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 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, 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 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));

  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;
      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:	1.7
Committed:	Fri Mar 23 15:57:18 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
Changes since 1.6:	+18 -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			croak ("malformed UTF-8 character in string, cannot convert to JSON");
139			}
140			else
141			{
142			uch = ch;
143			clen = 1;
144			}
145
146			if (uch < 0x80 \|\| enc->flags & F_ASCII)
147			{
148			if (uch > 0xFFFFUL)
149			{
150			need (enc, len += 11);
151			sprintf (enc->cur, "\\u%04x\\u%04x",
152			(uch - 0x10000) / 0x400 + 0xD800,
153			(uch - 0x10000) % 0x400 + 0xDC00);
154			enc->cur += 12;
155			}
156			else
157			{
158			static char hexdigit [16] = "0123456789abcdef";
159			need (enc, len += 5);
160			*enc->cur++ = '\\';
161			*enc->cur++ = 'u';
162			*enc->cur++ = hexdigit [ uch >> 12 ];
163			*enc->cur++ = hexdigit [(uch >> 8) & 15];
164			*enc->cur++ = hexdigit [(uch >> 4) & 15];
165			*enc->cur++ = hexdigit [(uch >> 0) & 15];
166			}
167	root	1.4
168	root	1.6	str += clen;
169			}
170			else if (is_utf8)
171			{
172			need (enc, len += clen);
173			do
174			{
175			enc->cur++ = str++;
176			}
177			while (--clen);
178			}
179			else
180			{
181	root	1.7	need (enc, len += 10); // never more than 11 bytes needed
182	root	1.6	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183			++str;
184			}
185	root	1.5	}
186	root	1.4	}
187	root	1.1	}
188
189			--len;
190			}
191			}
192
193			#define INDENT SB \
194			if (enc->flags & F_INDENT) \
195			{ \
196			int i_; \
197			need (enc, enc->indent); \
198			for (i_ = enc->indent * 3; i_--; )\
199			encode_ch (enc, ' '); \
200			} \
201			SE
202
203	root	1.2	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE
204	root	1.1	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
205			#define COMMA SB \
206			encode_ch (enc, ','); \
207			if (enc->flags & F_INDENT) \
208			NL; \
209			else if (enc->flags & F_SPACE_AFTER) \
210			SPACE; \
211			SE
212
213			static void encode_sv (enc_t enc, SV sv);
214
215			static void
216			encode_av (enc_t enc, AV av)
217			{
218			int i, len = av_len (av);
219
220			encode_ch (enc, '['); NL;
221			++enc->indent;
222
223			for (i = 0; i <= len; ++i)
224			{
225			INDENT;
226			encode_sv (enc, *av_fetch (av, i, 0));
227
228			if (i < len)
229			COMMA;
230			}
231
232			NL;
233
234			--enc->indent;
235			INDENT; encode_ch (enc, ']');
236			}
237
238			static void
239			encode_he (enc_t enc, HE he)
240			{
241			encode_ch (enc, '"');
242
243			if (HeKLEN (he) == HEf_SVKEY)
244			{
245			SV *sv = HeSVKEY (he);
246			STRLEN len;
247	root	1.4	char *str;
248
249			SvGETMAGIC (sv);
250			str = SvPV (sv, len);
251	root	1.1
252			encode_str (enc, str, len, SvUTF8 (sv));
253			}
254			else
255			encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256
257			encode_ch (enc, '"');
258
259			if (enc->flags & F_SPACE_BEFORE) SPACE;
260			encode_ch (enc, ':');
261			if (enc->flags & F_SPACE_AFTER ) SPACE;
262			encode_sv (enc, HeVAL (he));
263			}
264
265			// compare hash entries, used when all keys are bytestrings
266			static int
267			he_cmp_fast (const void a_, const void b_)
268			{
269			int cmp;
270
271			HE a = (HE **)a_;
272			HE b = (HE **)b_;
273
274			STRLEN la = HeKLEN (a);
275			STRLEN lb = HeKLEN (b);
276
277			if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
278			cmp = la < lb ? -1 : la == lb ? 0 : 1;
279
280			return cmp;
281			}
282
283			// compare hash entries, used when some keys are sv's or utf-x
284			static int
285			he_cmp_slow (const void a, const void b)
286			{
287			return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));
288			}
289
290			static void
291			encode_hv (enc_t enc, HV hv)
292			{
293			int count, i;
294
295			encode_ch (enc, '{'); NL; ++enc->indent;
296
297			if ((count = hv_iterinit (hv)))
298			{
299			// for canonical output we have to sort by keys first
300			// actually, this is mostly due to the stupid so-called
301			// security workaround added somewhere in 5.8.x.
302			// that randomises hash orderings
303			if (enc->flags & F_CANONICAL)
304			{
305			HE he, hes [count];
306			int fast = 1;
307
308			i = 0;
309			while ((he = hv_iternext (hv)))
310			{
311			hes [i++] = he;
312			if (HeKLEN (he) < 0 \|\| HeKUTF8 (he))
313			fast = 0;
314			}
315
316			assert (i == count);
317
318			if (fast)
319			qsort (hes, count, sizeof (HE *), he_cmp_fast);
320			else
321			{
322			// hack to disable "use bytes"
323			COP *oldcop = PL_curcop, cop;
324			cop.op_private = 0;
325			PL_curcop = &cop;
326
327			SAVETMPS;
328			qsort (hes, count, sizeof (HE *), he_cmp_slow);
329			FREETMPS;
330
331			PL_curcop = oldcop;
332			}
333
334			for (i = 0; i < count; ++i)
335			{
336			INDENT;
337			encode_he (enc, hes [i]);
338
339			if (i < count - 1)
340			COMMA;
341			}
342
343			NL;
344			}
345			else
346			{
347			SV *sv;
348			HE *he = hv_iternext (hv);
349
350			for (;;)
351			{
352			INDENT;
353			encode_he (enc, he);
354
355			if (!(he = hv_iternext (hv)))
356			break;
357
358			COMMA;
359			}
360
361			NL;
362			}
363			}
364
365			--enc->indent; INDENT; encode_ch (enc, '}');
366			}
367
368			static void
369			encode_sv (enc_t enc, SV sv)
370			{
371	root	1.4	SvGETMAGIC (sv);
372
373	root	1.1	if (SvPOKp (sv))
374			{
375			STRLEN len;
376			char *str = SvPV (sv, len);
377			encode_ch (enc, '"');
378			encode_str (enc, str, len, SvUTF8 (sv));
379			encode_ch (enc, '"');
380			}
381			else if (SvNOKp (sv))
382			{
383			need (enc, NV_DIG + 32);
384			Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385			enc->cur += strlen (enc->cur);
386			}
387			else if (SvIOKp (sv))
388			{
389			need (enc, 64);
390			enc->cur +=
391			SvIsUV(sv)
392			? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
393			: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
394			}
395			else if (SvROK (sv))
396			{
397			if (!--enc->max_recurse)
398			croak ("data structure too deep (hit recursion limit)");
399
400			sv = SvRV (sv);
401
402			switch (SvTYPE (sv))
403			{
404			case SVt_PVAV: encode_av (enc, (AV *)sv); break;
405			case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
406
407			default:
408			croak ("JSON can only represent references to arrays or hashes");
409			}
410			}
411			else if (!SvOK (sv))
412			encode_str (enc, "null", 4, 0);
413			else
414			croak ("encountered perl type that JSON cannot handle");
415			}
416
417			static SV *
418			encode_json (SV *scalar, UV flags)
419			{
420	root	1.3	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421			croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
422
423	root	1.1	enc_t enc;
424			enc.flags = flags;
425			enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426			enc.cur = SvPVX (enc.sv);
427			enc.end = SvEND (enc.sv);
428			enc.max_recurse = 0;
429			enc.indent = 0;
430
431			SvPOK_only (enc.sv);
432			encode_sv (&enc, scalar);
433
434			if (!(flags & (F_ASCII \| F_UTF8)))
435			SvUTF8_on (enc.sv);
436
437			SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438	root	1.6
439			if (enc.flags & F_SHRINK)
440	root	1.7	shrink (enc.sv);
441
442	root	1.1	return enc.sv;
443			}
444
445			/////////////////////////////////////////////////////////////////////////////
446
447			#define WS \
448			for (;;) \
449			{ \
450			char ch = *dec->cur; \
451			if (ch > 0x20 \
452			\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
453			break; \
454			++dec->cur; \
455			}
456
457			#define ERR(reason) SB dec->err = reason; goto fail; SE
458			#define EXPECT_CH(ch) SB \
459			if (*dec->cur != ch) \
460			ERR (# ch " expected"); \
461			++dec->cur; \
462			SE
463
464			static SV decode_sv (dec_t dec);
465
466			static signed char decode_hexdigit[256];
467
468			static UV
469			decode_4hex (dec_t *dec)
470			{
471			signed char d1, d2, d3, d4;
472
473			d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
474			if (d1 < 0) ERR ("four hexadecimal digits expected");
475			d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
476			if (d2 < 0) ERR ("four hexadecimal digits expected");
477			d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
478			if (d3 < 0) ERR ("four hexadecimal digits expected");
479			d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
480			if (d4 < 0) ERR ("four hexadecimal digits expected");
481
482			dec->cur += 4;
483
484			return ((UV)d1) << 12
485			\| ((UV)d2) << 8
486			\| ((UV)d3) << 4
487			\| ((UV)d4);
488
489			fail:
490			return (UV)-1;
491			}
492
493	root	1.4	#define APPEND_GROW(n) SB \
494			if (cur + (n) >= end) \
495			{ \
496			STRLEN ofs = cur - SvPVX (sv); \
497			SvGROW (sv, ofs + (n) + 1); \
498			cur = SvPVX (sv) + ofs; \
499			end = SvEND (sv); \
500			} \
501			SE
502
503			#define APPEND_CH(ch) SB \
504			APPEND_GROW (1); \
505			*cur++ = (ch); \
506			SE
507
508	root	1.1	static SV *
509			decode_str (dec_t *dec)
510			{
511			SV *sv = NEWSV (0,2);
512			int utf8 = 0;
513	root	1.4	char *cur = SvPVX (sv);
514			char *end = SvEND (sv);
515	root	1.1
516			for (;;)
517			{
518			unsigned char ch = (unsigned char )dec->cur;
519
520			if (ch == '"')
521			break;
522			else if (ch == '\\')
523			{
524			switch (*++dec->cur)
525			{
526			case '\\':
527			case '/':
528			case '"': APPEND_CH (*dec->cur++); break;
529
530			case 'b': APPEND_CH ('\010'); ++dec->cur; break;
531			case 't': APPEND_CH ('\011'); ++dec->cur; break;
532			case 'n': APPEND_CH ('\012'); ++dec->cur; break;
533			case 'f': APPEND_CH ('\014'); ++dec->cur; break;
534			case 'r': APPEND_CH ('\015'); ++dec->cur; break;
535
536			case 'u':
537			{
538			UV lo, hi;
539			++dec->cur;
540
541			hi = decode_4hex (dec);
542			if (hi == (UV)-1)
543			goto fail;
544
545			// possibly a surrogate pair
546			if (hi >= 0xd800 && hi < 0xdc00)
547			{
548			if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
549	root	1.5	ERR ("missing low surrogate character in surrogate pair");
550	root	1.1
551			dec->cur += 2;
552
553			lo = decode_4hex (dec);
554			if (lo == (UV)-1)
555			goto fail;
556
557			if (lo < 0xdc00 \|\| lo >= 0xe000)
558			ERR ("surrogate pair expected");
559
560			hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
561			}
562	root	1.5	else if (hi >= 0xdc00 && hi < 0xe000)
563			ERR ("missing high surrogate character in surrogate pair");
564	root	1.1
565			if (hi >= 0x80)
566			{
567			utf8 = 1;
568
569	root	1.4	APPEND_GROW (4); // at most 4 bytes for 21 bits
570			cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	root	1.1	}
572			else
573			APPEND_CH (hi);
574			}
575			break;
576	root	1.5
577			default:
578			--dec->cur;
579			ERR ("illegal backslash escape sequence in string");
580	root	1.1	}
581			}
582			else if (ch >= 0x20 && ch <= 0x7f)
583			APPEND_CH (*dec->cur++);
584			else if (ch >= 0x80)
585			{
586			STRLEN clen;
587			UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
588	root	1.5	if (clen == (STRLEN)-1)
589	root	1.7	ERR ("malformed UTF-8 character in JSON string");
590	root	1.1
591	root	1.4	APPEND_GROW (clen);
592	root	1.5	do
593			{
594			cur++ = dec->cur++;
595			}
596			while (--clen);
597
598			utf8 = 1;
599	root	1.1	}
600	root	1.5	else if (dec->cur == dec->end)
601			ERR ("unexpected end of string while parsing json string");
602	root	1.1	else
603			ERR ("invalid character encountered");
604			}
605
606			++dec->cur;
607
608	root	1.4	SvCUR_set (sv, cur - SvPVX (sv));
609
610	root	1.1	SvPOK_only (sv);
611			*SvEND (sv) = 0;
612
613			if (utf8)
614			SvUTF8_on (sv);
615
616	root	1.6	if (dec->flags & F_SHRINK)
617	root	1.7	shrink (sv);
618	root	1.6
619	root	1.1	return sv;
620
621			fail:
622			SvREFCNT_dec (sv);
623			return 0;
624			}
625
626			static SV *
627			decode_num (dec_t *dec)
628			{
629			int is_nv = 0;
630			char *start = dec->cur;
631
632			// [minus]
633			if (*dec->cur == '-')
634			++dec->cur;
635
636			if (*dec->cur == '0')
637			{
638			++dec->cur;
639			if (dec->cur >= '0' && dec->cur <= '9')
640			ERR ("malformed number (leading zero must not be followed by another digit)");
641			}
642	root	1.5	else if (dec->cur < '0' \|\| dec->cur > '9')
643			ERR ("malformed number (no digits after initial minus)");
644			else
645			do
646			{
647			++dec->cur;
648			}
649			while (dec->cur >= '0' && dec->cur <= '9');
650	root	1.1
651			// [frac]
652			if (*dec->cur == '.')
653			{
654	root	1.5	++dec->cur;
655
656			if (dec->cur < '0' \|\| dec->cur > '9')
657			ERR ("malformed number (no digits after decimal point)");
658	root	1.1
659			do
660			{
661			++dec->cur;
662			}
663			while (dec->cur >= '0' && dec->cur <= '9');
664	root	1.5
665			is_nv = 1;
666	root	1.1	}
667
668			// [exp]
669			if (dec->cur == 'e' \|\| dec->cur == 'E')
670			{
671	root	1.5	++dec->cur;
672	root	1.1
673			if (dec->cur == '-' \|\| dec->cur == '+')
674			++dec->cur;
675
676	root	1.5	if (dec->cur < '0' \|\| dec->cur > '9')
677			ERR ("malformed number (no digits after exp sign)");
678
679			do
680			{
681			++dec->cur;
682			}
683			while (dec->cur >= '0' && dec->cur <= '9');
684
685			is_nv = 1;
686	root	1.1	}
687
688			if (!is_nv)
689			{
690			UV uv;
691			int numtype = grok_number (start, dec->cur - start, &uv);
692			if (numtype & IS_NUMBER_IN_UV)
693			if (numtype & IS_NUMBER_NEG)
694			{
695			if (uv < (UV)IV_MIN)
696			return newSViv (-(IV)uv);
697			}
698			else
699			return newSVuv (uv);
700			}
701
702			return newSVnv (Atof (start));
703
704			fail:
705			return 0;
706			}
707
708			static SV *
709			decode_av (dec_t *dec)
710			{
711			AV *av = newAV ();
712
713	root	1.5	WS;
714			if (*dec->cur == ']')
715			++dec->cur;
716			else
717			for (;;)
718			{
719			SV *value;
720	root	1.1
721	root	1.5	value = decode_sv (dec);
722			if (!value)
723			goto fail;
724	root	1.1
725	root	1.5	av_push (av, value);
726	root	1.1
727	root	1.5	WS;
728	root	1.1
729	root	1.5	if (*dec->cur == ']')
730			{
731			++dec->cur;
732			break;
733			}
734
735			if (*dec->cur != ',')
736			ERR (", or ] expected while parsing array");
737	root	1.1
738	root	1.5	++dec->cur;
739			}
740	root	1.1
741			return newRV_noinc ((SV *)av);
742
743			fail:
744			SvREFCNT_dec (av);
745			return 0;
746			}
747
748			static SV *
749			decode_hv (dec_t *dec)
750			{
751			HV *hv = newHV ();
752
753	root	1.5	WS;
754			if (*dec->cur == '}')
755			++dec->cur;
756			else
757			for (;;)
758			{
759			SV key, value;
760	root	1.1
761	root	1.5	WS; EXPECT_CH ('"');
762	root	1.1
763	root	1.5	key = decode_str (dec);
764			if (!key)
765			goto fail;
766	root	1.1
767	root	1.5	WS; EXPECT_CH (':');
768	root	1.1
769	root	1.5	value = decode_sv (dec);
770			if (!value)
771			{
772			SvREFCNT_dec (key);
773			goto fail;
774			}
775	root	1.1
776	root	1.5	//TODO: optimise
777			hv_store_ent (hv, key, value, 0);
778	root	1.1
779	root	1.5	WS;
780	root	1.1
781	root	1.5	if (*dec->cur == '}')
782			{
783			++dec->cur;
784			break;
785			}
786	root	1.1
787	root	1.5	if (*dec->cur != ',')
788			ERR (", or } expected while parsing object/hash");
789	root	1.1
790	root	1.5	++dec->cur;
791			}
792	root	1.1
793			return newRV_noinc ((SV *)hv);
794
795			fail:
796			SvREFCNT_dec (hv);
797			return 0;
798			}
799
800			static SV *
801			decode_sv (dec_t *dec)
802			{
803			WS;
804			switch (*dec->cur)
805			{
806			case '"': ++dec->cur; return decode_str (dec);
807			case '[': ++dec->cur; return decode_av (dec);
808			case '{': ++dec->cur; return decode_hv (dec);
809
810			case '-':
811			case '0': case '1': case '2': case '3': case '4':
812			case '5': case '6': case '7': case '8': case '9':
813			return decode_num (dec);
814
815			case 't':
816			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817			{
818			dec->cur += 4;
819			return newSViv (1);
820			}
821			else
822			ERR ("'true' expected");
823
824			break;
825
826			case 'f':
827			if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828			{
829			dec->cur += 5;
830			return newSViv (0);
831			}
832			else
833			ERR ("'false' expected");
834
835			break;
836
837			case 'n':
838			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
839			{
840			dec->cur += 4;
841	root	1.5	return newSVsv (&PL_sv_undef);
842	root	1.1	}
843			else
844			ERR ("'null' expected");
845
846			break;
847
848			default:
849	root	1.7	ERR ("malformed json string, neither array, object, number, string or atom");
850	root	1.1	break;
851			}
852
853			fail:
854			return 0;
855			}
856
857			static SV *
858			decode_json (SV *string, UV flags)
859			{
860			SV *sv;
861
862	root	1.5	if (flags & F_UTF8)
863			sv_utf8_downgrade (string, 0);
864			else
865	root	1.1	sv_utf8_upgrade (string);
866
867			SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868
869			dec_t dec;
870			dec.flags = flags;
871			dec.cur = SvPVX (string);
872			dec.end = SvEND (string);
873			dec.err = 0;
874
875			sv = decode_sv (&dec);
876
877			if (!sv)
878			{
879	root	1.7	IV offset = dec.flags & F_UTF8
880			? dec.cur - SvPVX (string)
881			: utf8_distance (dec.cur, SvPVX (string));
882	root	1.1	SV *uni = sv_newmortal ();
883	root	1.5	// horrible hack to silence warning inside pv_uni_display
884			COP cop;
885			memset (&cop, 0, sizeof (cop));
886			cop.cop_warnings = pWARN_NONE;
887			SAVEVPTR (PL_curcop);
888			PL_curcop = &cop;
889	root	1.1
890			pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
891	root	1.5	croak ("%s, at character offset %d (%s)",
892	root	1.1	dec.err,
893			(int)offset,
894			dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895			}
896
897	root	1.3	sv = sv_2mortal (sv);
898
899			if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
900			croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");
901
902			return sv;
903	root	1.1	}
904
905			MODULE = JSON::XS PACKAGE = JSON::XS
906
907			BOOT:
908			{
909			int i;
910
911			memset (decode_hexdigit, 0xff, 256);
912			for (i = 10; i--; )
913			decode_hexdigit ['0' + i] = i;
914
915	root	1.4	for (i = 7; i--; )
916	root	1.1	{
917			decode_hexdigit ['a' + i] = 10 + i;
918			decode_hexdigit ['A' + i] = 10 + i;
919			}
920
921			json_stash = gv_stashpv ("JSON::XS", 1);
922			}
923
924	root	1.4	PROTOTYPES: DISABLE
925
926	root	1.1	SV new (char dummy)
927			CODE:
928			RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
929			OUTPUT:
930			RETVAL
931
932	root	1.6	SV ascii (SV self, int enable = 1)
933	root	1.1	ALIAS:
934			ascii = F_ASCII
935			utf8 = F_UTF8
936			indent = F_INDENT
937			canonical = F_CANONICAL
938			space_before = F_SPACE_BEFORE
939			space_after = F_SPACE_AFTER
940			json_rpc = F_JSON_RPC
941	root	1.2	pretty = F_PRETTY
942	root	1.3	allow_nonref = F_ALLOW_NONREF
943	root	1.6	shrink = F_SHRINK
944	root	1.1	CODE:
945			{
946			UV *uv = SvJSON (self);
947			if (enable)
948			*uv \|= ix;
949			else
950			*uv &= ~ix;
951
952			RETVAL = newSVsv (self);
953			}
954			OUTPUT:
955			RETVAL
956
957			void encode (SV self, SV scalar)
958			PPCODE:
959			XPUSHs (encode_json (scalar, *SvJSON (self)));
960
961	root	1.2	void decode (SV self, SV jsonstr)
962	root	1.1	PPCODE:
963	root	1.2	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
964
965	root	1.4	PROTOTYPES: ENABLE
966
967	root	1.2	void to_json (SV *scalar)
968			PPCODE:
969			XPUSHs (encode_json (scalar, F_UTF8));
970
971			void from_json (SV *jsonstr)
972			PPCODE:
973			XPUSHs (decode_json (jsonstr, F_UTF8));
974	root	1.1