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
shrink (SV *sv)
{
  sv_utf8_downgrade (sv, 1);
#ifdef SvPV_shrink_to_cur
  SvPV_shrink_to_cur (sv);
#endif
}

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

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 or illegal unicode character in string [%.11s], cannot convert to JSON", str);
                    }
                  else
                    {
                      uch = ch;
                      clen = 1;
                    }

                  if (uch > 0x10FFFFUL)
                    croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);

                  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, len += 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 forcefully disable "use bytes"
              COP cop = *PL_curcop;
              cop.op_private = 0;

              ENTER;
              SAVETMPS;

              SAVEVPTR (PL_curcop);
              PL_curcop = &cop;

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

              FREETMPS;
              LEAVE;
            }

          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))
    {
      SV *rv = SvRV (sv);

      if (!--enc->max_recurse)
        croak ("data structure too deep (hit recursion limit)");

      switch (SvTYPE (rv))
        {
          case SVt_PVAV: encode_av (enc, (AV *)rv); break;
          case SVt_PVHV: encode_hv (enc, (HV *)rv); break;

          default:
            croak ("encountered %s, but JSON can only represent references to arrays or hashes",
                   SvPV_nolen (sv));
        }
    }
  else if (!SvOK (sv))
    encode_str (enc, "null", 4, 0);
  else
    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)
{
  if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
    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;

  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));

  if (enc.flags & F_SHRINK)
    shrink (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 JSON string");

          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);

  if (dec->flags & F_SHRINK)
    shrink (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, neither array, object, number, string or atom");
        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 = dec.flags & F_UTF8
                  ? dec.cur - SvPVX (string)
                  : utf8_distance (dec.cur, SvPVX (string));
      SV *uni = sv_newmortal ();

      // horrible hack to silence warning inside pv_uni_display
      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);
      LEAVE;

      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 text must be an object or array (but found number, string, true, false or null, 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:	1.9
Committed:	Fri Mar 23 17:40:29 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
CVS Tags:	rel-0_3
Changes since 1.8:	+15 -10 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	root	1.7	static void
60			shrink (SV *sv)
61			{
62			sv_utf8_downgrade (sv, 1);
63			#ifdef SvPV_shrink_to_cur
64			SvPV_shrink_to_cur (sv);
65			#endif
66			}
67
68	root	1.1	/////////////////////////////////////////////////////////////////////////////
69
70			static void
71			need (enc_t *enc, STRLEN len)
72			{
73			if (enc->cur + len >= enc->end)
74			{
75			STRLEN cur = enc->cur - SvPVX (enc->sv);
76			SvGROW (enc->sv, cur + len + 1);
77			enc->cur = SvPVX (enc->sv) + cur;
78	root	1.4	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);
79	root	1.1	}
80			}
81
82			static void
83			encode_ch (enc_t *enc, char ch)
84			{
85			need (enc, 1);
86			*enc->cur++ = ch;
87			}
88
89			static void
90			encode_str (enc_t enc, char str, STRLEN len, int is_utf8)
91			{
92			char *end = str + len;
93
94	root	1.4	need (enc, len);
95
96	root	1.1	while (str < end)
97			{
98			unsigned char ch = (unsigned char )str;
99	root	1.4
100	root	1.6	if (ch >= 0x20 && ch < 0x80) // most common case
101	root	1.4	{
102	root	1.6	if (ch == '"') // but with slow exceptions
103			{
104			need (enc, len += 1);
105			*enc->cur++ = '\\';
106			*enc->cur++ = '"';
107			}
108			else if (ch == '\\')
109			{
110			need (enc, len += 1);
111			*enc->cur++ = '\\';
112			*enc->cur++ = '\\';
113			}
114			else
115			*enc->cur++ = ch;
116
117	root	1.4	++str;
118	root	1.1	}
119			else
120			{
121	root	1.6	switch (ch)
122	root	1.1	{
123	root	1.6	case '\010': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'b'; ++str; break;
124			case '\011': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 't'; ++str; break;
125			case '\012': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'n'; ++str; break;
126			case '\014': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'f'; ++str; break;
127			case '\015': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'r'; ++str; break;
128	root	1.1
129	root	1.6	default:
130	root	1.1	{
131	root	1.6	STRLEN clen;
132			UV uch;
133
134			if (is_utf8)
135			{
136			uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
137			if (clen == (STRLEN)-1)
138	root	1.9	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	root	1.6	}
140			else
141			{
142			uch = ch;
143			clen = 1;
144			}
145
146	root	1.9	if (uch > 0x10FFFFUL)
147			croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
148
149	root	1.6	if (uch < 0x80 \|\| enc->flags & F_ASCII)
150			{
151			if (uch > 0xFFFFUL)
152			{
153			need (enc, len += 11);
154			sprintf (enc->cur, "\\u%04x\\u%04x",
155			(uch - 0x10000) / 0x400 + 0xD800,
156			(uch - 0x10000) % 0x400 + 0xDC00);
157			enc->cur += 12;
158			}
159			else
160			{
161			static char hexdigit [16] = "0123456789abcdef";
162			need (enc, len += 5);
163			*enc->cur++ = '\\';
164			*enc->cur++ = 'u';
165			*enc->cur++ = hexdigit [ uch >> 12 ];
166			*enc->cur++ = hexdigit [(uch >> 8) & 15];
167			*enc->cur++ = hexdigit [(uch >> 4) & 15];
168			*enc->cur++ = hexdigit [(uch >> 0) & 15];
169			}
170	root	1.4
171	root	1.6	str += clen;
172			}
173			else if (is_utf8)
174			{
175			need (enc, len += clen);
176			do
177			{
178			enc->cur++ = str++;
179			}
180			while (--clen);
181			}
182			else
183			{
184	root	1.7	need (enc, len += 10); // never more than 11 bytes needed
185	root	1.6	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
186			++str;
187			}
188	root	1.5	}
189	root	1.4	}
190	root	1.1	}
191
192			--len;
193			}
194			}
195
196			#define INDENT SB \
197			if (enc->flags & F_INDENT) \
198			{ \
199			int i_; \
200			need (enc, enc->indent); \
201			for (i_ = enc->indent * 3; i_--; )\
202			encode_ch (enc, ' '); \
203			} \
204			SE
205
206	root	1.2	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE
207	root	1.1	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
208			#define COMMA SB \
209			encode_ch (enc, ','); \
210			if (enc->flags & F_INDENT) \
211			NL; \
212			else if (enc->flags & F_SPACE_AFTER) \
213			SPACE; \
214			SE
215
216			static void encode_sv (enc_t enc, SV sv);
217
218			static void
219			encode_av (enc_t enc, AV av)
220			{
221			int i, len = av_len (av);
222
223			encode_ch (enc, '['); NL;
224			++enc->indent;
225
226			for (i = 0; i <= len; ++i)
227			{
228			INDENT;
229			encode_sv (enc, *av_fetch (av, i, 0));
230
231			if (i < len)
232			COMMA;
233			}
234
235			NL;
236
237			--enc->indent;
238			INDENT; encode_ch (enc, ']');
239			}
240
241			static void
242			encode_he (enc_t enc, HE he)
243			{
244			encode_ch (enc, '"');
245
246			if (HeKLEN (he) == HEf_SVKEY)
247			{
248			SV *sv = HeSVKEY (he);
249			STRLEN len;
250	root	1.4	char *str;
251
252			SvGETMAGIC (sv);
253			str = SvPV (sv, len);
254	root	1.1
255			encode_str (enc, str, len, SvUTF8 (sv));
256			}
257			else
258			encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
259
260			encode_ch (enc, '"');
261
262			if (enc->flags & F_SPACE_BEFORE) SPACE;
263			encode_ch (enc, ':');
264			if (enc->flags & F_SPACE_AFTER ) SPACE;
265			encode_sv (enc, HeVAL (he));
266			}
267
268			// compare hash entries, used when all keys are bytestrings
269			static int
270			he_cmp_fast (const void a_, const void b_)
271			{
272			int cmp;
273
274			HE a = (HE **)a_;
275			HE b = (HE **)b_;
276
277			STRLEN la = HeKLEN (a);
278			STRLEN lb = HeKLEN (b);
279
280			if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
281			cmp = la < lb ? -1 : la == lb ? 0 : 1;
282
283			return cmp;
284			}
285
286			// compare hash entries, used when some keys are sv's or utf-x
287			static int
288			he_cmp_slow (const void a, const void b)
289			{
290			return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));
291			}
292
293			static void
294			encode_hv (enc_t enc, HV hv)
295			{
296			int count, i;
297
298			encode_ch (enc, '{'); NL; ++enc->indent;
299
300			if ((count = hv_iterinit (hv)))
301			{
302			// for canonical output we have to sort by keys first
303			// actually, this is mostly due to the stupid so-called
304			// security workaround added somewhere in 5.8.x.
305			// that randomises hash orderings
306			if (enc->flags & F_CANONICAL)
307			{
308			HE he, hes [count];
309			int fast = 1;
310
311			i = 0;
312			while ((he = hv_iternext (hv)))
313			{
314			hes [i++] = he;
315			if (HeKLEN (he) < 0 \|\| HeKUTF8 (he))
316			fast = 0;
317			}
318
319			assert (i == count);
320
321			if (fast)
322			qsort (hes, count, sizeof (HE *), he_cmp_fast);
323			else
324			{
325	root	1.8	// hack to forcefully disable "use bytes"
326			COP cop = *PL_curcop;
327	root	1.1	cop.op_private = 0;
328	root	1.8
329			ENTER;
330			SAVETMPS;
331
332			SAVEVPTR (PL_curcop);
333	root	1.1	PL_curcop = &cop;
334
335			qsort (hes, count, sizeof (HE *), he_cmp_slow);
336	root	1.8
337	root	1.1	FREETMPS;
338	root	1.8	LEAVE;
339	root	1.1	}
340
341			for (i = 0; i < count; ++i)
342			{
343			INDENT;
344			encode_he (enc, hes [i]);
345
346			if (i < count - 1)
347			COMMA;
348			}
349
350			NL;
351			}
352			else
353			{
354			SV *sv;
355			HE *he = hv_iternext (hv);
356
357			for (;;)
358			{
359			INDENT;
360			encode_he (enc, he);
361
362			if (!(he = hv_iternext (hv)))
363			break;
364
365			COMMA;
366			}
367
368			NL;
369			}
370			}
371
372			--enc->indent; INDENT; encode_ch (enc, '}');
373			}
374
375			static void
376			encode_sv (enc_t enc, SV sv)
377			{
378	root	1.4	SvGETMAGIC (sv);
379
380	root	1.1	if (SvPOKp (sv))
381			{
382			STRLEN len;
383			char *str = SvPV (sv, len);
384			encode_ch (enc, '"');
385			encode_str (enc, str, len, SvUTF8 (sv));
386			encode_ch (enc, '"');
387			}
388			else if (SvNOKp (sv))
389			{
390			need (enc, NV_DIG + 32);
391			Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
392			enc->cur += strlen (enc->cur);
393			}
394			else if (SvIOKp (sv))
395			{
396			need (enc, 64);
397			enc->cur +=
398			SvIsUV(sv)
399			? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
400			: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
401			}
402			else if (SvROK (sv))
403			{
404	root	1.9	SV *rv = SvRV (sv);
405
406	root	1.1	if (!--enc->max_recurse)
407			croak ("data structure too deep (hit recursion limit)");
408
409	root	1.9	switch (SvTYPE (rv))
410	root	1.1	{
411	root	1.9	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
412			case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
413	root	1.1
414			default:
415	root	1.9	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
416			SvPV_nolen (sv));
417	root	1.1	}
418			}
419			else if (!SvOK (sv))
420			encode_str (enc, "null", 4, 0);
421			else
422	root	1.9	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
423			SvPV_nolen (sv), SvFLAGS (sv));
424	root	1.1	}
425
426			static SV *
427			encode_json (SV *scalar, UV flags)
428			{
429	root	1.3	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
430	root	1.9	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
431	root	1.3
432	root	1.1	enc_t enc;
433			enc.flags = flags;
434			enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
435			enc.cur = SvPVX (enc.sv);
436			enc.end = SvEND (enc.sv);
437			enc.max_recurse = 0;
438			enc.indent = 0;
439
440			SvPOK_only (enc.sv);
441			encode_sv (&enc, scalar);
442
443			if (!(flags & (F_ASCII \| F_UTF8)))
444			SvUTF8_on (enc.sv);
445
446			SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
447	root	1.6
448			if (enc.flags & F_SHRINK)
449	root	1.7	shrink (enc.sv);
450
451	root	1.1	return enc.sv;
452			}
453
454			/////////////////////////////////////////////////////////////////////////////
455
456			#define WS \
457			for (;;) \
458			{ \
459			char ch = *dec->cur; \
460			if (ch > 0x20 \
461			\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
462			break; \
463			++dec->cur; \
464			}
465
466			#define ERR(reason) SB dec->err = reason; goto fail; SE
467			#define EXPECT_CH(ch) SB \
468			if (*dec->cur != ch) \
469			ERR (# ch " expected"); \
470			++dec->cur; \
471			SE
472
473			static SV decode_sv (dec_t dec);
474
475			static signed char decode_hexdigit[256];
476
477			static UV
478			decode_4hex (dec_t *dec)
479			{
480			signed char d1, d2, d3, d4;
481
482			d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
483			if (d1 < 0) ERR ("four hexadecimal digits expected");
484			d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
485			if (d2 < 0) ERR ("four hexadecimal digits expected");
486			d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
487			if (d3 < 0) ERR ("four hexadecimal digits expected");
488			d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
489			if (d4 < 0) ERR ("four hexadecimal digits expected");
490
491			dec->cur += 4;
492
493			return ((UV)d1) << 12
494			\| ((UV)d2) << 8
495			\| ((UV)d3) << 4
496			\| ((UV)d4);
497
498			fail:
499			return (UV)-1;
500			}
501
502	root	1.4	#define APPEND_GROW(n) SB \
503			if (cur + (n) >= end) \
504			{ \
505			STRLEN ofs = cur - SvPVX (sv); \
506			SvGROW (sv, ofs + (n) + 1); \
507			cur = SvPVX (sv) + ofs; \
508			end = SvEND (sv); \
509			} \
510			SE
511
512			#define APPEND_CH(ch) SB \
513			APPEND_GROW (1); \
514			*cur++ = (ch); \
515			SE
516
517	root	1.1	static SV *
518			decode_str (dec_t *dec)
519			{
520			SV *sv = NEWSV (0,2);
521			int utf8 = 0;
522	root	1.4	char *cur = SvPVX (sv);
523			char *end = SvEND (sv);
524	root	1.1
525			for (;;)
526			{
527			unsigned char ch = (unsigned char )dec->cur;
528
529			if (ch == '"')
530			break;
531			else if (ch == '\\')
532			{
533			switch (*++dec->cur)
534			{
535			case '\\':
536			case '/':
537			case '"': APPEND_CH (*dec->cur++); break;
538
539			case 'b': APPEND_CH ('\010'); ++dec->cur; break;
540			case 't': APPEND_CH ('\011'); ++dec->cur; break;
541			case 'n': APPEND_CH ('\012'); ++dec->cur; break;
542			case 'f': APPEND_CH ('\014'); ++dec->cur; break;
543			case 'r': APPEND_CH ('\015'); ++dec->cur; break;
544
545			case 'u':
546			{
547			UV lo, hi;
548			++dec->cur;
549
550			hi = decode_4hex (dec);
551			if (hi == (UV)-1)
552			goto fail;
553
554			// possibly a surrogate pair
555			if (hi >= 0xd800 && hi < 0xdc00)
556			{
557			if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
558	root	1.5	ERR ("missing low surrogate character in surrogate pair");
559	root	1.1
560			dec->cur += 2;
561
562			lo = decode_4hex (dec);
563			if (lo == (UV)-1)
564			goto fail;
565
566			if (lo < 0xdc00 \|\| lo >= 0xe000)
567			ERR ("surrogate pair expected");
568
569			hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
570			}
571	root	1.5	else if (hi >= 0xdc00 && hi < 0xe000)
572			ERR ("missing high surrogate character in surrogate pair");
573	root	1.1
574			if (hi >= 0x80)
575			{
576			utf8 = 1;
577
578	root	1.4	APPEND_GROW (4); // at most 4 bytes for 21 bits
579			cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
580	root	1.1	}
581			else
582			APPEND_CH (hi);
583			}
584			break;
585	root	1.5
586			default:
587			--dec->cur;
588			ERR ("illegal backslash escape sequence in string");
589	root	1.1	}
590			}
591			else if (ch >= 0x20 && ch <= 0x7f)
592			APPEND_CH (*dec->cur++);
593			else if (ch >= 0x80)
594			{
595			STRLEN clen;
596			UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
597	root	1.5	if (clen == (STRLEN)-1)
598	root	1.7	ERR ("malformed UTF-8 character in JSON string");
599	root	1.1
600	root	1.4	APPEND_GROW (clen);
601	root	1.5	do
602			{
603			cur++ = dec->cur++;
604			}
605			while (--clen);
606
607			utf8 = 1;
608	root	1.1	}
609	root	1.5	else if (dec->cur == dec->end)
610			ERR ("unexpected end of string while parsing json string");
611	root	1.1	else
612			ERR ("invalid character encountered");
613			}
614
615			++dec->cur;
616
617	root	1.4	SvCUR_set (sv, cur - SvPVX (sv));
618
619	root	1.1	SvPOK_only (sv);
620			*SvEND (sv) = 0;
621
622			if (utf8)
623			SvUTF8_on (sv);
624
625	root	1.6	if (dec->flags & F_SHRINK)
626	root	1.7	shrink (sv);
627	root	1.6
628	root	1.1	return sv;
629
630			fail:
631			SvREFCNT_dec (sv);
632			return 0;
633			}
634
635			static SV *
636			decode_num (dec_t *dec)
637			{
638			int is_nv = 0;
639			char *start = dec->cur;
640
641			// [minus]
642			if (*dec->cur == '-')
643			++dec->cur;
644
645			if (*dec->cur == '0')
646			{
647			++dec->cur;
648			if (dec->cur >= '0' && dec->cur <= '9')
649			ERR ("malformed number (leading zero must not be followed by another digit)");
650			}
651	root	1.5	else if (dec->cur < '0' \|\| dec->cur > '9')
652			ERR ("malformed number (no digits after initial minus)");
653			else
654			do
655			{
656			++dec->cur;
657			}
658			while (dec->cur >= '0' && dec->cur <= '9');
659	root	1.1
660			// [frac]
661			if (*dec->cur == '.')
662			{
663	root	1.5	++dec->cur;
664
665			if (dec->cur < '0' \|\| dec->cur > '9')
666			ERR ("malformed number (no digits after decimal point)");
667	root	1.1
668			do
669			{
670			++dec->cur;
671			}
672			while (dec->cur >= '0' && dec->cur <= '9');
673	root	1.5
674			is_nv = 1;
675	root	1.1	}
676
677			// [exp]
678			if (dec->cur == 'e' \|\| dec->cur == 'E')
679			{
680	root	1.5	++dec->cur;
681	root	1.1
682			if (dec->cur == '-' \|\| dec->cur == '+')
683			++dec->cur;
684
685	root	1.5	if (dec->cur < '0' \|\| dec->cur > '9')
686			ERR ("malformed number (no digits after exp sign)");
687
688			do
689			{
690			++dec->cur;
691			}
692			while (dec->cur >= '0' && dec->cur <= '9');
693
694			is_nv = 1;
695	root	1.1	}
696
697			if (!is_nv)
698			{
699			UV uv;
700			int numtype = grok_number (start, dec->cur - start, &uv);
701			if (numtype & IS_NUMBER_IN_UV)
702			if (numtype & IS_NUMBER_NEG)
703			{
704			if (uv < (UV)IV_MIN)
705			return newSViv (-(IV)uv);
706			}
707			else
708			return newSVuv (uv);
709			}
710
711			return newSVnv (Atof (start));
712
713			fail:
714			return 0;
715			}
716
717			static SV *
718			decode_av (dec_t *dec)
719			{
720			AV *av = newAV ();
721
722	root	1.5	WS;
723			if (*dec->cur == ']')
724			++dec->cur;
725			else
726			for (;;)
727			{
728			SV *value;
729	root	1.1
730	root	1.5	value = decode_sv (dec);
731			if (!value)
732			goto fail;
733	root	1.1
734	root	1.5	av_push (av, value);
735	root	1.1
736	root	1.5	WS;
737	root	1.1
738	root	1.5	if (*dec->cur == ']')
739			{
740			++dec->cur;
741			break;
742			}
743
744			if (*dec->cur != ',')
745			ERR (", or ] expected while parsing array");
746	root	1.1
747	root	1.5	++dec->cur;
748			}
749	root	1.1
750			return newRV_noinc ((SV *)av);
751
752			fail:
753			SvREFCNT_dec (av);
754			return 0;
755			}
756
757			static SV *
758			decode_hv (dec_t *dec)
759			{
760			HV *hv = newHV ();
761
762	root	1.5	WS;
763			if (*dec->cur == '}')
764			++dec->cur;
765			else
766			for (;;)
767			{
768			SV key, value;
769	root	1.1
770	root	1.5	WS; EXPECT_CH ('"');
771	root	1.1
772	root	1.5	key = decode_str (dec);
773			if (!key)
774			goto fail;
775	root	1.1
776	root	1.5	WS; EXPECT_CH (':');
777	root	1.1
778	root	1.5	value = decode_sv (dec);
779			if (!value)
780			{
781			SvREFCNT_dec (key);
782			goto fail;
783			}
784	root	1.1
785	root	1.5	//TODO: optimise
786			hv_store_ent (hv, key, value, 0);
787	root	1.1
788	root	1.5	WS;
789	root	1.1
790	root	1.5	if (*dec->cur == '}')
791			{
792			++dec->cur;
793			break;
794			}
795	root	1.1
796	root	1.5	if (*dec->cur != ',')
797			ERR (", or } expected while parsing object/hash");
798	root	1.1
799	root	1.5	++dec->cur;
800			}
801	root	1.1
802			return newRV_noinc ((SV *)hv);
803
804			fail:
805			SvREFCNT_dec (hv);
806			return 0;
807			}
808
809			static SV *
810			decode_sv (dec_t *dec)
811			{
812			WS;
813			switch (*dec->cur)
814			{
815			case '"': ++dec->cur; return decode_str (dec);
816			case '[': ++dec->cur; return decode_av (dec);
817			case '{': ++dec->cur; return decode_hv (dec);
818
819			case '-':
820			case '0': case '1': case '2': case '3': case '4':
821			case '5': case '6': case '7': case '8': case '9':
822			return decode_num (dec);
823
824			case 't':
825			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
826			{
827			dec->cur += 4;
828			return newSViv (1);
829			}
830			else
831			ERR ("'true' expected");
832
833			break;
834
835			case 'f':
836			if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
837			{
838			dec->cur += 5;
839			return newSViv (0);
840			}
841			else
842			ERR ("'false' expected");
843
844			break;
845
846			case 'n':
847			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
848			{
849			dec->cur += 4;
850	root	1.5	return newSVsv (&PL_sv_undef);
851	root	1.1	}
852			else
853			ERR ("'null' expected");
854
855			break;
856
857			default:
858	root	1.7	ERR ("malformed json string, neither array, object, number, string or atom");
859	root	1.1	break;
860			}
861
862			fail:
863			return 0;
864			}
865
866			static SV *
867			decode_json (SV *string, UV flags)
868			{
869			SV *sv;
870
871	root	1.5	if (flags & F_UTF8)
872			sv_utf8_downgrade (string, 0);
873			else
874	root	1.1	sv_utf8_upgrade (string);
875
876			SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
877
878			dec_t dec;
879			dec.flags = flags;
880			dec.cur = SvPVX (string);
881			dec.end = SvEND (string);
882			dec.err = 0;
883
884			sv = decode_sv (&dec);
885
886			if (!sv)
887			{
888	root	1.7	IV offset = dec.flags & F_UTF8
889			? dec.cur - SvPVX (string)
890			: utf8_distance (dec.cur, SvPVX (string));
891	root	1.1	SV *uni = sv_newmortal ();
892	root	1.8
893	root	1.5	// horrible hack to silence warning inside pv_uni_display
894	root	1.8	COP cop = *PL_curcop;
895	root	1.5	cop.cop_warnings = pWARN_NONE;
896	root	1.8	ENTER;
897	root	1.5	SAVEVPTR (PL_curcop);
898			PL_curcop = &cop;
899	root	1.8	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
900			LEAVE;
901	root	1.1
902	root	1.5	croak ("%s, at character offset %d (%s)",
903	root	1.1	dec.err,
904			(int)offset,
905			dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
906			}
907
908	root	1.3	sv = sv_2mortal (sv);
909
910			if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
911	root	1.9	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
912	root	1.3
913			return sv;
914	root	1.1	}
915
916			MODULE = JSON::XS PACKAGE = JSON::XS
917
918			BOOT:
919			{
920			int i;
921
922			memset (decode_hexdigit, 0xff, 256);
923			for (i = 10; i--; )
924			decode_hexdigit ['0' + i] = i;
925
926	root	1.4	for (i = 7; i--; )
927	root	1.1	{
928			decode_hexdigit ['a' + i] = 10 + i;
929			decode_hexdigit ['A' + i] = 10 + i;
930			}
931
932			json_stash = gv_stashpv ("JSON::XS", 1);
933			}
934
935	root	1.4	PROTOTYPES: DISABLE
936
937	root	1.1	SV new (char dummy)
938			CODE:
939			RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
940			OUTPUT:
941			RETVAL
942
943	root	1.6	SV ascii (SV self, int enable = 1)
944	root	1.1	ALIAS:
945			ascii = F_ASCII
946			utf8 = F_UTF8
947			indent = F_INDENT
948			canonical = F_CANONICAL
949			space_before = F_SPACE_BEFORE
950			space_after = F_SPACE_AFTER
951			json_rpc = F_JSON_RPC
952	root	1.2	pretty = F_PRETTY
953	root	1.3	allow_nonref = F_ALLOW_NONREF
954	root	1.6	shrink = F_SHRINK
955	root	1.1	CODE:
956			{
957			UV *uv = SvJSON (self);
958			if (enable)
959			*uv \|= ix;
960			else
961			*uv &= ~ix;
962
963			RETVAL = newSVsv (self);
964			}
965			OUTPUT:
966			RETVAL
967
968			void encode (SV self, SV scalar)
969			PPCODE:
970			XPUSHs (encode_json (scalar, *SvJSON (self)));
971
972	root	1.2	void decode (SV self, SV jsonstr)
973	root	1.1	PPCODE:
974	root	1.2	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
975
976	root	1.4	PROTOTYPES: ENABLE
977
978	root	1.2	void to_json (SV *scalar)
979			PPCODE:
980			XPUSHs (encode_json (scalar, F_UTF8));
981
982			void from_json (SV *jsonstr)
983			PPCODE:
984			XPUSHs (decode_json (jsonstr, F_UTF8));
985	root	1.1