Problem converting QByteArray to long
-
First of all, you need to understand endianness
Matching endianness between java and C++
void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); qint64 value = *reinterpret_cast<const qint64*>(msg.constData()); }Mismatching endianness between java and C++
void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); QByteArray reversed(sizeof(qint64),0); std::copy(msg.crbegin(),msg.crend(),reversed.begin()); qint64 value = *reinterpret_cast<const qint64*>(reversed.constData()); }@VRonin
Your "reversing" code involves actually copying the bytes in reverse order into a temporary area to "convert" it to a nativeint64. This is hideously inefficient if you have a file with billions of numbers in it :) Isn't there a solution which swaps bytes directly (e.g. so could be done on a register directly)? -
@VRonin
Your "reversing" code involves actually copying the bytes in reverse order into a temporary area to "convert" it to a nativeint64. This is hideously inefficient if you have a file with billions of numbers in it :) Isn't there a solution which swaps bytes directly (e.g. so could be done on a register directly)? -
@VRonin
LOL, of course you're smart enough!I was thinking of e.g. https://linux.die.net/man/3/htonl. These functions are "byte-swappers", and they were defined in macros with lots of
<<&>>s :) But they only go up toint32size :(EDIT I found this sample:
unsigned long int Endian_DWord_Conversion(unsigned long int dword) { return ((dword>>24)&0x000000FF) | ((dword>>8)&0x0000FF00) | ((dword<<8)&0x00FF0000) | ((dword<<24)&0xFF000000); }It's a question of expanding that approach from 4 to 8 bytes, and marking it
inline! -
@mpergand said in Problem converting QByteArray to long:
What about using QDataStream
Since QDataStream is a stream format it will not only store the 4 or 8 bytes but also some information to know what is saved - therefore it's totally useless here.
-
Hi,
Another important point, you don't check that you received all your data. The
readyReaddoesn't signal that the socket got the complete payload, only there's something that arrived. Your payload might be small enough now but it might very well change in the near future. -
@nico162
First let's start by having your actual code. Do you really have:for(int i=0;i<8;i++){ qInfo()<<(qint64)hashKeyValue[0]; }or would you like to correct that in either code or pasting here?
Second, have a look at the docs & example for http://doc.qt.io/qt-5/qbytearray.html#toLong. It's not doing anything like you're expecting it to do: it's for parsing a byte-array-string to a long, not creating a long out of a number already sitting there in binary bytes!
-
-
First of all, you need to understand endianness
Matching endianness between java and C++
void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); qint64 value = *reinterpret_cast<const qint64*>(msg.constData()); }Mismatching endianness between java and C++
void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); QByteArray reversed(sizeof(qint64),0); std::copy(msg.crbegin(),msg.crend(),reversed.begin()); qint64 value = *reinterpret_cast<const qint64*>(reversed.constData()); }@VRonin
Yes I though it could be that, but the bytes sent from Java to C++, are of same value and in the same order at the sending and receiving (I print the content in java and in C++, so I know the structure of the byte array).
So I admit that it wasn't an endianness error (maybe I am wrong ..). -
First of all, you need to understand endianness
Matching endianness between java and C++
void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); qint64 value = *reinterpret_cast<const qint64*>(msg.constData()); }Mismatching endianness between java and C++
void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); QByteArray reversed(sizeof(qint64),0); std::copy(msg.crbegin(),msg.crend(),reversed.begin()); qint64 value = *reinterpret_cast<const qint64*>(reversed.constData()); } -
@nico162 said in Problem converting QByteArray to long:
@VRonin
Ah no okay I misunderstood... Now I see what you mean :D! And it works fineSo as @JonB explained the QByteArray::toLong() is not adapted to my need (to convert a byte array to long).
@VRonin gave a solution to interpret the byteArray in the correct endianness mode (from Java to C++):void Receiver::on_receivingByteBuffMessage(const QByteArray& msg){ Q_ASSERT(msg.size()==sizeof(qint64)); QByteArray reversed(sizeof(qint64),0); std::copy(msg.crbegin(),msg.crend(),reversed.begin()); qint64 value = *reinterpret_cast<const qint64*>(reversed.constData()); }Maybe better solution could exist to avoid create a temp array to copy the byte in the reverse order as mentioned @JonB .
I personally have little bytebuffer received, so the solution that @VRonin gave it's enough efficient in my use case, that's why I close this topic.
@SGaist I take in account your remark thank you :).
I thank you all for your help :)! First time I posted a topic, and I am happy to see that there is an helpful and comprehensive community.See ya !
-
@VRonin
Your "reversing" code involves actually copying the bytes in reverse order into a temporary area to "convert" it to a nativeint64. This is hideously inefficient if you have a file with billions of numbers in it :) Isn't there a solution which swaps bytes directly (e.g. so could be done on a register directly)?@JonB said in Problem converting QByteArray to long:
Isn't there a solution which swaps bytes directly (e.g. so could be done on a register directly)?
Would by chance _byteswap_uint64 or __builtin_bswap64 do the trick?
`They did not know it was impossible, so they did it.'
-- Mark Twain -
@JonB said in Problem converting QByteArray to long:
Isn't there a solution which swaps bytes directly (e.g. so could be done on a register directly)?
Would by chance _byteswap_uint64 or __builtin_bswap64 do the trick?
@JohanSolo
They would indeed, but each of those is MSVC/GCC specific ... :( -
@JonB
I know, that's why I gave the two of them... Not the first time a#ifdef WIN32would help.@JohanSolo
Yeah, but I think (for Qt) we still have to support MinGW and whatever on MacOS and ...Unlike the functions for swapping 16 or 32 bytes, I couldn't find standard ones for 64 bytes anywhere when I searched. Shame!
