191			} \
192			SE
193
194	root	1.2	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE
195	root	1.1	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
196			#define COMMA SB \
197			encode_ch (enc, ','); \
198			if (enc->flags & F_INDENT) \
199			NL; \
200			else if (enc->flags & F_SPACE_AFTER) \
201			SPACE; \
202			SE
203
204			static void encode_sv (enc_t enc, SV sv);
205
206			static void
207			encode_av (enc_t enc, AV av)
208			{
209			int i, len = av_len (av);
210
211			encode_ch (enc, '['); NL;
212			++enc->indent;
213
214			for (i = 0; i <= len; ++i)
215			{
216			INDENT;
217			encode_sv (enc, *av_fetch (av, i, 0));
218
219			if (i < len)
220			COMMA;
221			}
222
223			NL;
224
225			--enc->indent;
226			INDENT; encode_ch (enc, ']');
227			}
228
229			static void
230			encode_he (enc_t enc, HE he)
231			{
232			encode_ch (enc, '"');
233
234			if (HeKLEN (he) == HEf_SVKEY)
235			{
236			SV *sv = HeSVKEY (he);
237			STRLEN len;
238	root	1.4	char *str;
239
240			SvGETMAGIC (sv);
241			str = SvPV (sv, len);
242	root	1.1
243			encode_str (enc, str, len, SvUTF8 (sv));
244			}
245			else
246			encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
247
248			encode_ch (enc, '"');
249
250			if (enc->flags & F_SPACE_BEFORE) SPACE;
251			encode_ch (enc, ':');
252			if (enc->flags & F_SPACE_AFTER ) SPACE;
253			encode_sv (enc, HeVAL (he));
254			}
255
256			// compare hash entries, used when all keys are bytestrings
257			static int
258			he_cmp_fast (const void a_, const void b_)
259			{
260			int cmp;
261
262			HE a = (HE **)a_;
263			HE b = (HE **)b_;
264
265			STRLEN la = HeKLEN (a);
266			STRLEN lb = HeKLEN (b);
267
268			if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
269			cmp = la < lb ? -1 : la == lb ? 0 : 1;
270
271			return cmp;
272			}
273
274			// compare hash entries, used when some keys are sv's or utf-x
275			static int
276			he_cmp_slow (const void a, const void b)
277			{
278			return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));
279			}
280
281			static void
282			encode_hv (enc_t enc, HV hv)
283			{
284			int count, i;
285
286			encode_ch (enc, '{'); NL; ++enc->indent;
287
288			if ((count = hv_iterinit (hv)))
289			{
290			// for canonical output we have to sort by keys first
291			// actually, this is mostly due to the stupid so-called
292			// security workaround added somewhere in 5.8.x.
293			// that randomises hash orderings
294			if (enc->flags & F_CANONICAL)
295			{
296			HE he, hes [count];
297			int fast = 1;
298
299			i = 0;
300			while ((he = hv_iternext (hv)))
301			{
302			hes [i++] = he;
303			if (HeKLEN (he) < 0 \|\| HeKUTF8 (he))
304			fast = 0;
305			}
306
307			assert (i == count);
308
309			if (fast)
310			qsort (hes, count, sizeof (HE *), he_cmp_fast);
311			else
312			{
313			// hack to disable "use bytes"
314			COP *oldcop = PL_curcop, cop;
315			cop.op_private = 0;
316			PL_curcop = &cop;
317
318			SAVETMPS;
319			qsort (hes, count, sizeof (HE *), he_cmp_slow);
320			FREETMPS;
321
322			PL_curcop = oldcop;
323			}
324
325			for (i = 0; i < count; ++i)
326			{
327			INDENT;
328			encode_he (enc, hes [i]);
329
330			if (i < count - 1)
331			COMMA;
332			}
333
334			NL;
335			}
336			else
337			{
338			SV *sv;
339			HE *he = hv_iternext (hv);
340
341			for (;;)
342			{
343			INDENT;
344			encode_he (enc, he);
345
346			if (!(he = hv_iternext (hv)))
347			break;
348
349			COMMA;
350			}
351
352			NL;
353			}
354			}
355
356			--enc->indent; INDENT; encode_ch (enc, '}');
357			}
358
359			static void
360			encode_sv (enc_t enc, SV sv)
361			{
362	root	1.4	SvGETMAGIC (sv);
363
364	root	1.1	if (SvPOKp (sv))
365			{
366			STRLEN len;
367			char *str = SvPV (sv, len);
368			encode_ch (enc, '"');
369			encode_str (enc, str, len, SvUTF8 (sv));
370			encode_ch (enc, '"');
371			}
372			else if (SvNOKp (sv))
373			{
374			need (enc, NV_DIG + 32);
375			Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
376			enc->cur += strlen (enc->cur);
377			}
378			else if (SvIOKp (sv))
379			{
380			need (enc, 64);
381			enc->cur +=
382			SvIsUV(sv)
383			? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
384			: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
385			}
386			else if (SvROK (sv))
387			{
388			if (!--enc->max_recurse)
389			croak ("data structure too deep (hit recursion limit)");
390
391			sv = SvRV (sv);
392
393			switch (SvTYPE (sv))
394			{
395			case SVt_PVAV: encode_av (enc, (AV *)sv); break;
396			case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
397
398			default:
399			croak ("JSON can only represent references to arrays or hashes");
400			}
401			}
402			else if (!SvOK (sv))
403			encode_str (enc, "null", 4, 0);
404			else
405			croak ("encountered perl type that JSON cannot handle");
406			}
407
408			static SV *
409			encode_json (SV *scalar, UV flags)
410			{
411	root	1.3	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
412			croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
413
414	root	1.1	enc_t enc;
415			enc.flags = flags;
416			enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
417			enc.cur = SvPVX (enc.sv);
418			enc.end = SvEND (enc.sv);
419			enc.max_recurse = 0;
420			enc.indent = 0;
421
422			SvPOK_only (enc.sv);
423			encode_sv (&enc, scalar);
424
425			if (!(flags & (F_ASCII \| F_UTF8)))
426			SvUTF8_on (enc.sv);
427
428			SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
429	root	1.6
430			#ifdef SvPV_shrink_to_cur
431			if (enc.flags & F_SHRINK)
432			SvPV_shrink_to_cur (enc.sv);
433			#endif
434	root	1.1	return enc.sv;
435			}
436
437			/////////////////////////////////////////////////////////////////////////////
438
439			#define WS \
440			for (;;) \
441			{ \
442			char ch = *dec->cur; \
443			if (ch > 0x20 \
444			\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
445			break; \
446			++dec->cur; \
447			}
448
449			#define ERR(reason) SB dec->err = reason; goto fail; SE
450			#define EXPECT_CH(ch) SB \
451			if (*dec->cur != ch) \
452			ERR (# ch " expected"); \
453			++dec->cur; \
454			SE
455
456			static SV decode_sv (dec_t dec);
457
458			static signed char decode_hexdigit[256];
459
460			static UV
461			decode_4hex (dec_t *dec)
462			{
463			signed char d1, d2, d3, d4;
464
465			d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
466			if (d1 < 0) ERR ("four hexadecimal digits expected");
467			d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
468			if (d2 < 0) ERR ("four hexadecimal digits expected");
469			d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
470			if (d3 < 0) ERR ("four hexadecimal digits expected");
471			d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
472			if (d4 < 0) ERR ("four hexadecimal digits expected");
473
474			dec->cur += 4;
475
476			return ((UV)d1) << 12
477			\| ((UV)d2) << 8
478			\| ((UV)d3) << 4
479			\| ((UV)d4);
480
481			fail:
482			return (UV)-1;
483			}
484
485	root	1.4	#define APPEND_GROW(n) SB \
486			if (cur + (n) >= end) \
487			{ \
488			STRLEN ofs = cur - SvPVX (sv); \
489			SvGROW (sv, ofs + (n) + 1); \
490			cur = SvPVX (sv) + ofs; \
491			end = SvEND (sv); \
492			} \
493			SE
494
495			#define APPEND_CH(ch) SB \
496			APPEND_GROW (1); \
497			*cur++ = (ch); \
498			SE
499
500	root	1.1	static SV *
501			decode_str (dec_t *dec)
502			{
503			SV *sv = NEWSV (0,2);
504			int utf8 = 0;
505	root	1.4	char *cur = SvPVX (sv);
506			char *end = SvEND (sv);
507	root	1.1
508			for (;;)
509			{
510			unsigned char ch = (unsigned char )dec->cur;
511
512			if (ch == '"')
513			break;
514			else if (ch == '\\')
515			{
516			switch (*++dec->cur)
517			{
518			case '\\':
519			case '/':
520			case '"': APPEND_CH (*dec->cur++); break;
521
522			case 'b': APPEND_CH ('\010'); ++dec->cur; break;
523			case 't': APPEND_CH ('\011'); ++dec->cur; break;
524			case 'n': APPEND_CH ('\012'); ++dec->cur; break;
525			case 'f': APPEND_CH ('\014'); ++dec->cur; break;
526			case 'r': APPEND_CH ('\015'); ++dec->cur; break;
527
528			case 'u':
529			{
530			UV lo, hi;
531			++dec->cur;
532
533			hi = decode_4hex (dec);
534			if (hi == (UV)-1)
535			goto fail;
536
537			// possibly a surrogate pair
538			if (hi >= 0xd800 && hi < 0xdc00)
539			{
540			if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
541	root	1.5	ERR ("missing low surrogate character in surrogate pair");
542	root	1.1
543			dec->cur += 2;
544
545			lo = decode_4hex (dec);
546			if (lo == (UV)-1)
547			goto fail;
548
549			if (lo < 0xdc00 \|\| lo >= 0xe000)
550			ERR ("surrogate pair expected");
551
552			hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
553			}
554	root	1.5	else if (hi >= 0xdc00 && hi < 0xe000)
555			ERR ("missing high surrogate character in surrogate pair");
556	root	1.1
557			if (hi >= 0x80)
558			{
559			utf8 = 1;
560
561	root	1.4	APPEND_GROW (4); // at most 4 bytes for 21 bits
562			cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
563	root	1.1	}
564			else
565			APPEND_CH (hi);
566			}
567			break;
568	root	1.5
569			default:
570			--dec->cur;
571			ERR ("illegal backslash escape sequence in string");
572	root	1.1	}
573			}
574			else if (ch >= 0x20 && ch <= 0x7f)
575			APPEND_CH (*dec->cur++);
576			else if (ch >= 0x80)
577			{
578			STRLEN clen;
579			UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
580	root	1.5	if (clen == (STRLEN)-1)
581	root	1.1	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
582
583	root	1.4	APPEND_GROW (clen);
584	root	1.5	do
585			{
586			cur++ = dec->cur++;
587			}
588			while (--clen);
589
590			utf8 = 1;
591	root	1.1	}
592	root	1.5	else if (dec->cur == dec->end)
593			ERR ("unexpected end of string while parsing json string");
594	root	1.1	else
595			ERR ("invalid character encountered");
596			}
597
598			++dec->cur;
599
600	root	1.4	SvCUR_set (sv, cur - SvPVX (sv));
601
602	root	1.1	SvPOK_only (sv);
603			*SvEND (sv) = 0;
604
605			if (utf8)
606			SvUTF8_on (sv);
607
608	root	1.6	#ifdef SvPV_shrink_to_cur
609			if (dec->flags & F_SHRINK)
610			SvPV_shrink_to_cur (sv);
611			#endif
612
613	root	1.1	return sv;
614
615			fail:
616			SvREFCNT_dec (sv);
617			return 0;
618			}
619
620			static SV *
621			decode_num (dec_t *dec)
622			{
623			int is_nv = 0;
624			char *start = dec->cur;
625
626			// [minus]
627			if (*dec->cur == '-')
628			++dec->cur;
629
630			if (*dec->cur == '0')
631			{
632			++dec->cur;
633			if (dec->cur >= '0' && dec->cur <= '9')
634			ERR ("malformed number (leading zero must not be followed by another digit)");
635			}
636	root	1.5	else if (dec->cur < '0' \|\| dec->cur > '9')
637			ERR ("malformed number (no digits after initial minus)");
638			else
639			do
640			{
641			++dec->cur;
642			}
643			while (dec->cur >= '0' && dec->cur <= '9');
644	root	1.1
645			// [frac]
646			if (*dec->cur == '.')
647			{
648	root	1.5	++dec->cur;
649
650			if (dec->cur < '0' \|\| dec->cur > '9')
651			ERR ("malformed number (no digits after decimal point)");
652	root	1.1
653			do
654			{
655			++dec->cur;
656			}
657			while (dec->cur >= '0' && dec->cur <= '9');
658	root	1.5
659			is_nv = 1;
660	root	1.1	}
661
662			// [exp]
663			if (dec->cur == 'e' \|\| dec->cur == 'E')
664			{
665	root	1.5	++dec->cur;
666	root	1.1
667			if (dec->cur == '-' \|\| dec->cur == '+')
668			++dec->cur;
669
670	root	1.5	if (dec->cur < '0' \|\| dec->cur > '9')
671			ERR ("malformed number (no digits after exp sign)");
672
673			do
674			{
675			++dec->cur;
676			}
677			while (dec->cur >= '0' && dec->cur <= '9');
678
679			is_nv = 1;
680	root	1.1	}
681
682			if (!is_nv)
683			{
684			UV uv;
685			int numtype = grok_number (start, dec->cur - start, &uv);
686			if (numtype & IS_NUMBER_IN_UV)
687			if (numtype & IS_NUMBER_NEG)
688			{
689			if (uv < (UV)IV_MIN)
690			return newSViv (-(IV)uv);
691			}
692			else
693			return newSVuv (uv);
694			}
695
696			return newSVnv (Atof (start));
697
698			fail:
699			return 0;
700			}
701
702			static SV *
703			decode_av (dec_t *dec)
704			{
705			AV *av = newAV ();
706
707	root	1.5	WS;
708			if (*dec->cur == ']')
709			++dec->cur;
710			else
711			for (;;)
712			{
713			SV *value;
714	root	1.1
715	root	1.5	value = decode_sv (dec);
716			if (!value)
717			goto fail;
718	root	1.1
719	root	1.5	av_push (av, value);
720	root	1.1
721	root	1.5	WS;
722	root	1.1
723	root	1.5	if (*dec->cur == ']')
724			{
725			++dec->cur;
726			break;
727			}
728
729			if (*dec->cur != ',')
730			ERR (", or ] expected while parsing array");
731	root	1.1
732	root	1.5	++dec->cur;
733			}
734	root	1.1
735			return newRV_noinc ((SV *)av);
736
737			fail:
738			SvREFCNT_dec (av);
739			return 0;
740			}
741
742			static SV *
743			decode_hv (dec_t *dec)
744			{
745			HV *hv = newHV ();
746
747	root	1.5	WS;
748			if (*dec->cur == '}')
749			++dec->cur;
750			else
751			for (;;)
752			{
753			SV key, value;
754	root	1.1
755	root	1.5	WS; EXPECT_CH ('"');
756	root	1.1
757	root	1.5	key = decode_str (dec);
758			if (!key)
759			goto fail;
760	root	1.1
761	root	1.5	WS; EXPECT_CH (':');
762	root	1.1
763	root	1.5	value = decode_sv (dec);
764			if (!value)
765			{
766			SvREFCNT_dec (key);
767			goto fail;
768			}
769	root	1.1
770	root	1.5	//TODO: optimise
771			hv_store_ent (hv, key, value, 0);
772	root	1.1
773	root	1.5	WS;
774	root	1.1
775	root	1.5	if (*dec->cur == '}')
776			{
777			++dec->cur;
778			break;
779			}
780	root	1.1
781	root	1.5	if (*dec->cur != ',')
782			ERR (", or } expected while parsing object/hash");
783	root	1.1
784	root	1.5	++dec->cur;
785			}
786	root	1.1
787			return newRV_noinc ((SV *)hv);
788
789			fail:
790			SvREFCNT_dec (hv);
791			return 0;
792			}
793
794			static SV *
795			decode_sv (dec_t *dec)
796			{
797			WS;
798			switch (*dec->cur)
799			{
800			case '"': ++dec->cur; return decode_str (dec);
801			case '[': ++dec->cur; return decode_av (dec);
802			case '{': ++dec->cur; return decode_hv (dec);
803
804			case '-':
805			case '0': case '1': case '2': case '3': case '4':
806			case '5': case '6': case '7': case '8': case '9':
807			return decode_num (dec);
808
809			case 't':
810			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
811			{
812			dec->cur += 4;
813			return newSViv (1);
814			}
815			else
816			ERR ("'true' expected");
817
818			break;
819
820			case 'f':
821			if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
822			{
823			dec->cur += 5;
824			return newSViv (0);
825			}
826			else
827			ERR ("'false' expected");
828
829			break;
830
831			case 'n':
832			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
833			{
834			dec->cur += 4;
835	root	1.5	return newSVsv (&PL_sv_undef);
836	root	1.1	}
837			else
838			ERR ("'null' expected");
839
840			break;
841
842			default:
843			ERR ("malformed json string");
844			break;
845			}
846
847			fail:
848			return 0;
849			}
850
851			static SV *
852			decode_json (SV *string, UV flags)
853			{
854			SV *sv;
855
856	root	1.5	if (flags & F_UTF8)
857			sv_utf8_downgrade (string, 0);
858			else
859	root	1.1	sv_utf8_upgrade (string);
860
861			SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
862
863			dec_t dec;
864			dec.flags = flags;
865			dec.cur = SvPVX (string);
866			dec.end = SvEND (string);
867			dec.err = 0;
868
869			sv = decode_sv (&dec);
870
871			if (!sv)
872			{
873			IV offset = utf8_distance (dec.cur, SvPVX (string));
874			SV *uni = sv_newmortal ();
875	root	1.5	// horrible hack to silence warning inside pv_uni_display
876			COP cop;
877			memset (&cop, 0, sizeof (cop));
878			cop.cop_warnings = pWARN_NONE;
879			SAVEVPTR (PL_curcop);
880			PL_curcop = &cop;
881	root	1.1
882			pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
883	root	1.5	croak ("%s, at character offset %d (%s)",
884	root	1.1	dec.err,
885			(int)offset,
886			dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
887			}
888
889	root	1.3	sv = sv_2mortal (sv);
890
891			if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
892			croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");
893
894			return sv;
895	root	1.1	}
896
897			MODULE = JSON::XS PACKAGE = JSON::XS
898
899			BOOT:
900			{
901			int i;
902
903			memset (decode_hexdigit, 0xff, 256);
904			for (i = 10; i--; )
905			decode_hexdigit ['0' + i] = i;
906
907	root	1.4	for (i = 7; i--; )
908	root	1.1	{
909			decode_hexdigit ['a' + i] = 10 + i;
910			decode_hexdigit ['A' + i] = 10 + i;
911			}
912
913			json_stash = gv_stashpv ("JSON::XS", 1);
914			}
915
916	root	1.4	PROTOTYPES: DISABLE
917
918	root	1.1	SV new (char dummy)
919			CODE:
920			RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
921			OUTPUT:
922			RETVAL
923
924	root	1.6	SV ascii (SV self, int enable = 1)
925	root	1.1	ALIAS:
926			ascii = F_ASCII
927			utf8 = F_UTF8
928			indent = F_INDENT
929			canonical = F_CANONICAL
930			space_before = F_SPACE_BEFORE
931			space_after = F_SPACE_AFTER
932			json_rpc = F_JSON_RPC
933	root	1.2	pretty = F_PRETTY
934	root	1.3	allow_nonref = F_ALLOW_NONREF
935	root	1.6	shrink = F_SHRINK
936	root	1.1	CODE:
937			{
938			UV *uv = SvJSON (self);
939			if (enable)
940			*uv \|= ix;
941			else
942			*uv &= ~ix;
943
944			RETVAL = newSVsv (self);
945			}
946			OUTPUT:
947			RETVAL
948
949			void encode (SV self, SV scalar)
950			PPCODE:
951			XPUSHs (encode_json (scalar, *SvJSON (self)));
952
953	root	1.2	void decode (SV self, SV jsonstr)
954	root	1.1	PPCODE:
955	root	1.2	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
956
957	root	1.4	PROTOTYPES: ENABLE
958
959	root	1.2	void to_json (SV *scalar)
960			PPCODE:
961			XPUSHs (encode_json (scalar, F_UTF8));
962
963			void from_json (SV *jsonstr)
964			PPCODE:
965			XPUSHs (decode_json (jsonstr, F_UTF8));
966	root	1.1