Use QLatin1String to optimize string comparisons in SolarSystem and Planet

[10110111]

In #3847 we can often see QUtf8::compareUtf8 appear in the profiler output. This PR makes string comparisons lighter due to the lack of need to convert from UTF-8 to UTF-16 or whatever the comparison function does.

The QLatin1String function is abbreviated to L1S by a #define in StelUtils, because oftentimes we have multiple string literals in a single if statement, for which using the full function name would make the line much longer.

This might make some of the changes in #3847 no longer important.

[github-actions[bot]]

Great PR! Please pay attention to the following items before merging:

Files matching src/**/*.cpp:

• [ ] Are possibly unused includes removed?

This is an automatically generated QA checklist based on modified files.

[gzotti]

Interesting! I had meanwhile considered using qHash(name) which may also compare faster. Before I run this: are there any dangers now using Latin1? I mean, all our text files including object names are now UTF8. But I think the string compare avoidances in #3847 are here to stay.

[10110111]

Before I run this: are there any dangers now using Latin1? I mean, all our text files including object names are now UTF8.

No, this only affects string literals. And all the places I changed have only Latin1 characters: like "Sun", "none" etc.

(Ideally, we should also edit other cpp files, but this PR only deals with the ones that seem to most frequently appear in the profiler output.)

[gzotti]

Ah, I misunderstood the intention of QLatin1String. But according to https://doc.qt.io/qt-6/qlatin1stringview.html I don't expect a change. Will profile a bit later. (Should this StelUtils macro become QLatin1StringView in Qt6 builds?)

[10110111]

I don't expect a change

Yes, that page doesn't make the main point explicit. The point is that QString("asdf")=="bbb" invokes QUtf8::compareUtf8, while QString("asdf")==QLatin1String("bbb") skips the comparison loop early by using the fact that in Latin1 encoding is fixed-length, implying that number of bytes is equal to string length. And even if the lengths of the strings compared coincide, it uses a simple loop in ucstrncmp instead of doing UTF-8 unpacking magic in QUtf8::compareUtf8.

[10110111]

Should this StelUtils macro become QLatin1StringView in Qt6 builds?

They might, but I don't think it will noticeably improve performance, since QLatin1String already avoids making a deep copy of the data. Note the documentation in the constructor from const char*:

The string data is not copied. The caller must be able to guarantee that str will not be deleted or modified as long as the QLatin1String object exists.

[gzotti]

Why not just apply it to #3847 ? I don't even want to profile master again after all these improvements.

[10110111]

Why not just apply it to #3847 ?

This change makes sense regardless of whether we'll accept all of the changes in that PR or only some of them, and I want to avoid cluttering the already hard to analyze diff there even further. I can (and actually will have to) add this commit there too for profiling, but I'm going to merge this one independently if there're no objections.

[github-actions[bot]]

Hello @10110111!

Please check the fresh version (development snapshot) of Stellarium: https://github.com/Stellarium/stellarium-data/releases/tag/weekly-snapshot

[github-actions[bot]]

Hello @10110111!

Please check the latest stable version of Stellarium: https://github.com/Stellarium/stellarium/releases/latest