I analyzed the usage of toLowerCase and toUpperCase in OpenJDK more carefully, and found that none of the use cases really expected locale-sensitive behavior, **all** use cases were misused without exception.
I think more third-party libraries and applications should be investigated to determine how common this situation is. If most of the use cases are misusing it, I think its behavior should be modified and marked as deprecated to avoid confusion. Then, we can create new APIs as their substitutes. On Thu, Jan 26, 2023 at 9:36 PM Remi Forax <fo...@univ-mlv.fr> wrote: > > > ------------------------------ > > *From: *"Glavo" <zjx001...@gmail.com> > *To: *"core-libs-dev" <core-libs-dev@openjdk.org> > *Sent: *Thursday, January 26, 2023 1:35:06 PM > *Subject: *[Proposal] Make toLowerCase and toUpperCase based on > Locale.ROOT by default > > At present, the no-parameter toLowerCase and toUpperCase methods of String > are based on the default locale. > > I checked all the uses of this method in OpenJDK, and found that most of > the use cases are suspicious, and even there are some hidden bugs. > For example, I just found that jdk.incubator.vector.LaneType will trigger > assertion failure on Turkish locale, and opened a PR for this problem: > > https://github.com/openjdk/panama-vector/pull/210 > > In addition to such obvious problems, some use cases behave suspiciously > after calling Locale.setDefault. > I am investigating these problems and preparing to open a PR to solve > these hidden bugs. > > In all the third-party libraries I have used, I have never seen the > correct use of these two methods. > Although the behavior of modifying API methods is destructive, I think it > is worthwhile to consider whether to modify its behavior for such a > suspicious method. > If users need locale-sensitive case conversion, it may be better to > explicitly use Locale.getDefault(). > Using Locale.ROOT as the default value also helps to keep the behavior of > these two methods consistent with equalsIgnoreCase, > Character.toLowerCase/toUpperCase and other methods. > > This is my rough idea. I hope to get your suggestions. > > > yes, it's a very very common bug, IntelliJ already warns when > toLowerCase()/toUpperCase() is used without a Locale. > I think the logical nest step is more to deprecate > toLowerCase()/toUpperCase() than to change their implementation. > > regards, > Rémi > >
