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 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))
    {
      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 = *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 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:	1.8
Committed:	Fri Mar 23 16:13:59 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
Changes since 1.7:	+14 -8 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	root	1.8	// hack to forcefully disable "use bytes"
323			COP cop = *PL_curcop;
324	root	1.1	cop.op_private = 0;
325	root	1.8
326			ENTER;
327			SAVETMPS;
328
329			SAVEVPTR (PL_curcop);
330	root	1.1	PL_curcop = &cop;
331
332			qsort (hes, count, sizeof (HE *), he_cmp_slow);
333	root	1.8
334	root	1.1	FREETMPS;
335	root	1.8	LEAVE;
336	root	1.1	}
337
338			for (i = 0; i < count; ++i)
339			{
340			INDENT;
341			encode_he (enc, hes [i]);
342
343			if (i < count - 1)
344			COMMA;
345			}
346
347			NL;
348			}
349			else
350			{
351			SV *sv;
352			HE *he = hv_iternext (hv);
353
354			for (;;)
355			{
356			INDENT;
357			encode_he (enc, he);
358
359			if (!(he = hv_iternext (hv)))
360			break;
361
362			COMMA;
363			}
364
365			NL;
366			}
367			}
368
369			--enc->indent; INDENT; encode_ch (enc, '}');
370			}
371
372			static void
373			encode_sv (enc_t enc, SV sv)
374			{
375	root	1.4	SvGETMAGIC (sv);
376
377	root	1.1	if (SvPOKp (sv))
378			{
379			STRLEN len;
380			char *str = SvPV (sv, len);
381			encode_ch (enc, '"');
382			encode_str (enc, str, len, SvUTF8 (sv));
383			encode_ch (enc, '"');
384			}
385			else if (SvNOKp (sv))
386			{
387			need (enc, NV_DIG + 32);
388			Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
389			enc->cur += strlen (enc->cur);
390			}
391			else if (SvIOKp (sv))
392			{
393			need (enc, 64);
394			enc->cur +=
395			SvIsUV(sv)
396			? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
397			: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
398			}
399			else if (SvROK (sv))
400			{
401			if (!--enc->max_recurse)
402			croak ("data structure too deep (hit recursion limit)");
403
404			sv = SvRV (sv);
405
406			switch (SvTYPE (sv))
407			{
408			case SVt_PVAV: encode_av (enc, (AV *)sv); break;
409			case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
410
411			default:
412			croak ("JSON can only represent references to arrays or hashes");
413			}
414			}
415			else if (!SvOK (sv))
416			encode_str (enc, "null", 4, 0);
417			else
418			croak ("encountered perl type that JSON cannot handle");
419			}
420
421			static SV *
422			encode_json (SV *scalar, UV flags)
423			{
424	root	1.3	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
425			croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
426
427	root	1.1	enc_t enc;
428			enc.flags = flags;
429			enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
430			enc.cur = SvPVX (enc.sv);
431			enc.end = SvEND (enc.sv);
432			enc.max_recurse = 0;
433			enc.indent = 0;
434
435			SvPOK_only (enc.sv);
436			encode_sv (&enc, scalar);
437
438			if (!(flags & (F_ASCII \| F_UTF8)))
439			SvUTF8_on (enc.sv);
440
441			SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
442	root	1.6
443			if (enc.flags & F_SHRINK)
444	root	1.7	shrink (enc.sv);
445
446	root	1.1	return enc.sv;
447			}
448
449			/////////////////////////////////////////////////////////////////////////////
450
451			#define WS \
452			for (;;) \
453			{ \
454			char ch = *dec->cur; \
455			if (ch > 0x20 \
456			\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
457			break; \
458			++dec->cur; \
459			}
460
461			#define ERR(reason) SB dec->err = reason; goto fail; SE
462			#define EXPECT_CH(ch) SB \
463			if (*dec->cur != ch) \
464			ERR (# ch " expected"); \
465			++dec->cur; \
466			SE
467
468			static SV decode_sv (dec_t dec);
469
470			static signed char decode_hexdigit[256];
471
472			static UV
473			decode_4hex (dec_t *dec)
474			{
475			signed char d1, d2, d3, d4;
476
477			d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
478			if (d1 < 0) ERR ("four hexadecimal digits expected");
479			d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
480			if (d2 < 0) ERR ("four hexadecimal digits expected");
481			d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
482			if (d3 < 0) ERR ("four hexadecimal digits expected");
483			d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
484			if (d4 < 0) ERR ("four hexadecimal digits expected");
485
486			dec->cur += 4;
487
488			return ((UV)d1) << 12
489			\| ((UV)d2) << 8
490			\| ((UV)d3) << 4
491			\| ((UV)d4);
492
493			fail:
494			return (UV)-1;
495			}
496
497	root	1.4	#define APPEND_GROW(n) SB \
498			if (cur + (n) >= end) \
499			{ \
500			STRLEN ofs = cur - SvPVX (sv); \
501			SvGROW (sv, ofs + (n) + 1); \
502			cur = SvPVX (sv) + ofs; \
503			end = SvEND (sv); \
504			} \
505			SE
506
507			#define APPEND_CH(ch) SB \
508			APPEND_GROW (1); \
509			*cur++ = (ch); \
510			SE
511
512	root	1.1	static SV *
513			decode_str (dec_t *dec)
514			{
515			SV *sv = NEWSV (0,2);
516			int utf8 = 0;
517	root	1.4	char *cur = SvPVX (sv);
518			char *end = SvEND (sv);
519	root	1.1
520			for (;;)
521			{
522			unsigned char ch = (unsigned char )dec->cur;
523
524			if (ch == '"')
525			break;
526			else if (ch == '\\')
527			{
528			switch (*++dec->cur)
529			{
530			case '\\':
531			case '/':
532			case '"': APPEND_CH (*dec->cur++); break;
533
534			case 'b': APPEND_CH ('\010'); ++dec->cur; break;
535			case 't': APPEND_CH ('\011'); ++dec->cur; break;
536			case 'n': APPEND_CH ('\012'); ++dec->cur; break;
537			case 'f': APPEND_CH ('\014'); ++dec->cur; break;
538			case 'r': APPEND_CH ('\015'); ++dec->cur; break;
539
540			case 'u':
541			{
542			UV lo, hi;
543			++dec->cur;
544
545			hi = decode_4hex (dec);
546			if (hi == (UV)-1)
547			goto fail;
548
549			// possibly a surrogate pair
550			if (hi >= 0xd800 && hi < 0xdc00)
551			{
552			if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
553	root	1.5	ERR ("missing low surrogate character in surrogate pair");
554	root	1.1
555			dec->cur += 2;
556
557			lo = decode_4hex (dec);
558			if (lo == (UV)-1)
559			goto fail;
560
561			if (lo < 0xdc00 \|\| lo >= 0xe000)
562			ERR ("surrogate pair expected");
563
564			hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
565			}
566	root	1.5	else if (hi >= 0xdc00 && hi < 0xe000)
567			ERR ("missing high surrogate character in surrogate pair");
568	root	1.1
569			if (hi >= 0x80)
570			{
571			utf8 = 1;
572
573	root	1.4	APPEND_GROW (4); // at most 4 bytes for 21 bits
574			cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
575	root	1.1	}
576			else
577			APPEND_CH (hi);
578			}
579			break;
580	root	1.5
581			default:
582			--dec->cur;
583			ERR ("illegal backslash escape sequence in string");
584	root	1.1	}
585			}
586			else if (ch >= 0x20 && ch <= 0x7f)
587			APPEND_CH (*dec->cur++);
588			else if (ch >= 0x80)
589			{
590			STRLEN clen;
591			UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
592	root	1.5	if (clen == (STRLEN)-1)
593	root	1.7	ERR ("malformed UTF-8 character in JSON string");
594	root	1.1
595	root	1.4	APPEND_GROW (clen);
596	root	1.5	do
597			{
598			cur++ = dec->cur++;
599			}
600			while (--clen);
601
602			utf8 = 1;
603	root	1.1	}
604	root	1.5	else if (dec->cur == dec->end)
605			ERR ("unexpected end of string while parsing json string");
606	root	1.1	else
607			ERR ("invalid character encountered");
608			}
609
610			++dec->cur;
611
612	root	1.4	SvCUR_set (sv, cur - SvPVX (sv));
613
614	root	1.1	SvPOK_only (sv);
615			*SvEND (sv) = 0;
616
617			if (utf8)
618			SvUTF8_on (sv);
619
620	root	1.6	if (dec->flags & F_SHRINK)
621	root	1.7	shrink (sv);
622	root	1.6
623	root	1.1	return sv;
624
625			fail:
626			SvREFCNT_dec (sv);
627			return 0;
628			}
629
630			static SV *
631			decode_num (dec_t *dec)
632			{
633			int is_nv = 0;
634			char *start = dec->cur;
635
636			// [minus]
637			if (*dec->cur == '-')
638			++dec->cur;
639
640			if (*dec->cur == '0')
641			{
642			++dec->cur;
643			if (dec->cur >= '0' && dec->cur <= '9')
644			ERR ("malformed number (leading zero must not be followed by another digit)");
645			}
646	root	1.5	else if (dec->cur < '0' \|\| dec->cur > '9')
647			ERR ("malformed number (no digits after initial minus)");
648			else
649			do
650			{
651			++dec->cur;
652			}
653			while (dec->cur >= '0' && dec->cur <= '9');
654	root	1.1
655			// [frac]
656			if (*dec->cur == '.')
657			{
658	root	1.5	++dec->cur;
659
660			if (dec->cur < '0' \|\| dec->cur > '9')
661			ERR ("malformed number (no digits after decimal point)");
662	root	1.1
663			do
664			{
665			++dec->cur;
666			}
667			while (dec->cur >= '0' && dec->cur <= '9');
668	root	1.5
669			is_nv = 1;
670	root	1.1	}
671
672			// [exp]
673			if (dec->cur == 'e' \|\| dec->cur == 'E')
674			{
675	root	1.5	++dec->cur;
676	root	1.1
677			if (dec->cur == '-' \|\| dec->cur == '+')
678			++dec->cur;
679
680	root	1.5	if (dec->cur < '0' \|\| dec->cur > '9')
681			ERR ("malformed number (no digits after exp sign)");
682
683			do
684			{
685			++dec->cur;
686			}
687			while (dec->cur >= '0' && dec->cur <= '9');
688
689			is_nv = 1;
690	root	1.1	}
691
692			if (!is_nv)
693			{
694			UV uv;
695			int numtype = grok_number (start, dec->cur - start, &uv);
696			if (numtype & IS_NUMBER_IN_UV)
697			if (numtype & IS_NUMBER_NEG)
698			{
699			if (uv < (UV)IV_MIN)
700			return newSViv (-(IV)uv);
701			}
702			else
703			return newSVuv (uv);
704			}
705
706			return newSVnv (Atof (start));
707
708			fail:
709			return 0;
710			}
711
712			static SV *
713			decode_av (dec_t *dec)
714			{
715			AV *av = newAV ();
716
717	root	1.5	WS;
718			if (*dec->cur == ']')
719			++dec->cur;
720			else
721			for (;;)
722			{
723			SV *value;
724	root	1.1
725	root	1.5	value = decode_sv (dec);
726			if (!value)
727			goto fail;
728	root	1.1
729	root	1.5	av_push (av, value);
730	root	1.1
731	root	1.5	WS;
732	root	1.1
733	root	1.5	if (*dec->cur == ']')
734			{
735			++dec->cur;
736			break;
737			}
738
739			if (*dec->cur != ',')
740			ERR (", or ] expected while parsing array");
741	root	1.1
742	root	1.5	++dec->cur;
743			}
744	root	1.1
745			return newRV_noinc ((SV *)av);
746
747			fail:
748			SvREFCNT_dec (av);
749			return 0;
750			}
751
752			static SV *
753			decode_hv (dec_t *dec)
754			{
755			HV *hv = newHV ();
756
757	root	1.5	WS;
758			if (*dec->cur == '}')
759			++dec->cur;
760			else
761			for (;;)
762			{
763			SV key, value;
764	root	1.1
765	root	1.5	WS; EXPECT_CH ('"');
766	root	1.1
767	root	1.5	key = decode_str (dec);
768			if (!key)
769			goto fail;
770	root	1.1
771	root	1.5	WS; EXPECT_CH (':');
772	root	1.1
773	root	1.5	value = decode_sv (dec);
774			if (!value)
775			{
776			SvREFCNT_dec (key);
777			goto fail;
778			}
779	root	1.1
780	root	1.5	//TODO: optimise
781			hv_store_ent (hv, key, value, 0);
782	root	1.1
783	root	1.5	WS;
784	root	1.1
785	root	1.5	if (*dec->cur == '}')
786			{
787			++dec->cur;
788			break;
789			}
790	root	1.1
791	root	1.5	if (*dec->cur != ',')
792			ERR (", or } expected while parsing object/hash");
793	root	1.1
794	root	1.5	++dec->cur;
795			}
796	root	1.1
797			return newRV_noinc ((SV *)hv);
798
799			fail:
800			SvREFCNT_dec (hv);
801			return 0;
802			}
803
804			static SV *
805			decode_sv (dec_t *dec)
806			{
807			WS;
808			switch (*dec->cur)
809			{
810			case '"': ++dec->cur; return decode_str (dec);
811			case '[': ++dec->cur; return decode_av (dec);
812			case '{': ++dec->cur; return decode_hv (dec);
813
814			case '-':
815			case '0': case '1': case '2': case '3': case '4':
816			case '5': case '6': case '7': case '8': case '9':
817			return decode_num (dec);
818
819			case 't':
820			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
821			{
822			dec->cur += 4;
823			return newSViv (1);
824			}
825			else
826			ERR ("'true' expected");
827
828			break;
829
830			case 'f':
831			if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
832			{
833			dec->cur += 5;
834			return newSViv (0);
835			}
836			else
837			ERR ("'false' expected");
838
839			break;
840
841			case 'n':
842			if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
843			{
844			dec->cur += 4;
845	root	1.5	return newSVsv (&PL_sv_undef);
846	root	1.1	}
847			else
848			ERR ("'null' expected");
849
850			break;
851
852			default:
853	root	1.7	ERR ("malformed json string, neither array, object, number, string or atom");
854	root	1.1	break;
855			}
856
857			fail:
858			return 0;
859			}
860
861			static SV *
862			decode_json (SV *string, UV flags)
863			{
864			SV *sv;
865
866	root	1.5	if (flags & F_UTF8)
867			sv_utf8_downgrade (string, 0);
868			else
869	root	1.1	sv_utf8_upgrade (string);
870
871			SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
872
873			dec_t dec;
874			dec.flags = flags;
875			dec.cur = SvPVX (string);
876			dec.end = SvEND (string);
877			dec.err = 0;
878
879			sv = decode_sv (&dec);
880
881			if (!sv)
882			{
883	root	1.7	IV offset = dec.flags & F_UTF8
884			? dec.cur - SvPVX (string)
885			: utf8_distance (dec.cur, SvPVX (string));
886	root	1.1	SV *uni = sv_newmortal ();
887	root	1.8
888	root	1.5	// horrible hack to silence warning inside pv_uni_display
889	root	1.8	COP cop = *PL_curcop;
890	root	1.5	cop.cop_warnings = pWARN_NONE;
891	root	1.8	ENTER;
892	root	1.5	SAVEVPTR (PL_curcop);
893			PL_curcop = &cop;
894	root	1.8	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
895			LEAVE;
896	root	1.1
897	root	1.5	croak ("%s, at character offset %d (%s)",
898	root	1.1	dec.err,
899			(int)offset,
900			dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
901			}
902
903	root	1.3	sv = sv_2mortal (sv);
904
905			if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
906			croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");
907
908			return sv;
909	root	1.1	}
910
911			MODULE = JSON::XS PACKAGE = JSON::XS
912
913			BOOT:
914			{
915			int i;
916
917			memset (decode_hexdigit, 0xff, 256);
918			for (i = 10; i--; )
919			decode_hexdigit ['0' + i] = i;
920
921	root	1.4	for (i = 7; i--; )
922	root	1.1	{
923			decode_hexdigit ['a' + i] = 10 + i;
924			decode_hexdigit ['A' + i] = 10 + i;
925			}
926
927			json_stash = gv_stashpv ("JSON::XS", 1);
928			}
929
930	root	1.4	PROTOTYPES: DISABLE
931
932	root	1.1	SV new (char dummy)
933			CODE:
934			RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
935			OUTPUT:
936			RETVAL
937
938	root	1.6	SV ascii (SV self, int enable = 1)
939	root	1.1	ALIAS:
940			ascii = F_ASCII
941			utf8 = F_UTF8
942			indent = F_INDENT
943			canonical = F_CANONICAL
944			space_before = F_SPACE_BEFORE
945			space_after = F_SPACE_AFTER
946			json_rpc = F_JSON_RPC
947	root	1.2	pretty = F_PRETTY
948	root	1.3	allow_nonref = F_ALLOW_NONREF
949	root	1.6	shrink = F_SHRINK
950	root	1.1	CODE:
951			{
952			UV *uv = SvJSON (self);
953			if (enable)
954			*uv \|= ix;
955			else
956			*uv &= ~ix;
957
958			RETVAL = newSVsv (self);
959			}
960			OUTPUT:
961			RETVAL
962
963			void encode (SV self, SV scalar)
964			PPCODE:
965			XPUSHs (encode_json (scalar, *SvJSON (self)));
966
967	root	1.2	void decode (SV self, SV jsonstr)
968	root	1.1	PPCODE:
969	root	1.2	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
970
971	root	1.4	PROTOTYPES: ENABLE
972
973	root	1.2	void to_json (SV *scalar)
974			PPCODE:
975			XPUSHs (encode_json (scalar, F_UTF8));
976
977			void from_json (SV *jsonstr)
978			PPCODE:
979			XPUSHs (decode_json (jsonstr, F_UTF8));
980	root	1.1