binary format change: encode numbers as ULEB128 instead of original uint8/16/32/64 format v_0
authorFrantišek Kučera <franta-hg@frantovo.cz>
Tue, 10 Dec 2019 19:37:50 +0100
branchv_0
changeset 45 24a506eb97b5
parent 44 bcff4ebca71d
child 46 12c329f5524f
binary format change: encode numbers as ULEB128 instead of original uint8/16/32/64 format
src/types/IntegerDataTypeReader.h
--- a/src/types/IntegerDataTypeReader.h	Tue Dec 03 16:30:02 2019 +0100
+++ b/src/types/IntegerDataTypeReader.h	Tue Dec 10 19:37:50 2019 +0100
@@ -38,61 +38,16 @@
 using namespace relpipe::reader;
 
 /**
- * The prototype does not have various integer and other numeric data types,
- * it just works with one type of integer.
- * But this integer has variable length -- smaller values occupy only one byte, bigger ones, more bytes 1,2,4,8 + first byte (contains length signalization).
- * In the real implementation of relational pipes, there will be DataTypes for particular numeric types.
- * 
- * TODO: support also big endian architectures.
- * TODO: throw exception if a value was stored in bigger type than needed (while reading – there should be only one supported way how to encode a single value)
- * 
- * Example of encoded values:
- * -------------------------------------------------------------------------------------------------
- * $ for n in 0 1 10 250 251 252 65535 65536 4294967295 4294967296 18446744073709551615; do printf '%20s = ' $n; dist/Debug/GNU-Linux/rp-prototype write integer $n | hd | head -n 1; done
- *                    0 = 00000000  00                                                |.|
- *                    1 = 00000000  01                                                |.|
- *                   10 = 00000000  0a                                                |.|
- *                  250 = 00000000  fa                                                |.|
- *                  251 = 00000000  fb fb                                             |..|
- *                  252 = 00000000  fb fc                                             |..|
- *                65535 = 00000000  fc ff ff                                          |...|
- *                65536 = 00000000  fd 00 00 01 00                                    |.....|
- *           4294967295 = 00000000  fd ff ff ff ff                                    |.....|
- *           4294967296 = 00000000  fe 00 00 00 00 01 00 00  00                       |.........|
- * 18446744073709551615 = 00000000  fe ff ff ff ff ff ff ff  ff                       |.........|
- * -------------------------------------------------------------------------------------------------
- * 
- * Example of decoded values:
- * -------------------------------------------------------------------------------------------------
- * $ for n in 0 1 10 250 251 252 65535 65536 4294967295 4294967296 18446744073709551615; do dist/Debug/GNU-Linux/rp-prototype write integer $n | dist/Debug/GNU-Linux/rp-prototype read integer; done;
- * 0
- * 1
- * 10
- * 250
- * 251
- * 252
- * 65535
- * 65536
- * 4294967295
- * 4294967296
- * 18446744073709551615
- * -------------------------------------------------------------------------------------------------
- * 
- * Note: similar format as original idea: https://en.wikipedia.org/wiki/X.690#Length_octets
- * 
+ * Unsigned variable-length integer.
+ * ULEB128
  */
 class IntegerDataTypeReader : public DataTypeReader<integer_t> {
 private:
-	static const uint8_t INTEGER_TYPE_UINT8 = 251;
-	static const uint8_t INTEGER_TYPE_UINT16 = 252;
-	static const uint8_t INTEGER_TYPE_UINT32 = 253;
-	static const uint8_t INTEGER_TYPE_UINT64 = 254;
-	static const uint8_t INTEGER_TYPE_RESERVED = 255;
 
-	template<typename T> integer_t read(std::istream &input) {
-		T value = 0;
-		input.read(reinterpret_cast<char *> (&value), sizeof (value));
-		return value;
+	uint8_t readNextOctet(std::istream &input) {
+		uint8_t value = input.get();
+		if (input.good()) return value;
+		else throw RelpipeReaderException(L"Unable to read next octet of the integer.");
 	}
 
 public:
@@ -101,19 +56,15 @@
 	}
 
 	integer_t readValue(std::istream &input) override {
-		uint8_t first;
-		input.read(reinterpret_cast<char *> (&first), sizeof (first));
-		if (input.good()) {
-
-			if (first < INTEGER_TYPE_UINT8) return first;
-			else if (first == INTEGER_TYPE_UINT8) return read<uint8_t>(input);
-			else if (first == INTEGER_TYPE_UINT16) return read<uint16_t>(input);
-			else if (first == INTEGER_TYPE_UINT32) return read<uint32_t>(input);
-			else if (first == INTEGER_TYPE_UINT64) return read<uint64_t>(input);
-			else throw RelpipeReaderException(L"Error while parsing integer type: unsupported type");
-		} else {
-			throw RelpipeReaderException(L"Error while reading integer from the stream.");
-		}
+		integer_t value = 0;
+		integer_t shift = 0;
+		octet_t octet;
+		do {
+			octet = readNextOctet(input);
+			value += integer_t(octet & 0x7F) << shift;
+			shift += 7;
+		} while (octet >= 128);
+		return value;
 	}
 
 	string_t toString(const integer_t &value) override {
@@ -124,4 +75,4 @@
 
 }
 }
-}
\ No newline at end of file
+}