scylladb/utils/serialization.hh

#pragma once

/*
 * Copyright (C) 2014 ScyllaDB
 */

/*
 * This file is part of Scylla.
 *
 * Scylla is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Affero General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * Scylla is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with Scylla.  If not, see <http://www.gnu.org/licenses/>.
 */

// The following is a redesigned subset of Java's DataOutput,
// DataOutputStream, DataInput, DataInputStream, etc. It allows serializing
// several primitive types (e.g., integer, string, etc.) to an object which
// is only capable of write()ing a single byte (write(char)) or an array of
// bytes (write(char *, int)), and deserializing the same data from an object
// with a char read() interface.
//
// The format of this serialization is identical to the format used by
// Java's DataOutputStream class. This is important to allow us communicate
// with nodes running Java version of the code.
//
// We only support the subset actually used in Cassandra, and the subset
// that is reversible, i.e., can be read back by data_input. For example,
// we only support DataOutput.writeUTF(string) and not
// DataOutput.writeChars(string) - because the latter does not include
// the length, which is necessary for reading the string back.

#include <stdint.h>

#include <seastar/util/gcc6-concepts.hh>
#include <seastar/core/sstring.hh>
#include <seastar/net/byteorder.hh>
#include "bytes.hh"
#include <iosfwd>
#include <iterator>


class UTFDataFormatException { };
class EOFException { };

static constexpr size_t serialize_int8_size = 1;
static constexpr size_t serialize_bool_size = 1;
static constexpr size_t serialize_int16_size = 2;
static constexpr size_t serialize_int32_size = 4;
static constexpr size_t serialize_int64_size = 8;

namespace internal_impl {

template <typename ExplicitIntegerType, typename CharOutputIterator, typename IntegerType>
GCC6_CONCEPT(requires std::is_integral<ExplicitIntegerType>::value && std::is_integral<IntegerType>::value && requires (CharOutputIterator it) {
    *it++ = 'a';
})
inline
void serialize_int(CharOutputIterator& out, IntegerType val) {
    ExplicitIntegerType nval = net::hton(ExplicitIntegerType(val));
    out = std::copy_n(reinterpret_cast<const char*>(&nval), sizeof(nval), out);
}

}

template <typename CharOutputIterator>
inline
void serialize_int8(CharOutputIterator& out, uint8_t val) {
    internal_impl::serialize_int<uint8_t>(out, val);
}

template <typename CharOutputIterator>
inline
void serialize_int16(CharOutputIterator& out, uint16_t val) {
    internal_impl::serialize_int<uint16_t>(out, val);
}

template <typename CharOutputIterator>
inline
void serialize_int32(CharOutputIterator& out, uint32_t val) {
    internal_impl::serialize_int<uint32_t>(out, val);
}

template <typename CharOutputIterator>
inline
void serialize_int64(CharOutputIterator& out, uint64_t val) {
    internal_impl::serialize_int<uint64_t>(out, val);
}

template <typename CharOutputIterator>
inline
void serialize_bool(CharOutputIterator& out, bool val) {
    serialize_int8(out, val ? 1 : 0);
}

// The following serializer is compatible with Java's writeUTF().
// In our C++ implementation, we assume the string is already UTF-8
// encoded. Unfortunately, Java's implementation is a bit different from
// UTF-8 for encoding characters above 16 bits in unicode (see
// http://docs.oracle.com/javase/7/docs/api/java/io/DataInput.html#modified-utf-8)
// For now we'll just assume those aren't in the string...
// TODO: fix the compatibility with Java even in this case.
template <typename CharOutputIterator>
GCC6_CONCEPT(requires requires (CharOutputIterator it) {
    *it++ = 'a';
})
inline
void serialize_string(CharOutputIterator& out, const sstring& s) {
    // Java specifies that nulls in the string need to be replaced by the
    // two bytes 0xC0, 0x80. Let's not bother with such transformation
    // now, but just verify wasn't needed.
    for (char c : s) {
        if (c == '\0') {
            throw UTFDataFormatException();
        }
    }
    if (s.size() > std::numeric_limits<uint16_t>::max()) {
        // Java specifies the string length is written as uint16_t, so we
        // can't serialize longer strings.
        throw UTFDataFormatException();
    }
    serialize_int16(out, s.size());
    out = std::copy(s.begin(), s.end(), out);
}

template <typename CharOutputIterator>
GCC6_CONCEPT(requires requires (CharOutputIterator it) {
    *it++ = 'a';
})
inline
void serialize_string(CharOutputIterator& out, const char* s) {
    // TODO: like above, need to change UTF-8 when above 16-bit.
    auto len = strlen(s);
    if (len > std::numeric_limits<uint16_t>::max()) {
        // Java specifies the string length is written as uint16_t, so we
        // can't serialize longer strings.
        throw UTFDataFormatException();
    }
    serialize_int16(out, len);
    out = std::copy_n(s, len, out);
}

inline
size_t serialize_string_size(const sstring& s) {;
    // As above, this code is missing the case of modified utf-8
    return serialize_int16_size + s.size();
}

template<typename T>
static inline
void write(bytes::iterator& out, const T& val) {
    auto v = net::ntoh(val);
    out = std::copy_n(reinterpret_cast<char*>(&v), sizeof(v), out);
}