pg 1.0.0 → 1.5.9

data/ext/pg_text_decoder.c

@@ -1,6 +1,6 @@
 /*
  * pg_text_decoder.c - PG::TextDecoder module
- * $Id: pg_text_decoder.c,v fcf731d3dff7 2015/09/08 12:25:06 jfali $
+ * $Id$
  *
  */
 
@@ -12,13 +12,13 @@
  * assigns a decoder function.
  *
  * Signature of all type cast decoders is:
- *    VALUE decoder_function(t_pg_coder *this, char *val, int len, int tuple, int field, int enc_idx)
+ *    VALUE decoder_function(t_pg_coder *this, const char *val, int len, int tuple, int field, int enc_idx)
  *
  * Params:
  *   this     - The data part of the coder object that belongs to the decoder function.
- *   val, len - The text or binary data to decode. The caller ensures, that the data is
- *              zero terminated ( that is val[len] = 0 ). The memory should be used read
- *              only by the callee.
+ *   val, len - The text or binary data to decode.
+ *              The caller ensures, that text data (format=0) is zero terminated so that val[len]=0.
+ *              The memory should be used read-only by the callee.
  *   tuple    - Row of the value within the result set.
  *   field    - Column of the value within the result set.
  *   enc_idx  - Index of the Encoding that any output String should get assigned.
@@ -28,15 +28,38 @@
  *
  */
 
+#include "ruby/version.h"
 #include "pg.h"
-#include "util.h"
+#include "pg_util.h"
 #ifdef HAVE_INTTYPES_H
 #include <inttypes.h>
 #endif
+#include <ctype.h>
+#include <time.h>
+#if !defined(_WIN32)
+#include <arpa/inet.h>
+#include <sys/socket.h>
+#endif
+#include <string.h>
 
 VALUE rb_mPG_TextDecoder;
-static ID s_id_decode;
-
+static ID s_id_Rational;
+static ID s_id_new;
+static ID s_id_utc;
+static ID s_id_getlocal;
+static ID s_id_BigDecimal;
+
+static VALUE s_IPAddr;
+static VALUE s_vmasks4;
+static VALUE s_vmasks6;
+static VALUE s_nan, s_pos_inf, s_neg_inf;
+static int use_ipaddr_alloc;
+static ID s_id_lshift;
+static ID s_id_add;
+static ID s_id_mask;
+static ID s_ivar_family;
+static ID s_ivar_addr;
+static ID s_ivar_mask_addr;
 
 /*
  * Document-class: PG::TextDecoder::Boolean < PG::SimpleDecoder
@@ -46,7 +69,7 @@ static ID s_id_decode;
  *
  */
 static VALUE
-pg_text_dec_boolean(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_boolean(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
 	if (len < 1) {
 		rb_raise( rb_eTypeError, "wrong data for text boolean converter in tuple %d field %d", tuple, field);
@@ -63,9 +86,9 @@ pg_text_dec_boolean(t_pg_coder *conv, char *val, int len, int tuple, int field,
  *
  */
 VALUE
-pg_text_dec_string(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_string(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
-	VALUE ret = rb_tainted_str_new( val, len );
+	VALUE ret = rb_str_new( val, len );
 	PG_ENCODING_SET_NOCHECK( ret, enc_idx );
 	return ret;
 }
@@ -78,7 +101,7 @@ pg_text_dec_string(t_pg_coder *conv, char *val, int len, int tuple, int field, i
  *
  */
 static VALUE
-pg_text_dec_integer(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_integer(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
 	long i;
 	int max_len;
@@ -103,7 +126,7 @@ pg_text_dec_integer(t_pg_coder *conv, char *val, int len, int tuple, int field,
 		 *     conn.exec("select generate_series(1,1000000)").values }
 		 *   end
 		 */
-		char *val_pos = val;
+		const char *val_pos = val;
 		char digit = *val_pos;
 		int neg;
 		int error = 0;
@@ -134,6 +157,34 @@ pg_text_dec_integer(t_pg_coder *conv, char *val, int len, int tuple, int field,
 	return rb_cstr2inum(val, 10);
 }
 
+/*
+ * Document-class: PG::TextDecoder::Numeric < PG::SimpleDecoder
+ *
+ * This is a decoder class for conversion of PostgreSQL numeric types
+ * to Ruby BigDecimal objects.
+ *
+ * As soon as this class is used, it requires the 'bigdecimal' gem.
+ *
+ */
+static VALUE
+pg_text_dec_numeric(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
+{
+	return rb_funcall(rb_cObject, s_id_BigDecimal, 1, rb_str_new(val, len));
+}
+
+/* called per autoload when TextDecoder::Numeric is used */
+static VALUE
+init_pg_text_decoder_numeric(VALUE rb_mPG_TextDecoder)
+{
+	rb_funcall(rb_mPG, rb_intern("require_bigdecimal_without_warning"), 0);
+	s_id_BigDecimal = rb_intern("BigDecimal");
+
+	/* dummy = rb_define_class_under( rb_mPG_TextDecoder, "Numeric", rb_cPG_SimpleDecoder ); */
+	pg_define_coder( "Numeric", pg_text_dec_numeric, rb_cPG_SimpleDecoder, rb_mPG_TextDecoder );
+
+	return Qnil;
+}
+
 /*
  * Document-class: PG::TextDecoder::Float < PG::SimpleDecoder
  *
@@ -142,9 +193,37 @@ pg_text_dec_integer(t_pg_coder *conv, char *val, int len, int tuple, int field,
  *
  */
 static VALUE
-pg_text_dec_float(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_float(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
-	return rb_float_new(strtod(val, NULL));
+	switch(*val) {
+		case 'N':
+			return s_nan;
+		case 'I':
+			return s_pos_inf;
+		case '-':
+			if (val[1] == 'I') {
+				return s_neg_inf;
+			} else {
+				return rb_float_new(rb_cstr_to_dbl(val, Qfalse));
+			}
+		default:
+			return rb_float_new(rb_cstr_to_dbl(val, Qfalse));
+	}
+}
+
+struct pg_blob_initialization {
+	char *blob_string;
+	size_t length;
+};
+
+static VALUE pg_create_blob(VALUE v) {
+	struct pg_blob_initialization *bi = (struct pg_blob_initialization *)v;
+	return rb_str_new(bi->blob_string, bi->length);
+}
+
+static VALUE pg_pq_freemem(VALUE mem) {
+  PQfreemem((void *)mem);
+  return Qfalse;
 }
 
 /*
@@ -155,18 +234,54 @@ pg_text_dec_float(t_pg_coder *conv, char *val, int len, int tuple, int field, in
  *
  */
 static VALUE
-pg_text_dec_bytea(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_bytea(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
-	unsigned char *to;
-	size_t to_len;
-	VALUE ret;
+	struct pg_blob_initialization bi;
 
-	to = PQunescapeBytea( (unsigned char *)val, &to_len);
+	bi.blob_string = (char *)PQunescapeBytea((unsigned char*)val, &bi.length);
+	if (bi.blob_string == NULL) {
+		rb_raise(rb_eNoMemError, "PQunescapeBytea failure: probably not enough memory");
+	}
+	return rb_ensure(pg_create_blob, (VALUE)&bi, pg_pq_freemem, (VALUE)bi.blob_string);
+}
+
+/*
+ * array_isspace() --- a non-locale-dependent isspace()
+ *
+ * We used to use isspace() for parsing array values, but that has
+ * undesirable results: an array value might be silently interpreted
+ * differently depending on the locale setting.  Now we just hard-wire
+ * the traditional ASCII definition of isspace().
+ */
+static int
+array_isspace(char ch)
+{
+	if (ch == ' ' ||
+		ch == '\t' ||
+		ch == '\n' ||
+		ch == '\r' ||
+		ch == '\v' ||
+		ch == '\f')
+		return 1;
+	return 0;
+}
 
-	ret = rb_tainted_str_new((char*)to, to_len);
-	PQfreemem(to);
+static int
+array_isdim(char ch)
+{
+	if ( (ch >= '0' && ch <= '9') ||
+		(ch == '-') ||
+		(ch == '+') ||
+		(ch == ':') )
+		return 1;
+	return 0;
+}
 
-	return ret;
+static void
+array_parser_error(t_pg_composite_coder *this, const char *text){
+	if( (this->comp.flags & PG_CODER_FORMAT_ERROR_MASK) == PG_CODER_FORMAT_ERROR_TO_RAISE ){
+		rb_raise( rb_eTypeError, "%s", text );
+	}
 }
 
 /*
@@ -174,7 +289,7 @@ pg_text_dec_bytea(t_pg_coder *conv, char *val, int len, int tuple, int field, in
  * https://github.com/dockyard/pg_array_parser
  */
 static VALUE
-read_array(t_pg_composite_coder *this, int *index, char *c_pg_array_string, int array_string_length, char *word, int enc_idx, int tuple, int field, t_pg_coder_dec_func dec_func)
+read_array_without_dim(t_pg_composite_coder *this, int *index, const char *c_pg_array_string, int array_string_length, char *word, int enc_idx, int tuple, int field, t_pg_coder_dec_func dec_func)
 {
 	/* Return value: array */
 	VALUE array;
@@ -198,7 +313,7 @@ read_array(t_pg_composite_coder *this, int *index, char *c_pg_array_string, int
 
 	/* Special case the empty array, so it doesn't need to be handled manually inside
 	* the loop. */
-	if(((*index) < array_string_length) && c_pg_array_string[(*index)] == '}')
+	if(((*index) < array_string_length) && c_pg_array_string[*index] == '}')
 	{
 		return array;
 	}
@@ -238,10 +353,17 @@ read_array(t_pg_composite_coder *this, int *index, char *c_pg_array_string, int
 			}
 			else if(c == '{')
 			{
+				VALUE subarray;
 				(*index)++;
-				rb_ary_push(array, read_array(this, index, c_pg_array_string, array_string_length, word, enc_idx, tuple, field, dec_func));
+				subarray = read_array_without_dim(this, index, c_pg_array_string, array_string_length, word, enc_idx, tuple, field, dec_func);
+				rb_ary_push(array, subarray);
 				escapeNext = 1;
 			}
+			else if(c == 0)
+			{
+				array_parser_error( this, "premature end of the array string" );
+				return array;
+			}
 			else
 			{
 				word[word_index] = c;
@@ -268,30 +390,112 @@ read_array(t_pg_composite_coder *this, int *index, char *c_pg_array_string, int
 		}
 	}
 
+	array_parser_error( this, "premature end of the array string" );
 	return array;
 }
 
 /*
  * Document-class: PG::TextDecoder::Array < PG::CompositeDecoder
  *
- * This is the decoder class for PostgreSQL array types.
+ * This is a decoder class for PostgreSQL array types.
  *
- * All values are decoded according to the #elements_type
- * accessor. Sub-arrays are decoded recursively.
+ * It returns an Array with possibly an arbitrary number of sub-Arrays.
+ * All values are decoded according to the #elements_type accessor.
+ * Sub-arrays are decoded recursively.
+ *
+ * This decoder simply ignores any dimension decorations preceding the array values.
+ * It returns all array values as regular ruby Array with a zero based index, regardless of the index given in the dimension decoration.
+ *
+ * An array decoder which respects dimension decorations is waiting to be implemented.
  *
  */
 static VALUE
-pg_text_dec_array(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_array(t_pg_coder *conv, const char *c_pg_array_string, int array_string_length, int tuple, int field, int enc_idx)
 {
+	int index = 0;
+	int ndim = 0;
+	VALUE ret;
 	t_pg_composite_coder *this = (t_pg_composite_coder *)conv;
-	t_pg_coder_dec_func dec_func = pg_coder_dec_func(this->elem, 0);
-	/* create a buffer of the same length, as that will be the worst case */
-	char *word = xmalloc(len + 1);
-	int index = 1;
-
-	VALUE return_value = read_array(this, &index, val, len, word, enc_idx, tuple, field, dec_func);
-	free(word);
-	return return_value;
+
+	/*
+	 * If the input string starts with dimension info, read and use that.
+	 * Otherwise, we require the input to be in curly-brace style, and we
+	 * prescan the input to determine dimensions.
+	 *
+	 * Dimension info takes the form of one or more [n] or [m:n] items. The
+	 * outer loop iterates once per dimension item.
+	 */
+	for (;;)
+	{
+		/*
+		 * Note: we currently allow whitespace between, but not within,
+		 * dimension items.
+		 */
+		while (array_isspace(c_pg_array_string[index]))
+			index++;
+		if (c_pg_array_string[index] != '[')
+			break;				/* no more dimension items */
+		index++;
+
+		while (array_isdim(c_pg_array_string[index]))
+			index++;
+
+		if (c_pg_array_string[index] != ']'){
+			array_parser_error( this, "missing \"]\" in array dimensions");
+			break;
+		}
+		index++;
+
+		ndim++;
+	}
+
+	if (ndim == 0)
+	{
+		/* No array dimensions */
+	}
+	else
+	{
+		/* If array dimensions are given, expect '=' operator */
+		if (c_pg_array_string[index] != '=') {
+			array_parser_error( this, "missing assignment operator");
+			index-=2; /* jump back to before "]" so that we don't break behavior to pg < 1.1 */
+		}
+		index++;
+
+		while (array_isspace(c_pg_array_string[index]))
+			index++;
+	}
+
+	if (c_pg_array_string[index] != '{')
+		array_parser_error( this, "array value must start with \"{\" or dimension information");
+	index++;
+
+	if ( index < array_string_length && c_pg_array_string[index] == '}' ) {
+		/* avoid buffer allocation for empty array */
+		ret = rb_ary_new();
+	} else {
+		t_pg_coder_dec_func dec_func = pg_coder_dec_func(this->elem, 0);
+		/* create a buffer of the same length, as that will be the worst case */
+		VALUE buf = rb_str_new(NULL, array_string_length);
+		char *word = RSTRING_PTR(buf);
+
+		ret = read_array_without_dim(this, &index, c_pg_array_string, array_string_length, word, enc_idx, tuple, field, dec_func);
+
+		RB_GC_GUARD(buf);
+	}
+
+	if (c_pg_array_string[index] != '}' )
+		array_parser_error( this, "array value must end with \"}\"");
+	index++;
+
+	/* only whitespace is allowed after the closing brace */
+	for(;index < array_string_length; ++index)
+	{
+		if (!array_isspace(c_pg_array_string[index]))
+			array_parser_error( this, "malformed array literal: Junk after closing right brace.");
+	}
+
+	return ret;
 }
 
 /*
@@ -305,7 +509,7 @@ pg_text_dec_array(t_pg_coder *conv, char *val, int len, int tuple, int field, in
  *
  */
 static VALUE
-pg_text_dec_identifier(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_identifier(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
 	/* Return value: array */
 	VALUE array;
@@ -367,13 +571,13 @@ pg_text_dec_identifier(t_pg_coder *conv, char *val, int len, int tuple, int fiel
  *
  */
 static VALUE
-pg_text_dec_from_base64(t_pg_coder *conv, char *val, int len, int tuple, int field, int enc_idx)
+pg_text_dec_from_base64(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
 {
 	t_pg_composite_coder *this = (t_pg_composite_coder *)conv;
 	t_pg_coder_dec_func dec_func = pg_coder_dec_func(this->elem, this->comp.format);
 	int decoded_len;
 	/* create a buffer of the expected decoded length */
-	VALUE out_value = rb_tainted_str_new(NULL, BASE64_DECODED_SIZE(len));
+	VALUE out_value = rb_str_new(NULL, BASE64_DECODED_SIZE(len));
 
 	decoded_len = base64_decode( RSTRING_PTR(out_value), val, len );
 	rb_str_set_len(out_value, decoded_len);
@@ -392,13 +596,394 @@ pg_text_dec_from_base64(t_pg_coder *conv, char *val, int len, int tuple, int fie
 	return out_value;
 }
 
+static inline int char_to_digit(char c)
+{
+	return c - '0';
+}
+
+static int str2_to_int(const char *str)
+{
+	return char_to_digit(str[0]) * 10
+			+ char_to_digit(str[1]);
+}
+
+static int parse_year(const char **str) {
+	int year = 0;
+	int i;
+	const char * p = *str;
+
+	for(i = 0; isdigit(*p) && i < 7; i++, p++) {
+		year = 10 * year + char_to_digit(*p);
+	}
+
+	*str = p;
+	return year;
+}
+
+#define TZ_NEG 1
+#define TZ_POS 2
+
+/*
+ * Document-class: PG::TextDecoder::Timestamp < PG::SimpleDecoder
+ *
+ * This is a decoder class for conversion of PostgreSQL text timestamps
+ * to Ruby Time objects.
+ *
+ * The following flags can be used to specify time interpretation when no timezone is given:
+ * * +PG::Coder::TIMESTAMP_DB_UTC+ : Interpret timestamp as UTC time (default)
+ * * +PG::Coder::TIMESTAMP_DB_LOCAL+ : Interpret timestamp as local time
+ * * +PG::Coder::TIMESTAMP_APP_UTC+ : Return timestamp as UTC time (default)
+ * * +PG::Coder::TIMESTAMP_APP_LOCAL+ : Return timestamp as local time
+ *
+ * Example:
+ *   deco = PG::TextDecoder::Timestamp.new(flags: PG::Coder::TIMESTAMP_DB_UTC | PG::Coder::TIMESTAMP_APP_LOCAL)
+ *   deco.decode("2000-01-01 00:00:00")  # => 2000-01-01 01:00:00 +0100
+ *   deco.decode("2000-01-01 00:00:00.123-06")  # => 2000-01-01 00:00:00 -0600
+ */
+static VALUE pg_text_dec_timestamp(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
+{
+	const char *str = val;
+	int year, mon, day;
+	int hour, min, sec;
+	int nsec = 0;
+	int tz_given = 0;
+	int tz_hour = 0;
+	int tz_min = 0;
+	int tz_sec = 0;
+
+	year = parse_year(&str);
+
+	if ( year > 0
+		&& str[0] == '-' && isdigit(str[1]) && isdigit(str[2])
+		&& str[3] == '-' && isdigit(str[4]) && isdigit(str[5])
+		&& str[6] == ' ' && isdigit(str[7]) && isdigit(str[8])
+		&& str[9] == ':' && isdigit(str[10]) && isdigit(str[11])
+		&& str[12] == ':' && isdigit(str[13]) && isdigit(str[14])
+	) {
+
+		mon = str2_to_int(str+1); str += 3;
+		day = str2_to_int(str+1); str += 3;
+		hour = str2_to_int(str+1); str += 3;
+		min = str2_to_int(str+1); str += 3;
+		sec = str2_to_int(str+1); str += 3;
+
+		if (str[0] == '.' && isdigit(str[1])) {
+			/* nano second part, up to 9 digits */
+			static const int coef[9] = {
+				100000000, 10000000, 1000000,
+				100000, 10000, 1000, 100, 10, 1
+			};
+			int i;
+
+			str++;
+			for (i = 0; i < 9 && isdigit(*str); i++)
+			{
+				nsec += coef[i] * char_to_digit(*str++);
+			}
+			/* consume digits smaller than nsec */
+			while(isdigit(*str)) str++;
+		}
+
+		if ((str[0] == '+' || str[0] == '-') && isdigit(str[1]) && isdigit(str[2])) {
+			tz_given = str[0] == '-' ? TZ_NEG : TZ_POS;
+			tz_hour = str2_to_int(str+1); str += 3;
+
+			if (str[0] == ':' && isdigit(str[1]) && isdigit(str[2]))
+			{
+				tz_min = str2_to_int(str+1); str += 3;
+			}
+			if (str[0] == ':' && isdigit(str[1]) && isdigit(str[2]))
+			{
+				tz_sec = str2_to_int(str+1); str += 3;
+			}
+		}
+
+		if (str[0] == ' ' && str[1] == 'B' && str[2] == 'C') {
+			year = -year + 1;
+			str += 3;
+		}
+
+		if (*str == '\0') { /* must have consumed all the string */
+			VALUE sec_value;
+			VALUE gmt_offset_value;
+			VALUE res;
+
+#if (RUBY_API_VERSION_MAJOR > 2 || (RUBY_API_VERSION_MAJOR == 2 && RUBY_API_VERSION_MINOR >= 3)) && defined(HAVE_TIMEGM)
+			/* Fast path for time conversion */
+			struct tm tm;
+			struct timespec ts;
+			tm.tm_year = year - 1900;
+			tm.tm_mon = mon - 1;
+			tm.tm_mday = day;
+			tm.tm_hour = hour;
+			tm.tm_min = min;
+			tm.tm_sec = sec;
+			tm.tm_isdst = -1;
+
+			if (tz_given) {
+				/* with timezone */
+				time_t time = timegm(&tm);
+				if (time != -1){
+					int gmt_offset;
+
+					gmt_offset = tz_hour * 3600 + tz_min * 60 + tz_sec;
+					if (tz_given == TZ_NEG)
+					{
+						gmt_offset = - gmt_offset;
+					}
+					ts.tv_sec = time - gmt_offset;
+					ts.tv_nsec = nsec;
+					return rb_time_timespec_new(&ts, gmt_offset);
+				}
+			} else {
+				/* without timezone */
+				time_t time;
+
+				if( conv->flags & PG_CODER_TIMESTAMP_DB_LOCAL ) {
+					time = mktime(&tm);
+				} else {
+					time = timegm(&tm);
+				}
+				if (time != -1){
+					ts.tv_sec = time;
+					ts.tv_nsec = nsec;
+					return rb_time_timespec_new(&ts, conv->flags & PG_CODER_TIMESTAMP_APP_LOCAL ? INT_MAX : INT_MAX-1);
+				}
+			}
+			/* Some libc implementations fail to convert certain values,
+			* so that we fall through to the slow path.
+			*/
+#endif
+			if (nsec) {
+				int sec_numerator = sec * 1000000 + nsec / 1000;
+				int sec_denominator = 1000000;
+				sec_value = rb_funcall(Qnil, s_id_Rational, 2,
+						INT2NUM(sec_numerator), INT2NUM(sec_denominator));
+			} else {
+				sec_value = INT2NUM(sec);
+			}
+
+			if (tz_given) {
+				/* with timezone */
+				int gmt_offset;
+
+				gmt_offset = tz_hour * 3600 + tz_min * 60 + tz_sec;
+				if (tz_given == TZ_NEG)
+				{
+					gmt_offset = - gmt_offset;
+				}
+				gmt_offset_value = INT2NUM(gmt_offset);
+			} else {
+				/* without timezone */
+				gmt_offset_value = conv->flags & PG_CODER_TIMESTAMP_DB_LOCAL ? Qnil : INT2NUM(0);
+			}
+
+			res = rb_funcall(rb_cTime, s_id_new, 7,
+					INT2NUM(year),
+					INT2NUM(mon),
+					INT2NUM(day),
+					INT2NUM(hour),
+					INT2NUM(min),
+					sec_value,
+					gmt_offset_value);
+
+			if (tz_given) {
+				/* with timezone */
+				return res;
+			} else {
+				/* without timezone */
+				if( (conv->flags & PG_CODER_TIMESTAMP_DB_LOCAL) && (conv->flags & PG_CODER_TIMESTAMP_APP_LOCAL) ) {
+					return res;
+				} else if( conv->flags & PG_CODER_TIMESTAMP_APP_LOCAL ) {
+					return rb_funcall(res, s_id_getlocal, 0);
+				} else {
+					return rb_funcall(res, s_id_utc, 0);
+				}
+			}
+		}
+	}
+
+	/* fall through to string conversion */
+	return pg_text_dec_string(conv, val, len, tuple, field, enc_idx);
+}
+
+/*
+ * Document-class: PG::TextDecoder::Inet < PG::SimpleDecoder
+ *
+ * This is a decoder class for conversion of PostgreSQL inet type
+ * to Ruby IPAddr values.
+ *
+ * As soon as this class is used, it requires the ruby standard library 'ipaddr'.
+ */
+static VALUE
+pg_text_dec_inet(t_pg_coder *conv, const char *val, int len, int tuple, int field, int enc_idx)
+{
+	VALUE ip;
+#if defined(_WIN32)
+	ip = rb_str_new(val, len);
+	ip = rb_class_new_instance(1, &ip, s_IPAddr);
+#else
+	VALUE ip_int;
+	VALUE vmasks;
+	char dst[16];
+	char buf[64];
+	int af = strchr(val, '.') ? AF_INET : AF_INET6;
+	int mask = -1;
+
+	if (len >= 64) {
+		rb_raise(rb_eTypeError, "too long data for text inet converter in tuple %d field %d", tuple, field);
+	}
+
+	if (len >= 4) {
+		if (val[len-2] == '/') {
+			mask = val[len-1] - '0';
+			memcpy(buf, val, len-2);
+			buf[len-2] = '\0';
+			val = buf;
+		} else if (val[len-3] == '/') {
+			mask = (val[len-2]- '0')*10 + val[len-1] - '0';
+			memcpy(buf, val, len-3);
+			buf[len-3] = '\0';
+			val = buf;
+		} else if (val[len-4] == '/') {
+			mask = (val[len-3]- '0')*100 + (val[len-2]- '0')*10 + val[len-1] - '0';
+			memcpy(buf, val, len-4);
+			buf[len-4] = '\0';
+			val = buf;
+		}
+	}
+
+	if (1 != inet_pton(af, val, dst)) {
+		rb_raise(rb_eTypeError, "wrong data for text inet converter in tuple %d field %d val", tuple, field);
+	}
+
+	if (af == AF_INET) {
+		unsigned int ip_int_native;
+
+		if (mask == -1) {
+			mask = 32;
+		} else if (mask < 0 || mask > 32) {
+			rb_raise(rb_eTypeError, "invalid mask for IPv4: %d", mask);
+		}
+		vmasks = s_vmasks4;
+
+		ip_int_native = read_nbo32(dst);
+
+		/* Work around broken IPAddr behavior of converting portion
+		   of address after netmask to 0 */
+		switch (mask) {
+			case 0:
+				ip_int_native = 0;
+				break;
+			case 32:
+				/* nothing to do */
+				break;
+			default:
+				ip_int_native &= ~((1UL<<(32-mask))-1);
+				break;
+		}
+
+		ip_int = UINT2NUM(ip_int_native);
+	} else {
+		unsigned long long * dstllp = (unsigned long long *)dst;
+		unsigned long long ip_int_native1;
+		unsigned long long ip_int_native2;
+
+		if (mask == -1) {
+			mask = 128;
+		} else if (mask < 0 || mask > 128) {
+			rb_raise(rb_eTypeError, "invalid mask for IPv6: %d", mask);
+		}
+		vmasks = s_vmasks6;
+
+		ip_int_native1 = read_nbo64(dstllp);
+		dstllp++;
+		ip_int_native2 = read_nbo64(dstllp);
+
+		if (mask == 128) {
+			/* nothing to do */
+		} else if (mask == 64) {
+			ip_int_native2 = 0;
+		} else if (mask == 0) {
+			ip_int_native1 = 0;
+			ip_int_native2 = 0;
+		} else if (mask < 64) {
+			ip_int_native1 &= ~((1ULL<<(64-mask))-1);
+			ip_int_native2 = 0;
+		} else {
+			ip_int_native2 &= ~((1ULL<<(128-mask))-1);
+		}
+
+		/* 4 Bignum allocations */
+		ip_int = ULL2NUM(ip_int_native1);
+		ip_int = rb_funcall(ip_int, s_id_lshift, 1, INT2NUM(64));
+		ip_int = rb_funcall(ip_int, s_id_add, 1, ULL2NUM(ip_int_native2));
+	}
+
+	if (use_ipaddr_alloc) {
+		ip = rb_obj_alloc(s_IPAddr);
+		rb_ivar_set(ip, s_ivar_family, INT2NUM(af));
+		rb_ivar_set(ip, s_ivar_addr, ip_int);
+		rb_ivar_set(ip, s_ivar_mask_addr, RARRAY_AREF(vmasks, mask));
+	} else {
+		VALUE ip_args[2];
+		ip_args[0] = ip_int;
+		ip_args[1] = INT2NUM(af);
+		ip = rb_class_new_instance(2, ip_args, s_IPAddr);
+		ip = rb_funcall(ip, s_id_mask, 1, INT2NUM(mask));
+	}
+
+#endif
+	return ip;
+}
+
+/* called per autoload when TextDecoder::Inet is used */
+static VALUE
+init_pg_text_decoder_inet(VALUE rb_mPG_TextDecoder)
+{
+	rb_require("ipaddr");
+	s_IPAddr = rb_funcall(rb_cObject, rb_intern("const_get"), 1, rb_str_new2("IPAddr"));
+	rb_global_variable(&s_IPAddr);
+	s_ivar_family = rb_intern("@family");
+	s_ivar_addr = rb_intern("@addr");
+	s_ivar_mask_addr = rb_intern("@mask_addr");
+	s_id_lshift = rb_intern("<<");
+	s_id_add = rb_intern("+");
+	s_id_mask = rb_intern("mask");
+
+	use_ipaddr_alloc = RTEST(rb_eval_string("IPAddr.new.instance_variables.sort == [:@addr, :@family, :@mask_addr]"));
+
+	s_vmasks4 = rb_eval_string("a = [0]*33; a[0] = 0; a[32] = 0xffffffff; 31.downto(1){|i| a[i] = a[i+1] - (1 << (31 - i))}; a.freeze");
+	rb_global_variable(&s_vmasks4);
+	s_vmasks6 = rb_eval_string("a = [0]*129; a[0] = 0; a[128] = 0xffffffffffffffffffffffffffffffff; 127.downto(1){|i| a[i] = a[i+1] - (1 << (127 - i))}; a.freeze");
+	rb_global_variable(&s_vmasks6);
+
+	/* dummy = rb_define_class_under( rb_mPG_TextDecoder, "Inet", rb_cPG_SimpleDecoder ); */
+	pg_define_coder( "Inet", pg_text_dec_inet, rb_cPG_SimpleDecoder, rb_mPG_TextDecoder);
+
+	return Qnil;
+}
+
+
 void
-init_pg_text_decoder()
+init_pg_text_decoder(void)
 {
-	s_id_decode = rb_intern("decode");
+	s_id_Rational = rb_intern("Rational");
+	s_id_new = rb_intern("new");
+	s_id_utc = rb_intern("utc");
+	s_id_getlocal = rb_intern("getlocal");
+
+	s_nan = rb_eval_string("0.0/0.0");
+	rb_global_variable(&s_nan);
+	s_pos_inf = rb_eval_string("1.0/0.0");
+	rb_global_variable(&s_pos_inf);
+	s_neg_inf = rb_eval_string("-1.0/0.0");
+	rb_global_variable(&s_neg_inf);
 
 	/* This module encapsulates all decoder classes with text input format */
 	rb_mPG_TextDecoder = rb_define_module_under( rb_mPG, "TextDecoder" );
+	rb_define_private_method(rb_singleton_class(rb_mPG_TextDecoder), "init_inet", init_pg_text_decoder_inet, 0);
+	rb_define_private_method(rb_singleton_class(rb_mPG_TextDecoder), "init_numeric", init_pg_text_decoder_numeric, 0);
 
 	/* Make RDoc aware of the decoder classes... */
 	/* dummy = rb_define_class_under( rb_mPG_TextDecoder, "Boolean", rb_cPG_SimpleDecoder ); */
@@ -413,6 +998,8 @@ init_pg_text_decoder()
 	pg_define_coder( "Bytea", pg_text_dec_bytea, rb_cPG_SimpleDecoder, rb_mPG_TextDecoder );
 	/* dummy = rb_define_class_under( rb_mPG_TextDecoder, "Identifier", rb_cPG_SimpleDecoder ); */
 	pg_define_coder( "Identifier", pg_text_dec_identifier, rb_cPG_SimpleDecoder, rb_mPG_TextDecoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextDecoder, "Timestamp", rb_cPG_SimpleDecoder ); */
+	pg_define_coder( "Timestamp", pg_text_dec_timestamp, rb_cPG_SimpleDecoder, rb_mPG_TextDecoder);
 
 	/* dummy = rb_define_class_under( rb_mPG_TextDecoder, "Array", rb_cPG_CompositeDecoder ); */
 	pg_define_coder( "Array", pg_text_dec_array, rb_cPG_CompositeDecoder, rb_mPG_TextDecoder );