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_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;
  char *err;
  UV flags;
} dec_t;

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

  return &SvUVX (SvRV (sv));
}

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

static void
need (enc_t *enc, STRLEN len)
{
  if (enc->cur + len >= enc->end)
    {
      STRLEN cur = enc->cur - SvPVX (enc->sv);
      SvGROW (enc->sv, cur + len + 1);
      enc->cur = SvPVX (enc->sv) + cur;
      enc->end = SvEND (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;

  while (str < end)
    {
      unsigned char ch = *(unsigned char *)str;
      if (ch >= 0x20 && ch < 0x80) // most common case
        {
          *enc->cur++ = ch;
          str++;
        }
      else
        {
          STRLEN clen;
          UV uch;

          if (is_utf8)
            {
              uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
              if (clen < 0)
                croak ("malformed UTF-8 character in string, cannot convert to JSON");
            }
          else
            {
              uch = ch;
              clen = 1;
            }

          need (enc, len += 6);

          if (uch < 0xa0 || enc->flags & F_ASCII)
            {
              if (uch > 0xFFFFUL)
                {
                  len += 6;
                  need (enc, len += 6);
                  sprintf (enc->cur, "\\u%04x\\u%04x",
                           (uch - 0x10000) / 0x400 + 0xD800,
                           (uch - 0x10000) % 0x400 + 0xDC00);
                  enc->cur += 12;
                }
              else
                {
                  sprintf (enc->cur, "\\u%04x", uch);
                  enc->cur += 6;
                }
            }
          else if (is_utf8)
            {
              memcpy (enc->cur, str, clen);
              enc->cur += clen;
            }
          else
            enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);

          str += clen;
        }

      --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 if (enc->flags & F_INDENT) { 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 = 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)
{
  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)
{
  enc_t enc;
  enc.flags       = flags;
  enc.sv          = sv_2mortal (NEWSV (0, INIT_SIZE));
  enc.cur         = SvPVX (enc.sv);
  enc.end         = SvEND (enc.sv);
  enc.max_recurse = 0;
  enc.indent      = 0;

  SvPOK_only (enc.sv);
  encode_sv (&enc, scalar);

  if (!(flags & (F_ASCII | F_UTF8)))
    SvUTF8_on (enc.sv);

  SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
  return enc.sv;
}

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

#define WS \
  for (;;)                              \
    {                                   \
      char ch = *dec->cur;              \
      if (ch > 0x20                     \
          || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
        break;                          \
      ++dec->cur;                       \
    }

#define ERR(reason) SB dec->err = reason; goto fail; SE
#define EXPECT_CH(ch) SB \
  if (*dec->cur != ch)          \
    ERR (# ch " expected");     \
  ++dec->cur;                   \
  SE

static SV *decode_sv (dec_t *dec);

#define APPEND_CH(ch) SB \
  SvGROW (sv, cur + 1 + 1);     \
  SvPVX (sv)[cur++] = (ch);     \
  SE

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;
}

static SV *
decode_str (dec_t *dec)
{
  SV *sv = NEWSV (0,2);
  STRLEN cur = 0;
  int utf8 = 0;

  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 ("illegal surrogate character");

                      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 (lo >= 0xdc00 && lo < 0xe000)
                    ERR ("illegal surrogate character");

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

                      SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
                      cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
                    }
                  else
                    APPEND_CH (hi);
                }
                break;
            }
        }
      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 < 0)
            ERR ("malformed UTF-8 character in string, cannot convert to JSON");

          SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
          memcpy (SvPVX (sv) + cur, dec->cur, clen);
          dec->cur += clen;
        }
      else
        ERR ("invalid character encountered");
    }

  ++dec->cur;

  SvPOK_only (sv);

  SvCUR_set (sv, cur);
  *SvEND (sv) = 0;

  if (utf8)
    SvUTF8_on (sv);

  return sv;

fail:
  SvREFCNT_dec (sv);
  return 0;
}

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

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

  if (*dec->cur == '0')
    {
      ++dec->cur;
      if (*dec->cur >= '0' && *dec->cur <= '9')
         ERR ("malformed number (leading zero must not be followed by another digit)");
    }

  // int
  while (*dec->cur >= '0' && *dec->cur <= '9')
    ++dec->cur;

  // [frac]
  if (*dec->cur == '.')
    {
      is_nv = 1;

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

  // [exp]
  if (*dec->cur == 'e' || *dec->cur == 'E')
    {
      is_nv = 1;

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

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

  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 ();

  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 ();

  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 newSViv (1);
          }
        else
          ERR ("'null' expected");

        break;

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

fail:
  return 0;
}

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

  if (!(flags & F_UTF8))
    sv_utf8_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;

  *dec.end = 1; // invalid anywhere
  sv = decode_sv (&dec);
  *dec.end = 0;

  if (!sv)
    {
      IV offset = utf8_distance (dec.cur, SvPVX (string));
      SV *uni = sv_newmortal ();

      pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
      croak ("%s, at character %d (%s)",
             dec.err,
             (int)offset,
             dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
    }

  sv_dump (sv);//D
  return sv_2mortal (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 = 6; --i; )
          {
            decode_hexdigit ['a' + i] = 10 + i;
            decode_hexdigit ['A' + i] = 10 + i;
          }

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

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

SV *ascii (SV *self, int enable)
        ALIAS:
        ascii        = F_ASCII
        utf8         = F_UTF8
        indent       = F_INDENT
        canonical    = F_CANONICAL
        space_before = F_SPACE_BEFORE
        space_after  = F_SPACE_AFTER
        json_rpc     = F_JSON_RPC
        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 *jsondata)
        PPCODE:
        XPUSHs (decode_json (jsondata, *SvJSON (self)));

Revision:	1.1
Committed:	Thu Mar 22 16:40:16 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
Log Message:	initial check-in
#	Content
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
17	#define F_DEFAULT 0
18
19	#define INIT_SIZE 32 // initial scalar size to be allocated
20
21	#define SB do {
22	#define SE } while (0)
23
24	static HV *json_stash;
25
26	// structure used for encoding JSON
27	typedef struct
28	{
29	char *cur;
30	STRLEN len; // SvLEN (sv)
31	char *end; // SvEND (sv)
32	SV *sv;
33	UV flags;
34	int max_recurse;
35	int indent;
36	} enc_t;
37
38	// structure used for decoding JSON
39	typedef struct
40	{
41	char *cur;
42	char *end;
43	char *err;
44	UV flags;
45	} dec_t;
46
47	static UV *
48	SvJSON (SV *sv)
49	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
51	croak ("object is not of type JSON::XS");
52
53	return &SvUVX (SvRV (sv));
54	}
55
56	/////////////////////////////////////////////////////////////////////////////
57
58	static void
59	need (enc_t *enc, STRLEN len)
60	{
61	if (enc->cur + len >= enc->end)
62	{
63	STRLEN cur = enc->cur - SvPVX (enc->sv);
64	SvGROW (enc->sv, cur + len + 1);
65	enc->cur = SvPVX (enc->sv) + cur;
66	enc->end = SvEND (enc->sv);
67	}
68	}
69
70	static void
71	encode_ch (enc_t *enc, char ch)
72	{
73	need (enc, 1);
74	*enc->cur++ = ch;
75	}
76
77	static void
78	encode_str (enc_t enc, char str, STRLEN len, int is_utf8)
79	{
80	char *end = str + len;
81
82	while (str < end)
83	{
84	unsigned char ch = (unsigned char )str;
85	if (ch >= 0x20 && ch < 0x80) // most common case
86	{
87	*enc->cur++ = ch;
88	str++;
89	}
90	else
91	{
92	STRLEN clen;
93	UV uch;
94
95	if (is_utf8)
96	{
97	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
98	if (clen < 0)
99	croak ("malformed UTF-8 character in string, cannot convert to JSON");
100	}
101	else
102	{
103	uch = ch;
104	clen = 1;
105	}
106
107	need (enc, len += 6);
108
109	if (uch < 0xa0 \|\| enc->flags & F_ASCII)
110	{
111	if (uch > 0xFFFFUL)
112	{
113	len += 6;
114	need (enc, len += 6);
115	sprintf (enc->cur, "\\u%04x\\u%04x",
116	(uch - 0x10000) / 0x400 + 0xD800,
117	(uch - 0x10000) % 0x400 + 0xDC00);
118	enc->cur += 12;
119	}
120	else
121	{
122	sprintf (enc->cur, "\\u%04x", uch);
123	enc->cur += 6;
124	}
125	}
126	else if (is_utf8)
127	{
128	memcpy (enc->cur, str, clen);
129	enc->cur += clen;
130	}
131	else
132	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
133
134	str += clen;
135	}
136
137	--len;
138	}
139	}
140
141	#define INDENT SB \
142	if (enc->flags & F_INDENT) \
143	{ \
144	int i_; \
145	need (enc, enc->indent); \
146	for (i_ = enc->indent * 3; i_--; )\
147	encode_ch (enc, ' '); \
148	} \
149	SE
150
151	#define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE
152	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
153	#define COMMA SB \
154	encode_ch (enc, ','); \
155	if (enc->flags & F_INDENT) \
156	NL; \
157	else if (enc->flags & F_SPACE_AFTER) \
158	SPACE; \
159	SE
160
161	static void encode_sv (enc_t enc, SV sv);
162
163	static void
164	encode_av (enc_t enc, AV av)
165	{
166	int i, len = av_len (av);
167
168	encode_ch (enc, '['); NL;
169	++enc->indent;
170
171	for (i = 0; i <= len; ++i)
172	{
173	INDENT;
174	encode_sv (enc, *av_fetch (av, i, 0));
175
176	if (i < len)
177	COMMA;
178	}
179
180	NL;
181
182	--enc->indent;
183	INDENT; encode_ch (enc, ']');
184	}
185
186	static void
187	encode_he (enc_t enc, HE he)
188	{
189	encode_ch (enc, '"');
190
191	if (HeKLEN (he) == HEf_SVKEY)
192	{
193	SV *sv = HeSVKEY (he);
194	STRLEN len;
195	char *str = SvPV (sv, len);
196
197	encode_str (enc, str, len, SvUTF8 (sv));
198	}
199	else
200	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
201
202	encode_ch (enc, '"');
203
204	if (enc->flags & F_SPACE_BEFORE) SPACE;
205	encode_ch (enc, ':');
206	if (enc->flags & F_SPACE_AFTER ) SPACE;
207	encode_sv (enc, HeVAL (he));
208	}
209
210	// compare hash entries, used when all keys are bytestrings
211	static int
212	he_cmp_fast (const void a_, const void b_)
213	{
214	int cmp;
215
216	HE a = (HE **)a_;
217	HE b = (HE **)b_;
218
219	STRLEN la = HeKLEN (a);
220	STRLEN lb = HeKLEN (b);
221
222	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
223	cmp = la < lb ? -1 : la == lb ? 0 : 1;
224
225	return cmp;
226	}
227
228	// compare hash entries, used when some keys are sv's or utf-x
229	static int
230	he_cmp_slow (const void a, const void b)
231	{
232	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));
233	}
234
235	static void
236	encode_hv (enc_t enc, HV hv)
237	{
238	int count, i;
239
240	encode_ch (enc, '{'); NL; ++enc->indent;
241
242	if ((count = hv_iterinit (hv)))
243	{
244	// for canonical output we have to sort by keys first
245	// actually, this is mostly due to the stupid so-called
246	// security workaround added somewhere in 5.8.x.
247	// that randomises hash orderings
248	if (enc->flags & F_CANONICAL)
249	{
250	HE he, hes [count];
251	int fast = 1;
252
253	i = 0;
254	while ((he = hv_iternext (hv)))
255	{
256	hes [i++] = he;
257	if (HeKLEN (he) < 0 \|\| HeKUTF8 (he))
258	fast = 0;
259	}
260
261	assert (i == count);
262
263	if (fast)
264	qsort (hes, count, sizeof (HE *), he_cmp_fast);
265	else
266	{
267	// hack to disable "use bytes"
268	COP *oldcop = PL_curcop, cop;
269	cop.op_private = 0;
270	PL_curcop = &cop;
271
272	SAVETMPS;
273	qsort (hes, count, sizeof (HE *), he_cmp_slow);
274	FREETMPS;
275
276	PL_curcop = oldcop;
277	}
278
279	for (i = 0; i < count; ++i)
280	{
281	INDENT;
282	encode_he (enc, hes [i]);
283
284	if (i < count - 1)
285	COMMA;
286	}
287
288	NL;
289	}
290	else
291	{
292	SV *sv;
293	HE *he = hv_iternext (hv);
294
295	for (;;)
296	{
297	INDENT;
298	encode_he (enc, he);
299
300	if (!(he = hv_iternext (hv)))
301	break;
302
303	COMMA;
304	}
305
306	NL;
307	}
308	}
309
310	--enc->indent; INDENT; encode_ch (enc, '}');
311	}
312
313	static void
314	encode_sv (enc_t enc, SV sv)
315	{
316	if (SvPOKp (sv))
317	{
318	STRLEN len;
319	char *str = SvPV (sv, len);
320	encode_ch (enc, '"');
321	encode_str (enc, str, len, SvUTF8 (sv));
322	encode_ch (enc, '"');
323	}
324	else if (SvNOKp (sv))
325	{
326	need (enc, NV_DIG + 32);
327	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
328	enc->cur += strlen (enc->cur);
329	}
330	else if (SvIOKp (sv))
331	{
332	need (enc, 64);
333	enc->cur +=
334	SvIsUV(sv)
335	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
336	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
337	}
338	else if (SvROK (sv))
339	{
340	if (!--enc->max_recurse)
341	croak ("data structure too deep (hit recursion limit)");
342
343	sv = SvRV (sv);
344
345	switch (SvTYPE (sv))
346	{
347	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
348	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
349
350	default:
351	croak ("JSON can only represent references to arrays or hashes");
352	}
353	}
354	else if (!SvOK (sv))
355	encode_str (enc, "null", 4, 0);
356	else
357	croak ("encountered perl type that JSON cannot handle");
358	}
359
360	static SV *
361	encode_json (SV *scalar, UV flags)
362	{
363	enc_t enc;
364	enc.flags = flags;
365	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
366	enc.cur = SvPVX (enc.sv);
367	enc.end = SvEND (enc.sv);
368	enc.max_recurse = 0;
369	enc.indent = 0;
370
371	SvPOK_only (enc.sv);
372	encode_sv (&enc, scalar);
373
374	if (!(flags & (F_ASCII \| F_UTF8)))
375	SvUTF8_on (enc.sv);
376
377	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
378	return enc.sv;
379	}
380
381	/////////////////////////////////////////////////////////////////////////////
382
383	#define WS \
384	for (;;) \
385	{ \
386	char ch = *dec->cur; \
387	if (ch > 0x20 \
388	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
389	break; \
390	++dec->cur; \
391	}
392
393	#define ERR(reason) SB dec->err = reason; goto fail; SE
394	#define EXPECT_CH(ch) SB \
395	if (*dec->cur != ch) \
396	ERR (# ch " expected"); \
397	++dec->cur; \
398	SE
399
400	static SV decode_sv (dec_t dec);
401
402	#define APPEND_CH(ch) SB \
403	SvGROW (sv, cur + 1 + 1); \
404	SvPVX (sv)[cur++] = (ch); \
405	SE
406
407	static signed char decode_hexdigit[256];
408
409	static UV
410	decode_4hex (dec_t *dec)
411	{
412	signed char d1, d2, d3, d4;
413
414	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
415	if (d1 < 0) ERR ("four hexadecimal digits expected");
416	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
417	if (d2 < 0) ERR ("four hexadecimal digits expected");
418	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
419	if (d3 < 0) ERR ("four hexadecimal digits expected");
420	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
421	if (d4 < 0) ERR ("four hexadecimal digits expected");
422
423	dec->cur += 4;
424
425	return ((UV)d1) << 12
426	\| ((UV)d2) << 8
427	\| ((UV)d3) << 4
428	\| ((UV)d4);
429
430	fail:
431	return (UV)-1;
432	}
433
434	static SV *
435	decode_str (dec_t *dec)
436	{
437	SV *sv = NEWSV (0,2);
438	STRLEN cur = 0;
439	int utf8 = 0;
440
441	for (;;)
442	{
443	unsigned char ch = (unsigned char )dec->cur;
444
445	if (ch == '"')
446	break;
447	else if (ch == '\\')
448	{
449	switch (*++dec->cur)
450	{
451	case '\\':
452	case '/':
453	case '"': APPEND_CH (*dec->cur++); break;
454
455	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
456	case 't': APPEND_CH ('\011'); ++dec->cur; break;
457	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
458	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
459	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
460
461	case 'u':
462	{
463	UV lo, hi;
464	++dec->cur;
465
466	hi = decode_4hex (dec);
467	if (hi == (UV)-1)
468	goto fail;
469
470	// possibly a surrogate pair
471	if (hi >= 0xd800 && hi < 0xdc00)
472	{
473	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
474	ERR ("illegal surrogate character");
475
476	dec->cur += 2;
477
478	lo = decode_4hex (dec);
479	if (lo == (UV)-1)
480	goto fail;
481
482	if (lo < 0xdc00 \|\| lo >= 0xe000)
483	ERR ("surrogate pair expected");
484
485	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
486	}
487	else if (lo >= 0xdc00 && lo < 0xe000)
488	ERR ("illegal surrogate character");
489
490	if (hi >= 0x80)
491	{
492	utf8 = 1;
493
494	SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
495	cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
496	}
497	else
498	APPEND_CH (hi);
499	}
500	break;
501	}
502	}
503	else if (ch >= 0x20 && ch <= 0x7f)
504	APPEND_CH (*dec->cur++);
505	else if (ch >= 0x80)
506	{
507	STRLEN clen;
508	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
509	if (clen < 0)
510	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
511
512	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
513	memcpy (SvPVX (sv) + cur, dec->cur, clen);
514	dec->cur += clen;
515	}
516	else
517	ERR ("invalid character encountered");
518	}
519
520	++dec->cur;
521
522	SvPOK_only (sv);
523
524	SvCUR_set (sv, cur);
525	*SvEND (sv) = 0;
526
527	if (utf8)
528	SvUTF8_on (sv);
529
530	return sv;
531
532	fail:
533	SvREFCNT_dec (sv);
534	return 0;
535	}
536
537	static SV *
538	decode_num (dec_t *dec)
539	{
540	int is_nv = 0;
541	char *start = dec->cur;
542
543	// [minus]
544	if (*dec->cur == '-')
545	++dec->cur;
546
547	if (*dec->cur == '0')
548	{
549	++dec->cur;
550	if (dec->cur >= '0' && dec->cur <= '9')
551	ERR ("malformed number (leading zero must not be followed by another digit)");
552	}
553
554	// int
555	while (dec->cur >= '0' && dec->cur <= '9')
556	++dec->cur;
557
558	// [frac]
559	if (*dec->cur == '.')
560	{
561	is_nv = 1;
562
563	do
564	{
565	++dec->cur;
566	}
567	while (dec->cur >= '0' && dec->cur <= '9');
568	}
569
570	// [exp]
571	if (dec->cur == 'e' \|\| dec->cur == 'E')
572	{
573	is_nv = 1;
574
575	++dec->cur;
576	if (dec->cur == '-' \|\| dec->cur == '+')
577	++dec->cur;
578
579	while (dec->cur >= '0' && dec->cur <= '9')
580	++dec->cur;
581	}
582
583	if (!is_nv)
584	{
585	UV uv;
586	int numtype = grok_number (start, dec->cur - start, &uv);
587	if (numtype & IS_NUMBER_IN_UV)
588	if (numtype & IS_NUMBER_NEG)
589	{
590	if (uv < (UV)IV_MIN)
591	return newSViv (-(IV)uv);
592	}
593	else
594	return newSVuv (uv);
595	}
596
597	return newSVnv (Atof (start));
598
599	fail:
600	return 0;
601	}
602
603	static SV *
604	decode_av (dec_t *dec)
605	{
606	AV *av = newAV ();
607
608	for (;;)
609	{
610	SV *value;
611
612	value = decode_sv (dec);
613	if (!value)
614	goto fail;
615
616	av_push (av, value);
617
618	WS;
619
620	if (*dec->cur == ']')
621	{
622	++dec->cur;
623	break;
624	}
625
626	if (*dec->cur != ',')
627	ERR (", or ] expected while parsing array");
628
629	++dec->cur;
630	}
631
632	return newRV_noinc ((SV *)av);
633
634	fail:
635	SvREFCNT_dec (av);
636	return 0;
637	}
638
639	static SV *
640	decode_hv (dec_t *dec)
641	{
642	HV *hv = newHV ();
643
644	for (;;)
645	{
646	SV key, value;
647
648	WS; EXPECT_CH ('"');
649
650	key = decode_str (dec);
651	if (!key)
652	goto fail;
653
654	WS; EXPECT_CH (':');
655
656	value = decode_sv (dec);
657	if (!value)
658	{
659	SvREFCNT_dec (key);
660	goto fail;
661	}
662
663	//TODO: optimise
664	hv_store_ent (hv, key, value, 0);
665
666	WS;
667
668	if (*dec->cur == '}')
669	{
670	++dec->cur;
671	break;
672	}
673
674	if (*dec->cur != ',')
675	ERR (", or } expected while parsing object/hash");
676
677	++dec->cur;
678	}
679
680	return newRV_noinc ((SV *)hv);
681
682	fail:
683	SvREFCNT_dec (hv);
684	return 0;
685	}
686
687	static SV *
688	decode_sv (dec_t *dec)
689	{
690	WS;
691	switch (*dec->cur)
692	{
693	case '"': ++dec->cur; return decode_str (dec);
694	case '[': ++dec->cur; return decode_av (dec);
695	case '{': ++dec->cur; return decode_hv (dec);
696
697	case '-':
698	case '0': case '1': case '2': case '3': case '4':
699	case '5': case '6': case '7': case '8': case '9':
700	return decode_num (dec);
701
702	case 't':
703	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
704	{
705	dec->cur += 4;
706	return newSViv (1);
707	}
708	else
709	ERR ("'true' expected");
710
711	break;
712
713	case 'f':
714	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
715	{
716	dec->cur += 5;
717	return newSViv (0);
718	}
719	else
720	ERR ("'false' expected");
721
722	break;
723
724	case 'n':
725	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
726	{
727	dec->cur += 4;
728	return newSViv (1);
729	}
730	else
731	ERR ("'null' expected");
732
733	break;
734
735	default:
736	ERR ("malformed json string");
737	break;
738	}
739
740	fail:
741	return 0;
742	}
743
744	static SV *
745	decode_json (SV *string, UV flags)
746	{
747	SV *sv;
748
749	if (!(flags & F_UTF8))
750	sv_utf8_upgrade (string);
751
752	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
753
754	dec_t dec;
755	dec.flags = flags;
756	dec.cur = SvPVX (string);
757	dec.end = SvEND (string);
758	dec.err = 0;
759
760	*dec.end = 1; // invalid anywhere
761	sv = decode_sv (&dec);
762	*dec.end = 0;
763
764	if (!sv)
765	{
766	IV offset = utf8_distance (dec.cur, SvPVX (string));
767	SV *uni = sv_newmortal ();
768
769	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
770	croak ("%s, at character %d (%s)",
771	dec.err,
772	(int)offset,
773	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
774	}
775
776	sv_dump (sv);//D
777	return sv_2mortal (sv);
778	}
779
780	MODULE = JSON::XS PACKAGE = JSON::XS
781
782	BOOT:
783	{
784	int i;
785
786	memset (decode_hexdigit, 0xff, 256);
787	for (i = 10; i--; )
788	decode_hexdigit ['0' + i] = i;
789
790	for (i = 6; --i; )
791	{
792	decode_hexdigit ['a' + i] = 10 + i;
793	decode_hexdigit ['A' + i] = 10 + i;
794	}
795
796	json_stash = gv_stashpv ("JSON::XS", 1);
797	}
798
799	SV new (char dummy)
800	CODE:
801	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
802	OUTPUT:
803	RETVAL
804
805	SV ascii (SV self, int enable)
806	ALIAS:
807	ascii = F_ASCII
808	utf8 = F_UTF8
809	indent = F_INDENT
810	canonical = F_CANONICAL
811	space_before = F_SPACE_BEFORE
812	space_after = F_SPACE_AFTER
813	json_rpc = F_JSON_RPC
814	CODE:
815	{
816	UV *uv = SvJSON (self);
817	if (enable)
818	*uv \|= ix;
819	else
820	*uv &= ~ix;
821
822	RETVAL = newSVsv (self);
823	}
824	OUTPUT:
825	RETVAL
826
827	void encode (SV self, SV scalar)
828	PPCODE:
829	XPUSHs (encode_json (scalar, *SvJSON (self)));
830
831	void decode (SV self, SV jsondata)
832	PPCODE:
833	XPUSHs (decode_json (jsondata, *SvJSON (self)));
834