spjson.c

/*
Это мой spj.h

Вставил заодно вместе с набросками .c
*/
#ifndef _SPJ_H
#define _SPJ_H

#include <stdlib.h>

struct jval;
#ifdef USE_OBJ_DICT
struct htab; // will be implemented sometime
#endif

typedef enum jValue_type {
  Unknown = 0,
  Object,
  Array,
  String,
  Number,
  Bool,
  Null
} jValue_type;

typedef struct jObject {
  struct jval *data; // array
  int size;
#ifdef USE_OBJ_DICT
  struct htab *hash;
#endif
} jObject;

typedef struct jArray {
  struct jval *data; // array
  int size;
} jArray;

typedef struct jString {
  char *data; // nil terminating
  int size;
} jString;

union value {
  struct jObject obj;
  struct jArray arr;
  struct jString str;
  double number;  // Bool, Null, integer and real (all in one)
};

typedef struct jval {
  jValue_type type;    // try to do it without flags
  struct jval *parent; // calculated AFTER the filling of the PARENT's container
  char *name;          // name pairs (valid only for members of the object)
  union value value;
} jValue;


typedef enum jtoken {
  Jeof,
  Jnumber,
  Jstring,
  Jbool,     // true false
  Jnull,     // null
  Jobjstart, // {
  Jarrstart, // [
  Jobjend,   // }
  Jarrend,   // ]
  Jcolon,    // : =
  Jcomma,    // ,
  Jname,     // extension for name : value (unquoted name)
  Jerror,
  Janymember,
  Jnomem = -1
} jToken;

typedef struct spjiter {
  const char *data;
  size_t dsize, curpos;
} jIter;

typedef struct jlexer {
  jIter it;
  jToken tok;
  int charin, charpos;
  char *errmsg;
  union value lv;
} jLexer;

jToken spj_getoken (jLexer *lex);
#endif

/*
  spj.c

   Это просто наброски кода лексера и двух функций парсера
   (парсер нужно рассматривать в качестве псевдокода)

   Станислав, я не стал класть сюда некоторые функции,
   возмжно самому будет интересно их написать (4/2/2014 я их длбавил сюда).
   Только у меня некоторые структуры немного другие.
*/

// функции для get_jutf8str и еще несколько для сообщений об ошибке

// эти .h я использую тут
#include <stdio.h>
#include <string.h>
#include <stdint.h>
#include <ctype.h>
#include <errno.h>
#include <stdarg.h>

// эта кладет в buf[] размером bs текст из исходного json, вызвавший ошибку при распознавании true,false,null
//   например, для tru или trueeee
static  char *
stsubstr (const char *src, const char *end, char buf[], int bs)
{
  int l = end - src;

  if (l > bs - 1) {
    strncpy(buf, src, bs - 4);
    strcpy(buf + bs - 4, "...");
  } else {
    strncpy(buf, src, l);
    buf[l] = 0;
  }
  return buf;
}

// like sprintf to dynamic string in jLexer
static jToken
spj_eprintf (jLexer *lex, jToken curtok, const char *fmt, ...)
{
  lex->errtok = curtok;
  int retl;
  jToken rc = Jnomem;
  va_list ap;

  va_start(ap, fmt);
  retl = vsnprintf(lex->errmsg, 0, fmt, ap);
  va_end(ap);
  if (lex->errmsg = (char *)malloc(retl + 1)) {
    va_start(ap, fmt);
    vsnprintf(lex->errmsg, retl + 1, fmt, ap);
    va_end(ap);
    rc = Jerror;
  }

  return rc;
}

// like sprintf to concat string with lex->errmsg
static jToken
spj_emsgcat (jLexer *lex, const char *fmt, ...)
{
  int retl, lemsg = lex->errmsg ? strlen(lex->errmsg) : 0;
  char *t = lex->errmsg;
  jToken rc = Jnomem;
  va_list ap;

  va_start(ap, fmt);
  retl = vsnprintf(t, 0, fmt, ap);
  va_end(ap);
  if (lex->errmsg = (char *)realloc(lex->errmsg, retl + lemsg + 1)) {
    va_start(ap, fmt);
    vsnprintf(lex->errmsg + lemsg, retl, fmt, ap);
    va_end(ap);
    rc = Jerror;
  } else {
    lex->errmsg = t;
  }

  return rc;
}

// check true,false,null
static int
check_tfn (jLexer *lex, const char *w)
{
  int c;
  const char *cw = w;

  while ((c = spj_getc(&lex->it)) && *w == c)
    w++;

  if (c && *w == 0)
    if (isspace(c) || strchr("]},", c)) {
      spj_seek(&lex->it, -1);
      return 1;
    }
  char buf[10];
  if (c) {
    spj_eprintf (lex, lex->charin == 'n' ? Jnull : Jbool,
         "unexpected '%c%s%c...' instead of '%c%s' at pos: %d\n",
         lex->charin, stsubstr(cw, w, buf, sizeof(buf)), c,
         lex->charin, cw, lex->charpos);
    spj_seek(&lex->it, -1);
  } else {
    spj_eprintf (lex, lex->charin == 'n' ? Jnull : Jbool,
         "unexpected EOF instead of '%c%s' at pos: %d\n",
         lex->charin, cw, lex->charpos);
  }
  return 0;
}

// Call it only for VALID c !!!
static inline int
hexdigit (int c)
{
  if (c <= '9')
    return c-'0';
  if (c <= 'F')
    return c-'A'+10;
  return c-'a'+10;
}

// returns 0 if surrogate pair is not valid
int
surpair_to_ucs (int sp[2])
{
  int ucs = 0;

  if ((sp[0] & 0xFFFFFC00) == 0xD800 && (sp[1] & 0xFFFFFC00) == 0xDC00)
    ucs = ((sp[0] & 0x3FF) << 10) | (sp[1] & 0x3FF);
  return ucs;
}

/*
  For a given UCS puts in memory bytes in UTF-8
  Returns the number of bytes, or 0 on error (UCS <0)
 */
int
ucs_to_utf8 (int uc, char *b)
{
  if (uc < 0)
    return 0;
  u_int ucs = uc;
  if (ucs < 128) {
    b[0] = ucs;
    return 1;
  }

  int n = 6, i;
  if (ucs < 0x800)
    n = 2;
  else if (ucs < 0x10000)
    n = 3;
  else if (ucs < 0x200000)
    n = 4;
  else if (ucs < 0x4000000)
    n = 5;

  char *u = b+n-1;
  static u_int
    mask1[7] = { 0x0, 0x0, 0x1f, 0x0f, 0x07, 0x03, 0x01},
    mask2[7] = { 0x0, 0x0, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc};

  for (i = 0; i < n-1; i++, u--, ucs >>= 6)
    *u = (ucs & 0x3f) | 0x80;
  *u = (ucs & mask1[n]) | mask2[n];

  return n;
}

// 4 hex digits to integer (returns -1 on error)
static int
spj_gethex4 (jLexer *lex)
{
  int i, ucs = 0, c;

  for (i = 0; i < 4; i++) {
    if (isxdigit(c = spj_getc(&lex->it)))
      ucs = (ucs << 4) | hexdigit(c);
    else {
      spj_eprintf (lex, Jstring,
           "invalid '%c' (0x02x) character in \\uHHHH in Jstring before pos: %d\n",
           c, c, lex->it.curpos);
      return -1;
    }
  }
  return ucs;
}


// json String encoded utf-8
jToken
get_jutf8str (jLexer *lex)
{
  jToken rc = Jeof;
  int ucs32, c, lutf8, lstr = 0, capacity, incr = JSTR_INIT_INCREMENT;
  char utf8[8], *str = (char *)malloc(capacity = JSTR_INIT_SIZE);
  if (!str)
    return Jnomem;

  while (c = spj_getc(&lex->it)) {
    ucs32 = -1;
    if (c == '"') { // return result
      lex->lv.str.size = lstr;
      str[lstr++] = 0;
      lex->lv.str.data = (char *)realloc(str, lstr); // free unused bytes
      return Jstring;
    }
    if (c == '\\') {
      // c == 0 after switch() means any error is found
      switch (c = spj_getc(&lex->it)) {
      case 0: break;
      case 't': c = '\t'; break;
      case 'r': c = '\r'; break;
      case 'n': c = '\n'; break;
      case 'f': c = '\f'; break;
      case 'b': c = '\b'; break;
      case '\\': c = '\\'; break;
      case '/': c = '/'; break;
      case '"': c = '"'; break;
      case 'u':
    {
      int sp[2]; // assume the appearance of a surrogate pair
      if ((ucs32 = sp[0] = spj_gethex4(lex)) == -1) {
        rc = Jerror; c = 0; // c == 0 error flag
            break;
      }
      if ((ucs32 & 0xFC00) == 0xd800) { // surrogate pair
        if ((c = spj_getc(&lex->it)) == '\\' &&
        (c = spj_getc(&lex->it)) == 'u') {
          if (!(sp[1] = spj_gethex4(lex))) {
        rc = Jerror; c = 0; break;
          }
          if (!(ucs32 = surpair_to_ucs(sp))) {
        fprintf (stderr, "invalid surrogate pair in Jstring before pos: %ld\n",
             lex->it.curpos);
            rc = Jerror; c = 0; break;
          }
        } else {
          fprintf (stderr, "unexpected character '\\%c' (0x%02x) in Jstring at pos: %ld\n",
               c, c, lex->it.curpos);
          rc = Jerror; c = 0; break;
        }
      }
    }
    break;
      default:
    rc = Jerror;
    fprintf (stderr, "invalid escape sequence '\\%c' (0x%02x) in Jstring at pos: %ld\n",
         c, c, lex->it.curpos);
    c = 0;
      }
      if (!c)
    break; // break while () loop, free mem, returns error
    }
    /*
      We assume that JSON encoded UTF-8.
      Therefore, all symbols except encoded by \uHHHH
      copied into the result
    */
    if (ucs32 != -1) { // utf-8 value of \uHHHH is in utf8[]
      if (!(lutf8 = ucs_to_utf8(ucs32, utf8))) {
    rc = Jerror;
    fprintf (stderr, "invalid ucs to utf8 conversion near pos: %ld\n",
         lex->it.curpos);
    break;
      }
    } else {
      lutf8 = 1;
      utf8[0] = c;
    }
    if (lutf8 + 1 + lstr > capacity) {
      char *t = (char *)realloc(str, capacity += (lutf8 + incr));
      if (!t) {
    rc = Jnomem;
    break;
      }
      str = t;
      if ((incr <<= 1) > JSTR_MAX_INCREMENT)
    incr = JSTR_MAX_INCREMENT;
    }
    memcpy(str + lstr, utf8, lutf8);
    lstr += lutf8;
  }

  free(str);
  return rc;
}

jToken
spj_getoken (jLexer *lex)
{
  int c, i;

  while (isspace(c = spj_getc(&lex->it)));

  lex->charin = c;
  lex->charpos = spj_seek(&lex->it, 0); // one position more !!!

  switch (c) {
  case 0:
    return lex->tok = Jeof;
  case '{':  return lex->tok = Jobjstart;
  case '}':  return lex->tok = Jobjend;
  case '[':  return lex->tok = Jarrstart;
  case ']':  return lex->tok = Jarrend;
  case ':':  case '=': return lex->tok = Jcolon;
  case ',':  return lex->tok = Jcomma;
  case 't': case 'f': case 'n':
    lex->lv.number = (c == 't'); // false, null = 0
    {
      const char *w = "rue";
      if (c == 'f')
    w = "alse";
      else if (c == 'n')
    w = "ull";
      return !check_tfn(lex, w) ? lex->tok = Jerror :
    (c == 'n') ? (lex->tok = Jnull) : (lex->tok = Jbool);
    }
  case '"':
    return lex->tok = get_jutf8str(lex);
  default :
    if (isdigit(c) || c == '-') {
      char *t, s[22];
      s[0] = c;
      for (i = 1;
       (isdigit(c = spj_getc(&lex->it)) || strchr("eE+-.", c)) && i < 20;
       i++)
    s[i] = c;
      s[i++] = c;
      s[i] = 0;
      errno = 0;
      lex->lv.number = strtod(s, &t);
      if (errno || s == t ||
      !(isspace(*t) || *t == ',' || *t == '}' || *t == ']')) {
    fprintf (stderr, "invalid number '%s' at pos: %ld\n",
         s, lex->it.curpos);
    return lex->tok = Jerror;
      }
      spj_seek(&lex->it, -1);
      return lex->tok = Jnumber;
    }
    fprintf (stderr, "unexpected character  '%c' (0x%02x) at pos: %ld\n",
         c, c, lex->it.curpos);
    return lex->tok = Jerror;
  }

  fprintf (stderr, "interror: spj_getoken() notreached\n");
  return lex->tok = Jerror;
}

/*
  это чисто заглушки, чтобы компилились spj_parse_object() и spj_parse_array()
*/
static void complete_obj (jValue *obj) {

}

static void delete_member (jValue *obj) {

}

static void complete_arr (jValue *arr) {

}

static void add_arrmember (jValue *arr, jValue *member, int n) {

}

static void put_objmember (jValue *obj, jValue *member) {

}

static jToken struct_error (jToken rc, jToken expected, jLexer *lex, jValue *obj) {

}

static jValue_type do_jtype (jToken rc) {

}

jToken spj_parse_array (jLexer *lex, jValue *obj);

jToken
spj_parse_object (jLexer *lex, jValue *obj)
{
  obj->type = Object;
  char *name;
  jToken rc;
  int n;

  for (n = 0;;n++) {
    if ((rc = spj_getoken(lex)) == Jobjend && n == 0)
      break;
    if (rc != Jstring)
      return struct_error(rc, Jstring, lex, obj);
    name = lex->lv.str.data;

    if ((rc = spj_getoken(lex)) != Jcolon)
      return struct_error(rc, Jcolon, lex, obj);

    jValue member = {0};
    member.name = name;
    switch (rc = spj_getoken(lex)) {
    case Jnumber: case Jstring: case Jbool: case Jnull:
      member.value = lex->lv;
      member.type = do_jtype(rc);
      break;
    case Jobjstart:
      if ((rc = spj_parse_object(lex, &member)) == Jobjend)
    break; // OK
      delete_member(&member);
      return rc;
    case Jarrstart:
      if ((rc = spj_parse_array(lex, &member)) == Jarrend)
    break; // OK
      delete_member(&member);
      return rc;
    default:
      return struct_error(rc, Janymember, lex, obj);
    }
    put_objmember(obj, &member);

    if ((rc = spj_getoken(lex)) == Jobjend)
      break;
    if (rc != Jcomma)
      return struct_error(rc, Jcomma, lex, obj);
  }

  complete_obj(obj);
  return rc;
}

jToken
spj_parse_array (jLexer *lex, jValue *obj)
{
  obj->type = Array;
  jToken rc;
  int n;

  for (n = 0;;n++) {
    if ((rc = spj_getoken(lex)) == Jarrend && n == 0)
      break;
    jValue member = {0};

    switch (rc) {
    case Jnumber: case Jstring: case Jbool: case Jnull:
      member.value = lex->lv;
      member.type = do_jtype(rc);
      break;
    case Jobjstart:
      if ((rc = spj_parse_object(lex, &member)) == Jobjend)
    break; // OK
      delete_member(&member);
      return rc;
    case Jarrstart:
      if ((rc = spj_parse_array(lex, &member)) == Jarrend)
    break; // OK
      delete_member(&member);
      return rc;
    default:
      return struct_error(rc, Janymember, lex, obj);
    }
    add_arrmember(obj, &member, n);

    if ((rc = spj_getoken(lex)) == Jarrend)
      break;
    if (rc != Jcomma)
      return struct_error(rc, Jcomma, lex, obj);
  }

  complete_arr(obj);
  return rc;
}