yugabytedb-ysql 0.3

data/ext/pg_text_encoder.c

@@ -0,0 +1,827 @@
+/*
+ * pg_text_encoder.c - PG::TextEncoder module
+ * $Id$
+ *
+ */
+
+/*
+ *
+ * Type casts for encoding Ruby objects to PostgreSQL string representations.
+ *
+ * Encoder classes are defined with pg_define_coder(). This creates a new coder class and
+ * assigns an encoder function. The encoder function can decide between two different options
+ * to return the encoded data. It can either return it as a Ruby String object or write the
+ * encoded data to a memory space provided by the caller. In the second case, the encoder
+ * function is called twice, once for deciding the encoding option and returning the expected
+ * data length, and a second time when the requested memory space was made available by the
+ * calling function, to do the actual conversion and writing. Parameter intermediate can be
+ * used to store data between these two calls.
+ *
+ * Signature of all type cast encoders is:
+ *    int encoder_function(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate)
+ *
+ * Params:
+ *   this  - The data part of the coder object that belongs to the encoder function.
+ *   value - The Ruby object to cast.
+ *   out   - NULL for the first call,
+ *           pointer to a buffer with the requested size for the second call.
+ *   intermediate - Pointer to a VALUE that might be set by the encoding function to some
+ *           value in the first call that can be retrieved later in the second call.
+ *           This VALUE is not yet initialized by the caller.
+ *   enc_idx  - Index of the output Encoding that strings should be converted to.
+ *
+ * Returns:
+ *   >= 0  - If out==NULL the encoder function must return the expected output buffer size.
+ *           This can be larger than the size of the second call, but may not be smaller.
+ *           If out!=NULL the encoder function must return the actually used output buffer size
+ *           without a termination character.
+ *   -1    - The encoder function can alternatively return -1 to indicate that no second call
+ *           is required, but the String value in *intermediate should be used instead.
+ */
+
+
+#include "pg.h"
+#include "pg_util.h"
+#ifdef HAVE_INTTYPES_H
+#include <inttypes.h>
+#endif
+#include <math.h>
+
+VALUE rb_mPG_TextEncoder;
+static ID s_id_encode;
+static ID s_id_to_i;
+static ID s_id_to_s;
+static ID s_cBigDecimal;
+static VALUE s_str_F;
+
+static int pg_text_enc_integer(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate, int enc_idx);
+
+VALUE
+pg_obj_to_i( VALUE value )
+{
+	switch (TYPE(value)) {
+		case T_FIXNUM:
+		case T_FLOAT:
+		case T_BIGNUM:
+			return value;
+		default:
+			return rb_funcall(value, s_id_to_i, 0);
+	}
+}
+
+/*
+ * Document-class: PG::TextEncoder::Boolean < PG::SimpleEncoder
+ *
+ * This is the encoder class for the PostgreSQL bool type.
+ *
+ * Ruby value false is encoded as SQL +FALSE+ value.
+ * Ruby value true is encoded as SQL +TRUE+ value.
+ * Any other value is sent as it's string representation.
+ *
+ */
+static int
+pg_text_enc_boolean(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	switch( TYPE(value) ){
+		case T_FALSE:
+			if(out) *out = 'f';
+			return 1;
+		case T_TRUE:
+			if(out) *out = 't';
+			return 1;
+		case T_FIXNUM:
+		case T_BIGNUM:
+			if( NUM2LONG(value) == 0 ){
+				if(out) *out = '0';
+				return 1;
+			} else if( NUM2LONG(value) == 1 ){
+				if(out) *out = '1';
+				return 1;
+			} else {
+				return pg_text_enc_integer(this, value, out, intermediate, enc_idx);
+			}
+		default:
+			return pg_coder_enc_to_s(this, value, out, intermediate, enc_idx);
+	}
+	/* never reached */
+	return 0;
+}
+
+
+/*
+ * Document-class: PG::TextEncoder::String < PG::SimpleEncoder
+ *
+ * This is the encoder class for the PostgreSQL text types.
+ *
+ * Non-String values are expected to have method +to_s+ defined.
+ *
+ */
+int
+pg_coder_enc_to_s(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	VALUE str = rb_obj_as_string(value);
+	if( ENCODING_GET(str) == enc_idx ){
+		*intermediate = str;
+	}else{
+		*intermediate = rb_str_export_to_enc(str, rb_enc_from_index(enc_idx));
+	}
+	return -1;
+}
+
+static int
+count_leading_zero_bits(unsigned long long x)
+{
+#if defined(__GNUC__) || defined(__clang__)
+	return __builtin_clzll(x);
+#elif defined(_MSC_VER)
+	DWORD r = 0;
+	_BitScanForward64(&r, x);
+	return (int)r;
+#else
+	unsigned int a;
+	for(a=0; a < sizeof(unsigned long long) * 8; a++){
+		if( x & (1 << (sizeof(unsigned long long) * 8 - 1))) return a;
+		x <<= 1;
+	}
+	return a;
+#endif
+}
+
+/*
+ * Document-class: PG::TextEncoder::Integer < PG::SimpleEncoder
+ *
+ * This is the encoder class for the PostgreSQL integer types.
+ *
+ * Non-Integer values are expected to have method +to_i+ defined.
+ *
+ */
+static int
+pg_text_enc_integer(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	if(out){
+		if(TYPE(*intermediate) == T_STRING){
+			return pg_coder_enc_to_s(this, value, out, intermediate, enc_idx);
+		}else{
+			char *start = out;
+			int len;
+			int neg = 0;
+			long long sll = NUM2LL(*intermediate);
+			unsigned long long ll;
+
+			if (sll < 0) {
+				/* Avoid problems with the most negative integer not being representable
+				 * as a positive integer, by using unsigned long long for encoding.
+				 */
+				ll = -sll;
+				neg = 1;
+			} else {
+				ll = sll;
+			}
+
+			/* Compute the result string backwards. */
+			do {
+				unsigned long long remainder;
+				unsigned long long oldval = ll;
+
+				ll /= 10;
+				remainder = oldval - ll * 10;
+				*out++ = '0' + remainder;
+			} while (ll != 0);
+
+			if (neg)
+				*out++ = '-';
+
+			len = (int)(out - start);
+
+			/* Reverse string. */
+			out--;
+			while (start < out)
+			{
+				char swap = *start;
+
+				*start++ = *out;
+				*out-- = swap;
+			}
+
+			return len;
+		}
+	}else{
+		*intermediate = pg_obj_to_i(value);
+		if(TYPE(*intermediate) == T_FIXNUM){
+			long long sll = NUM2LL(*intermediate);
+			unsigned long long ll = sll < 0 ? -sll : sll;
+			int len = (sizeof(unsigned long long) * 8 - count_leading_zero_bits(ll)) / 3;
+			return sll < 0 ? len+2 : len+1;
+		}else{
+			return pg_coder_enc_to_s(this, *intermediate, NULL, intermediate, enc_idx);
+		}
+	}
+}
+
+#define MAX_DOUBLE_DIGITS 16
+
+/*
+ * Document-class: PG::TextEncoder::Float < PG::SimpleEncoder
+ *
+ * This is the encoder class for the PostgreSQL float types.
+ *
+ */
+static int
+pg_text_enc_float(t_pg_coder *conv, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	if(out){
+		double dvalue = NUM2DBL(value);
+		int len = 0;
+		int neg = 0;
+		int exp2i, exp10i, i;
+		unsigned long long ll, remainder, oldval;
+		VALUE intermediate;
+
+		/* Cast to the same strings as value.to_s . */
+		if( isinf(dvalue) ){
+			if( dvalue < 0 ){
+				memcpy( out, "-Infinity", 9);
+				return 9;
+			} else {
+				memcpy( out, "Infinity", 8);
+				return 8;
+			}
+		} else if (isnan(dvalue)) {
+			memcpy( out, "NaN", 3);
+			return 3;
+		}
+
+		/*
+		 * The following computation is roughly a conversion kind of
+		 *   sprintf( out, "%.16E", dvalue);
+		 */
+
+		/* write the algebraic sign */
+		if( dvalue < 0 ) {
+			dvalue = -dvalue;
+			*out++ = '-';
+			neg++;
+		}
+
+		/* retrieve the power of 2 exponent */
+		frexp(dvalue, &exp2i);
+		/* compute the power of 10 exponent */
+		exp10i = (int)floor(exp2i * 0.30102999566398114); /* Math.log(2)/Math.log(10) */
+		/* move the decimal point, so that we get an integer of MAX_DOUBLE_DIGITS decimal digits */
+		ll = (unsigned long long)(dvalue * pow(10, MAX_DOUBLE_DIGITS - 1 - exp10i) + 0.5);
+
+		/* avoid leading zeros due to inaccuracy of deriving exp10i from exp2i */
+		/* otherwise we would print "09.0" instead of "9.0" */
+		if( ll < 1000000000000000 ){ /* pow(10, MAX_DOUBLE_DIGITS-1) */
+			exp10i--;
+			ll *= 10;
+		}
+
+		if( exp10i <= -5 || exp10i >= 15 ) {
+			/* Write the float in exponent format (1.23e45) */
+
+			/* write fraction digits from right to left */
+			for( i = MAX_DOUBLE_DIGITS; i > 1; i--){
+				oldval = ll;
+				ll /= 10;
+				remainder = oldval - ll * 10;
+				/* omit trailing zeros */
+				if(remainder != 0 || len ) {
+					out[i] = '0' + remainder;
+					len++;
+				}
+			}
+
+			/* write decimal point */
+			if( len ){
+				out[1] = '.';
+				len++;
+			}
+
+			/* write remaining single digit left to the decimal point */
+			oldval = ll;
+			ll /= 10;
+			remainder = oldval - ll * 10;
+			out[0] = '0' + remainder;
+			len++;
+
+			/* write exponent */
+			out[len++] = 'e';
+			intermediate = INT2NUM(exp10i);
+
+			return neg + len + pg_text_enc_integer(conv, Qnil, out + len, &intermediate, enc_idx);
+		} else {
+			/* write the float in non exponent format (0.001234 or 123450.0) */
+
+			/* write digits from right to left */
+			int lz = exp10i < 0 ? 0 : exp10i;
+			for( i = MAX_DOUBLE_DIGITS - (exp10i < 0 ? exp10i : 0); i >= 0; i-- ){
+				oldval = ll;
+				ll /= 10;
+				remainder = oldval - ll * 10;
+				/* write decimal point */
+				if( i - 1 == lz ){
+					out[i--] = '.';
+					len++;
+				}
+				/* if possible then omit trailing zeros */
+				if(remainder != 0 || len || i - 2 == lz) {
+					out[i] = '0' + remainder;
+					len++;
+				}
+			}
+			return neg + len;
+		}
+	}else{
+		return 1 /*sign*/ + MAX_DOUBLE_DIGITS + 1 /*dot*/ + 1 /*e*/ + 1 /*exp sign*/ + 3 /*exp digits*/;
+	}
+}
+
+
+/*
+ * Document-class: PG::TextEncoder::Numeric < PG::SimpleEncoder
+ *
+ * This is the encoder class for the PostgreSQL numeric types.
+ *
+ * It converts Integer, Float and BigDecimal objects.
+ * All other objects are expected to respond to +to_s+.
+ */
+static int
+pg_text_enc_numeric(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	switch(TYPE(value)){
+		case T_FIXNUM:
+		case T_BIGNUM:
+			return pg_text_enc_integer(this, value, out, intermediate, enc_idx);
+		case T_FLOAT:
+			return pg_text_enc_float(this, value, out, intermediate, enc_idx);
+		default:
+			if(out){ /* second pass */
+				rb_bug("unexpected value type: %d", TYPE(value));
+			} else { /* first pass */
+				if( rb_obj_is_kind_of(value, s_cBigDecimal) ){
+					/* value.to_s('F') */
+					*intermediate = rb_funcall(value, s_id_to_s, 1, s_str_F);
+					return -1; /* no second pass */
+				} else {
+					return pg_coder_enc_to_s(this, value, NULL, intermediate, enc_idx);
+					/* no second pass */
+				}
+			}
+	}
+}
+
+/* called per autoload when TextEncoder::Numeric is used */
+static VALUE
+init_pg_text_encoder_numeric(VALUE rb_mPG_TextDecoder)
+{
+	s_str_F = rb_str_freeze(rb_str_new_cstr("F"));
+	rb_global_variable(&s_str_F);
+	rb_require("bigdecimal");
+	s_cBigDecimal = rb_const_get(rb_cObject, rb_intern("BigDecimal"));
+
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Numeric", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "Numeric", pg_text_enc_numeric, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+
+	return Qnil;
+}
+
+
+static const char hextab[] = {
+	'0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'a', 'b', 'c', 'd', 'e', 'f'
+};
+
+/*
+ * Document-class: PG::TextEncoder::Bytea < PG::SimpleEncoder
+ *
+ * This is an encoder class for the PostgreSQL +bytea+ type.
+ *
+ * The binary String is converted to hexadecimal representation for transmission
+ * in text format. For query bind parameters it is recommended to use
+ * PG::BinaryEncoder::Bytea or the hash form <tt>{value: binary_string, format: 1}</tt> instead,
+ * in order to decrease network traffic and CPU usage.
+ * See PG::Connection#exec_params for using the hash form.
+ *
+ * This encoder is particular useful when PG::TextEncoder::CopyRow is used with the COPY command.
+ * In this case there's no way to change the format of a single column to binary, so that the data have to be converted to bytea hex representation.
+ *
+ */
+static int
+pg_text_enc_bytea(t_pg_coder *conv, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	if(out){
+		size_t strlen = RSTRING_LEN(*intermediate);
+		char *iptr = RSTRING_PTR(*intermediate);
+		char *eptr = iptr + strlen;
+		char *optr = out;
+		*optr++ = '\\';
+		*optr++ = 'x';
+
+		for( ; iptr < eptr; iptr++ ){
+			unsigned char c = *iptr;
+			*optr++ = hextab[c >> 4];
+			*optr++ = hextab[c & 0xf];
+		}
+		return (int)(optr - out);
+	}else{
+		*intermediate = rb_obj_as_string(value);
+		/* The output starts with "\x" and each character is converted to hex. */
+		return 2 + RSTRING_LENINT(*intermediate) * 2;
+	}
+}
+
+typedef int (*t_quote_func)( void *_this, char *p_in, int strlen, char *p_out );
+
+static int
+quote_array_buffer( void *_this, char *p_in, int strlen, char *p_out ){
+	t_pg_composite_coder *this = _this;
+	char *ptr1;
+	char *ptr2;
+	int backslashs = 0;
+	int needquote;
+
+	/* count data plus backslashes; detect chars needing quotes */
+	if (strlen == 0)
+		needquote = 1;   /* force quotes for empty string */
+	else if (strlen == 4 && rbpg_strncasecmp(p_in, "NULL", strlen) == 0)
+		needquote = 1;   /* force quotes for literal NULL */
+	else
+		needquote = 0;
+
+	/* count required backlashs */
+	for(ptr1 = p_in; ptr1 != p_in + strlen; ptr1++) {
+		char ch = *ptr1;
+
+		if (ch == '"' || ch == '\\'){
+			needquote = 1;
+			backslashs++;
+		} else if (ch == '{' || ch == '}' || ch == this->delimiter ||
+					ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r' || ch == '\v' || ch == '\f'){
+			needquote = 1;
+		}
+	}
+
+	if( needquote ){
+		ptr1 = p_in + strlen;
+		ptr2 = p_out + strlen + backslashs + 2;
+		/* Write end quote */
+		*--ptr2 = '"';
+
+		/* Then store the escaped string on the final position, walking
+			* right to left, until all backslashs are placed. */
+		while( ptr1 != p_in ) {
+			*--ptr2 = *--ptr1;
+			if(*ptr2 == '"' || *ptr2 == '\\'){
+				*--ptr2 = '\\';
+			}
+		}
+		/* Write start quote */
+		*p_out = '"';
+		return strlen + backslashs + 2;
+	} else {
+		if( p_in != p_out )
+			memcpy( p_out, p_in, strlen );
+		return strlen;
+	}
+}
+
+static char *
+quote_string(t_pg_coder *this, VALUE value, VALUE string, char *current_out, int with_quote, t_quote_func quote_buffer, void *func_data, int enc_idx)
+{
+	int strlen;
+	VALUE subint;
+	t_pg_coder_enc_func enc_func = pg_coder_enc_func(this);
+
+	strlen = enc_func(this, value, NULL, &subint, enc_idx);
+
+	if( strlen == -1 ){
+		/* we can directly use String value in subint */
+		strlen = RSTRING_LENINT(subint);
+
+		if(with_quote){
+			/* size of string assuming the worst case, that every character must be escaped. */
+			current_out = pg_rb_str_ensure_capa( string, strlen * 2 + 2, current_out, NULL );
+
+			current_out += quote_buffer( func_data, RSTRING_PTR(subint), strlen, current_out );
+		} else {
+			current_out = pg_rb_str_ensure_capa( string, strlen, current_out, NULL );
+			memcpy( current_out, RSTRING_PTR(subint), strlen );
+			current_out += strlen;
+		}
+
+	} else {
+
+		if(with_quote){
+			/* size of string assuming the worst case, that every character must be escaped
+			 * plus two bytes for quotation.
+			 */
+			current_out = pg_rb_str_ensure_capa( string, 2 * strlen + 2, current_out, NULL );
+
+			/* Place the unescaped string at current output position. */
+			strlen = enc_func(this, value, current_out, &subint, enc_idx);
+
+			current_out += quote_buffer( func_data, current_out, strlen, current_out );
+		}else{
+			/* size of the unquoted string */
+			current_out = pg_rb_str_ensure_capa( string, strlen, current_out, NULL );
+			current_out += enc_func(this, value, current_out, &subint, enc_idx);
+		}
+	}
+	return current_out;
+}
+
+static char *
+write_array(t_pg_composite_coder *this, VALUE value, char *current_out, VALUE string, int quote, int enc_idx)
+{
+	int i;
+
+	/* size of "{}" */
+	current_out = pg_rb_str_ensure_capa( string, 2, current_out, NULL );
+	*current_out++ = '{';
+
+	for( i=0; i<RARRAY_LEN(value); i++){
+		VALUE entry = rb_ary_entry(value, i);
+
+		if( i > 0 ){
+			current_out = pg_rb_str_ensure_capa( string, 1, current_out, NULL );
+			*current_out++ = this->delimiter;
+		}
+
+		switch(TYPE(entry)){
+			case T_ARRAY:
+				current_out = write_array(this, entry, current_out, string, quote, enc_idx);
+			break;
+			case T_NIL:
+				current_out = pg_rb_str_ensure_capa( string, 4, current_out, NULL );
+				*current_out++ = 'N';
+				*current_out++ = 'U';
+				*current_out++ = 'L';
+				*current_out++ = 'L';
+				break;
+			default:
+				current_out = quote_string( this->elem, entry, string, current_out, quote, quote_array_buffer, this, enc_idx );
+		}
+	}
+	current_out = pg_rb_str_ensure_capa( string, 1, current_out, NULL );
+	*current_out++ = '}';
+	return current_out;
+}
+
+
+/*
+ * Document-class: PG::TextEncoder::Array < PG::CompositeEncoder
+ *
+ * This is the encoder class for PostgreSQL array types.
+ *
+ * All values are encoded according to the #elements_type
+ * accessor. Sub-arrays are encoded recursively.
+ *
+ * This encoder expects an Array of values or sub-arrays as input.
+ * Other values are passed through as text without interpretation.
+ *
+ */
+static int
+pg_text_enc_array(t_pg_coder *conv, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	char *end_ptr;
+	t_pg_composite_coder *this = (t_pg_composite_coder *)conv;
+
+	if( TYPE(value) == T_ARRAY){
+		VALUE out_str = rb_str_new(NULL, 0);
+		PG_ENCODING_SET_NOCHECK(out_str, enc_idx);
+
+		end_ptr = write_array(this, value, RSTRING_PTR(out_str), out_str, this->needs_quotation, enc_idx);
+
+		rb_str_set_len( out_str, end_ptr - RSTRING_PTR(out_str) );
+		*intermediate = out_str;
+
+		return -1;
+	} else {
+		return pg_coder_enc_to_s( conv, value, out, intermediate, enc_idx );
+	}
+}
+
+static char *
+quote_identifier( VALUE value, VALUE out_string, char *current_out ){
+	char *p_in = RSTRING_PTR(value);
+	size_t strlen = RSTRING_LEN(value);
+	char *p_inend = p_in + strlen;
+	char *end_capa = current_out;
+
+	PG_RB_STR_ENSURE_CAPA( out_string, strlen + 2, current_out, end_capa );
+	*current_out++ = '"';
+	for(; p_in != p_inend; p_in++) {
+		char c = *p_in;
+		if (c == '"'){
+			PG_RB_STR_ENSURE_CAPA( out_string, p_inend - p_in + 2, current_out, end_capa );
+			*current_out++ = '"';
+		} else if (c == 0){
+			rb_raise(rb_eArgError, "string contains null byte");
+		}
+		*current_out++ = c;
+	}
+	PG_RB_STR_ENSURE_CAPA( out_string, 1, current_out, end_capa );
+	*current_out++ = '"';
+
+	return current_out;
+}
+
+static char *
+pg_text_enc_array_identifier(VALUE value, VALUE string, char *out, int enc_idx)
+{
+	long i;
+	long nr_elems;
+
+	Check_Type(value, T_ARRAY);
+	nr_elems = RARRAY_LEN(value);
+
+	for( i=0; i<nr_elems; i++){
+		VALUE entry = rb_ary_entry(value, i);
+
+		StringValue(entry);
+		if( ENCODING_GET(entry) != enc_idx ){
+			entry = rb_str_export_to_enc(entry, rb_enc_from_index(enc_idx));
+		}
+		out = quote_identifier(entry, string, out);
+		if( i < nr_elems-1 ){
+			out = pg_rb_str_ensure_capa( string, 1, out, NULL );
+			*out++ = '.';
+		}
+	}
+	return out;
+}
+
+/*
+ * Document-class: PG::TextEncoder::Identifier < PG::SimpleEncoder
+ *
+ * This is the encoder class for PostgreSQL identifiers.
+ *
+ * An Array value can be used for identifiers of the kind "schema.table.column".
+ * This ensures that each element is properly quoted:
+ *   PG::TextEncoder::Identifier.new.encode(['schema', 'table', 'column'])
+ *      => '"schema"."table"."column"'
+ *
+ *  This encoder can also be used per PG::Connection#quote_ident .
+ */
+int
+pg_text_enc_identifier(t_pg_coder *this, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	VALUE out_str;
+	UNUSED( this );
+	if( TYPE(value) == T_ARRAY){
+		out_str = rb_str_new(NULL, 0);
+		out = RSTRING_PTR(out_str);
+		out = pg_text_enc_array_identifier(value, out_str, out, enc_idx);
+	} else {
+		StringValue(value);
+		if( ENCODING_GET(value) != enc_idx ){
+			value = rb_str_export_to_enc(value, rb_enc_from_index(enc_idx));
+		}
+		out_str = rb_str_new(NULL, RSTRING_LEN(value) + 2);
+		out = RSTRING_PTR(out_str);
+		out = quote_identifier(value, out_str, out);
+	}
+	rb_str_set_len( out_str, out - RSTRING_PTR(out_str) );
+	PG_ENCODING_SET_NOCHECK(out_str, enc_idx);
+	*intermediate = out_str;
+	return -1;
+}
+
+
+static int
+quote_literal_buffer( void *_this, char *p_in, int strlen, char *p_out ){
+	char *ptr1;
+	char *ptr2;
+	int backslashs = 0;
+
+	/* count required backlashs */
+	for(ptr1 = p_in; ptr1 != p_in + strlen; ptr1++) {
+		if (*ptr1 == '\''){
+			backslashs++;
+		}
+	}
+
+	ptr1 = p_in + strlen;
+	ptr2 = p_out + strlen + backslashs + 2;
+	/* Write end quote */
+	*--ptr2 = '\'';
+
+	/* Then store the escaped string on the final position, walking
+		* right to left, until all backslashs are placed. */
+	while( ptr1 != p_in ) {
+		*--ptr2 = *--ptr1;
+		if(*ptr2 == '\''){
+			*--ptr2 = '\'';
+		}
+	}
+	/* Write start quote */
+	*p_out = '\'';
+	return strlen + backslashs + 2;
+}
+
+
+/*
+ * Document-class: PG::TextEncoder::QuotedLiteral < PG::CompositeEncoder
+ *
+ * This is the encoder class for PostgreSQL literals.
+ *
+ * A literal is quoted and escaped by the <tt>'</tt> character, so that it can be inserted into SQL queries.
+ * It works equal to PG::Connection#escape_literal, but integrates into the type cast system of ruby-pg.
+ *
+ * Both expressions have the same result:
+ *   conn.escape_literal(PG::TextEncoder::Array.new.encode(["v1","v2"])) # => "'{v1,v2}'"
+ *   PG::TextEncoder::QuotedLiteral.new(elements_type: PG::TextEncoder::Array.new).encode(["v1","v2"]) # => "'{v1,v2}'"
+ * While escape_literal requires a intermediate ruby string allocation, QuotedLiteral encodes the values directly to the result string.
+ *
+ */
+static int
+pg_text_enc_quoted_literal(t_pg_coder *conv, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	t_pg_composite_coder *this = (t_pg_composite_coder *)conv;
+	VALUE out_str = rb_str_new(NULL, 0);
+	PG_ENCODING_SET_NOCHECK(out_str, enc_idx);
+
+	out = RSTRING_PTR(out_str);
+	out = quote_string(this->elem, value, out_str, out, this->needs_quotation, quote_literal_buffer, this, enc_idx);
+	rb_str_set_len( out_str, out - RSTRING_PTR(out_str) );
+	*intermediate = out_str;
+	return -1;
+}
+
+/*
+ * Document-class: PG::TextEncoder::ToBase64 < PG::CompositeEncoder
+ *
+ * This is an encoder class for conversion of binary to base64 data.
+ *
+ */
+static int
+pg_text_enc_to_base64(t_pg_coder *conv, VALUE value, char *out, VALUE *intermediate, int enc_idx)
+{
+	int strlen;
+	VALUE subint;
+	t_pg_composite_coder *this = (t_pg_composite_coder *)conv;
+	t_pg_coder_enc_func enc_func = pg_coder_enc_func(this->elem);
+
+	if(out){
+		/* Second encoder pass, if required */
+		strlen = enc_func(this->elem, value, out, intermediate, enc_idx);
+		base64_encode( out, out, strlen );
+
+		return BASE64_ENCODED_SIZE(strlen);
+	} else {
+		/* First encoder pass */
+		strlen = enc_func(this->elem, value, NULL, &subint, enc_idx);
+
+		if( strlen == -1 ){
+			/* Encoded string is returned in subint */
+			VALUE out_str;
+
+			strlen = RSTRING_LENINT(subint);
+			out_str = rb_str_new(NULL, BASE64_ENCODED_SIZE(strlen));
+			PG_ENCODING_SET_NOCHECK(out_str, enc_idx);
+
+			base64_encode( RSTRING_PTR(out_str), RSTRING_PTR(subint), strlen);
+			*intermediate = out_str;
+
+			return -1;
+		} else {
+			*intermediate = subint;
+
+			return BASE64_ENCODED_SIZE(strlen);
+		}
+	}
+}
+
+
+void
+init_pg_text_encoder(void)
+{
+	s_id_encode = rb_intern("encode");
+	s_id_to_i = rb_intern("to_i");
+	s_id_to_s = rb_intern("to_s");
+
+	/* This module encapsulates all encoder classes with text output format */
+	rb_mPG_TextEncoder = rb_define_module_under( rb_mPG, "TextEncoder" );
+	rb_define_private_method(rb_singleton_class(rb_mPG_TextEncoder), "init_numeric", init_pg_text_encoder_numeric, 0);
+
+	/* Make RDoc aware of the encoder classes... */
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Boolean", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "Boolean", pg_text_enc_boolean, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Integer", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "Integer", pg_text_enc_integer, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Float", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "Float", pg_text_enc_float, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "String", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "String", pg_coder_enc_to_s, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Bytea", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "Bytea", pg_text_enc_bytea, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Identifier", rb_cPG_SimpleEncoder ); */
+	pg_define_coder( "Identifier", pg_text_enc_identifier, rb_cPG_SimpleEncoder, rb_mPG_TextEncoder );
+
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "Array", rb_cPG_CompositeEncoder ); */
+	pg_define_coder( "Array", pg_text_enc_array, rb_cPG_CompositeEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "QuotedLiteral", rb_cPG_CompositeEncoder ); */
+	pg_define_coder( "QuotedLiteral", pg_text_enc_quoted_literal, rb_cPG_CompositeEncoder, rb_mPG_TextEncoder );
+	/* dummy = rb_define_class_under( rb_mPG_TextEncoder, "ToBase64", rb_cPG_CompositeEncoder ); */
+	pg_define_coder( "ToBase64", pg_text_enc_to_base64, rb_cPG_CompositeEncoder, rb_mPG_TextEncoder );
+}