Hi, I think we should at least _somehow_ deprecate operator<; but I'm not sure if it should be a special deprecation macro, or a special one. Regarding point 4: Unfortunately, it's not that easy: If userType is actually a built-in type, then QMetaType::hasRegisteredComparators returns false, and QMetaType compare will not work. I'm afraid that in this case you would actually have to call the deprecated operator< [1]; unless we provide a replacement functionality in 5.15.3.
[1] Though operator< would still behave different for floating point values than QVariant::compare, as the former does a fuzzy compare for floating point values, and the latter and a strict comparison. Fabian -- Fabian Kosmale Software Engineer The Qt Company GmbH Erich-Thilo-Str. 10 D-12489 Berlin fabian.kosm...@qt.io +49 1638686070 http://qt.io Geschäftsführer: Mika Pälsi, Juha Varelius, Jouni Lintunen Sitz der Gesellschaft: Berlin Registergericht: Amtsgericht Charlottenburg, HRB 144331 B -- ________________________________________ Von: Development <development-boun...@qt-project.org> im Auftrag von Giuseppe D'Angelo via Development <development@qt-project.org> Gesendet: Sonntag, 22. November 2020 18:09 An: development@qt-project.org Betreff: [Development] QVariant::operator< deprecation fallout QVariant::operator< has been deprecated in 5.15 because its broken semantics, unable to establish a strict weak order (as expected by "operator<". cf. VALUE.16 of QUIP-19). The problem with this deprecation is that it doesn't offer a replacement. Users that are still on Qt 5.15 and want to get rid of deprecation warnings won't be able to get rid of this one. (Whether they do this in order to upgrade to Qt 6 or just to keep their codebase clean is irrelevant). Only Qt 6 introduced a replacement (QVariant::compare()). == Solutions == 1) Un-deprecate operator<. The operator has been broken since forever, we're not going to fix it anyhow, we just expect users to understand its limitations (e.g. only use it between QVariants containing the same kind of data, for for instance in order to sort a model's data). 2) Change the deprecation macro to another one that isn't included in 5.15 deprecations -- rather, one has to opt-in for. The opt-in is supposed to be used only when targeting Qt 6, and not by projects which want to stay on 5.15. (This kind of macro could be applied to any other similar deprecation in Qt 5.15 that doesn't have a corresponding replacement.) 3) Actually add a replacement for operator<. Possibly something that can be mechanically ported to Q6Variant::compare via search&replace. 4) At least document how to avoid the warning, i.e. how to compare "safely" two QVariants. Is there enough public API for this? Thinking of a) compare the variants' userType(), different -> uncomparable b) check QMetaType::hasRegisteredComparators(type), false -> uncomparable c) call QMetaType::compare(v1.data(), v2.data(), v1.userType(), result) and return *result This should be almost identical in Qt 6 (only step b) doesn't exist). Opinions? -- Giuseppe D'Angelo | giuseppe.dang...@kdab.com | Senior Software Engineer KDAB (France) S.A.S., a KDAB Group company Tel. France +33 (0)4 90 84 08 53, http://www.kdab.com KDAB - The Qt, C++ and OpenGL Experts _______________________________________________ Development mailing list Development@qt-project.org https://lists.qt-project.org/listinfo/development
