Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB
-
@DerReisende said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
Where is the mistake and how would you come to a QStringRef/QStringView version on Qt 6.4?
QString::splitRef() - create a QList/Vector with references to parts of a QString, no QString objects are created, nothing gets copied
QString::split() - create a QList/Vector with newly create QString objectsTo avoid the creation of the QString objects and get the same behavior as before, use QStringView::split()0
wrote on 30 Sept 2022, 19:24 last edited by@Christian-Ehrlicher
Ok I modified it with the following://const auto result = line.split(semicolon); const QStringView sv{line}; const auto result = sv.split(semicolon);
Doesn't make a difference in runtime. And looking at the QString split code I am almost sure that line.split already does this optimization through implicit sharing through QStringList.
-
@Christian-Ehrlicher
Ok I modified it with the following://const auto result = line.split(semicolon); const QStringView sv{line}; const auto result = sv.split(semicolon);
Doesn't make a difference in runtime. And looking at the QString split code I am almost sure that line.split already does this optimization through implicit sharing through QStringList.
@DerReisende said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
I am almost sure that line.split already does this optimization through implicit sharing through QStringList.
No, implicit sharing of a container has nothing to do with creating new QString objects - QStringList is a list of string objects, not a list of QStringViews ...
-
@JonB Unfortunately, no. The solutions provided are by far leaps and bounds better than before, but the file sizes I'm expecting will definitely cause an issue if it's all done in Qt. I'm currently writing a Rust lib to resolve the problem for now, but even a 2-3x increase with Rust will make a monumental difference.
wrote on 30 Sept 2022, 19:28 last edited by JoeCFD@TheLumbee That is what I did with Java code before. When I got any bottleneck in my Java apps, I tried to use C/C++ code to do the jobs. Good luck!
-
wrote on 30 Sept 2022, 21:29 last edited by TheLumbee
Update
I successfully created the Rust lib and it's passing all the data to Qt. Rust converts the dateTime portion to mSecsSinceEpoch and then it passes that to Qt to create a DateTime object. Much faster that way.
This wasn't the preferred choice, but it does the job faster than I've ever seen. I do appreciate everyone's input and help. Hopefully in the future Qt would be able to match the performance of Rust with this particular issue.
Also, I'm going to run some tests with this Rust lib on Windows to see if it's still glacial compared to Unix.
-
wrote on 30 Sept 2022, 22:19 last edited by
Windows Update
After testing on Windows, I'm actually getting the same performance as Linux. So, it must be a C++ issue with Windows. I've tested with MinGW and MSVC with parsing and it's nearly unusable.
-
Windows Update
After testing on Windows, I'm actually getting the same performance as Linux. So, it must be a C++ issue with Windows. I've tested with MinGW and MSVC with parsing and it's nearly unusable.
@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
So, it must be a C++ issue with Windows
Again - use plain C++ and not Qt - the QDateTime parsing is painful slow...
-
@JonB @J-Hilk @DerReisende Thanks for all the responses! Didn't actually expect much here. I apologize for not providing more details. I've been dealing with this file parsing issue in C++ for years. Same code in Windows takes >100x times to complete rather than using Linux for some odd reason which I've posted in C++ forums prior to using Qt, but what you've provided is actually the first significant improvement I've ever seen.
So thank you for that!
I've tested this with versions 512, 5.15, 6.0, 6.2.4, and 6.4. Never noticed a major difference between them regarding this issue. Current machine: i7-6700 with 32GB RAM. So not sure what y'all are working with but the results seem promising.
I was previously streaming into a QTextStream then reading line-by-line but came across this post: https://forum.qt.io/topic/98282/parsing-large-big-text-files-quickly and a couple of others that suggested that is more expensive that using a QByteArray. I didn't notice much difference to be quite honest.
I did comment out the QDateTime parsing just to check and it was a significant improvement. Not quite like Rust but I'll attribute that to @JonB comment:
That "naive" means it does not do any local time/daylight etc, conversions.
If any of you are interested, I'll test each of your solutions and provide an update. But this actually woke me up and got me excited to start my day so thank you.
wrote on 1 Oct 2022, 09:11 last edited by@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
I've tested this with versions 512, 5.15, 6.0, 6.2.4, and 6.4. Never noticed a major difference between them regarding this issue.
Do you have access to both Qt 6.4 and C++20? If so, can you try combining
https://doc.qt.io/qt-6/qdatetime.html#fromStdTimePoint-1 (QDateTime QDateTime::fromStdTimePoint(const std::chrono::local_time<std::chrono::milliseconds> &time)
)
https://en.cppreference.com/w/cpp/chrono/parse (std::chrono::parse()
)
to see whether the datetime conversion part now matches Rust's?For right or for wrong, I have appended a post into https://bugreports.qt.io/browse/QTBUG-97489 for this whole datetime parsing to
QDateTime
issue, as I am concerned it is a "show-stopper" if you have a large amountof string datetime data you need to get into Qt'sQDateTime
. -
Lifetime Qt Championwrote on 1 Oct 2022, 13:42 last edited by Christian Ehrlicher 10 Jan 2022, 13:43
Ok, some c++20 magic to use c++ instead c, but not really optimized
testFile.open(QFile::ReadOnly); instr.tickList.clear(); instr.tickList.reserve(1000000); QElapsedTimer parseTimer1; parseTimer1.start(); const QByteArrayList allData = testFile.readAll().split('\n'); for (const auto &line : allData) { const QByteArrayList data = line.split(';'); std::string str = data.at(0).data(); // TODO: use data.at(0).toStdString() str[15] = '.'; // sadly needed for correct msec parsing std::istringstream stream(str); std::chrono::sys_time<std::chrono::milliseconds> tTimePoint; std::chrono::from_stream(stream, "%Y%m%d %H%M%S", tTimePoint); instr.tickList.push_back({ QDateTime::fromMSecsSinceEpoch(tTimePoint.time_since_epoch().count()), data.at(1).toDouble(), data.at(2).toDouble(), data.at(3).toDouble(), data.at(4).toInt() }); } qDebug().noquote() << QString("Qt parse time: %1ms").arg(parseTimer1.elapsed());``` compared to @J-Hilk 's version: compiled with debug: Qt parse time: 32125ms Qt parse time: 154511ms compiled with release: Qt parse time: 17926ms Qt parse time: 157748ms Attention: Qt debug libs, the emplace_back() doesn't help much.
-
@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
So, it must be a C++ issue with Windows
Again - use plain C++ and not Qt - the QDateTime parsing is painful slow...
wrote on 1 Oct 2022, 13:53 last edited by@Christian-Ehrlicher Before I ever used Qt, I was facing the same issue with C++. I initially believed it was a filesystem difference and posted in a forum here: https://cplusplus.com/forum/general/254030/
Near the end of the thread you'll see that others noticed the same issue with Windows. I just find it odd that this blatant difference has never been noticed, at least in a major way.
-
@Christian-Ehrlicher Before I ever used Qt, I was facing the same issue with C++. I initially believed it was a filesystem difference and posted in a forum here: https://cplusplus.com/forum/general/254030/
Near the end of the thread you'll see that others noticed the same issue with Windows. I just find it odd that this blatant difference has never been noticed, at least in a major way.
wrote on 1 Oct 2022, 14:46 last edited by@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
Before I ever used Qt, I was facing the same issue with C++.
I do think this thread is getting confused. You certainly seem to talk in this thread about various different aspects of your speed with Rust/C++/Qt/Windows/file I/O all mixed into one. One has to deal with these separately. The issue @Christian-Ehrlicher and I, at least, are discussing now is specifically what to do about
QDateTime::fromString()
, which is by far the major contributor to your efficiency compared to Rust, other items are minor. The proposal is if one has Qt 6.4+ and C++ 20 thenstd::chrono
can be used to parse the string input to a "naive datetime" (and I have a hunch that is what Rust uses) and that converted to aQDateTime
in condirably better time thatQDateTime::fromString()
.This is quite distinct from e.g. the time taken to read the large file under Windows.
-
@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
Before I ever used Qt, I was facing the same issue with C++.
I do think this thread is getting confused. You certainly seem to talk in this thread about various different aspects of your speed with Rust/C++/Qt/Windows/file I/O all mixed into one. One has to deal with these separately. The issue @Christian-Ehrlicher and I, at least, are discussing now is specifically what to do about
QDateTime::fromString()
, which is by far the major contributor to your efficiency compared to Rust, other items are minor. The proposal is if one has Qt 6.4+ and C++ 20 thenstd::chrono
can be used to parse the string input to a "naive datetime" (and I have a hunch that is what Rust uses) and that converted to aQDateTime
in condirably better time thatQDateTime::fromString()
.This is quite distinct from e.g. the time taken to read the large file under Windows.
wrote on 1 Oct 2022, 15:21 last edited by@JonB said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
The issue @Christian-Ehrlicher and I, at least, are discussing now is specifically what to do about
QDateTime::fromString()
, which is by far the major contributor to your efficiency compared to Rust, other items are minor. The proposal is if one has Qt 6.4+ and C++ 20 thenstd::chrono
can be used to parse the string input to a "naive datetime" (and I have a hunch that is what Rust uses) and that converted to aQDateTime
in condirably better time thatQDateTime::fromString()
.Apologies. I don't have C++20, but I can set up an environment to test it. But even without the DateTime, Rust is parsing the file 2-3x faster than C++, with just floats and ints. Maybe C++20 has some improvements in that domain, but I'll set up an environment to test this.
-
@JonB said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
The issue @Christian-Ehrlicher and I, at least, are discussing now is specifically what to do about
QDateTime::fromString()
, which is by far the major contributor to your efficiency compared to Rust, other items are minor. The proposal is if one has Qt 6.4+ and C++ 20 thenstd::chrono
can be used to parse the string input to a "naive datetime" (and I have a hunch that is what Rust uses) and that converted to aQDateTime
in condirably better time thatQDateTime::fromString()
.Apologies. I don't have C++20, but I can set up an environment to test it. But even without the DateTime, Rust is parsing the file 2-3x faster than C++, with just floats and ints. Maybe C++20 has some improvements in that domain, but I'll set up an environment to test this.
wrote on 1 Oct 2022, 15:30 last edited by@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
But even without the DateTime, Rust is parsing the file 2-3x faster than C++, with just floats and ints.
I do understand this. But I suggest this is a separate issue from the
QDateTime
. You started with 40x faster. Dealing withQDateTime
is the first priority. File reading or parsing ints and floats is a separate issue requiring its own solution. -
@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
But even without the DateTime, Rust is parsing the file 2-3x faster than C++, with just floats and ints.
I do understand this. But I suggest this is a separate issue from the
QDateTime
. You started with 40x faster. Dealing withQDateTime
is the first priority. File reading or parsing ints and floats is a separate issue requiring its own solution. -
wrote on 6 Oct 2022, 15:34 last edited by
To provide an update, the solution provided by @JonB 's "final offering" in this post parses files a little quicker. Although better than my initial approach, still not nearly as fast as the Rust solution. For now, I'm sticking with the Rust lib I wrote but I do think this points out some performance enhancements that can be made on the C++ side of things.
-
To provide an update, the solution provided by @JonB 's "final offering" in this post parses files a little quicker. Although better than my initial approach, still not nearly as fast as the Rust solution. For now, I'm sticking with the Rust lib I wrote but I do think this points out some performance enhancements that can be made on the C++ side of things.
wrote on 6 Oct 2022, 17:05 last edited by@TheLumbee said in Rust file parsing significantly faster than Qt/C++ file parsing. Solutions for Qt implementation wanted. File size: 68.5 MB:
Although
For MSVC++ you should write a bug report for them with test data. There seems to be a major performance issue with the MS c++ library as it is much slower than other solutions. Maybe they can provide an update in a future version or hint to workarounds to improve performance on windows.
-
wrote on 11 Mar 2023, 00:43 last edited by
Hello. I know this is an old topic.
I couldn't test as the test file doesn't seem accessible anymore.
Assuming you were testing on Linux with glibc, could you try setting the TZ environment variable? e.g.:
export TZ=":/etc/localtime"See also:
https://sourceware.org/bugzilla/show_bug.cgi?id=24004
https://bugreports.qt.io/browse/QTBUG-77948