Rainer Schuetze

> Right. I think we can accept that if someone does that, it's intentional and they know what they're doing... Agreed. Wanted to use this feature on the dmd frontend,...

This is now available in https://github.com/dlang/visuald/releases/tag/v1.4.1-beta4

This now passes all the tests with all associative array operations templated and the runtime implementation removed. It is probably a bit large for review, though. I have rebased and...

> It is, but I think it's worth pulling in one go, unless you want to put in the work to split this up into incremental changes. A lot of...

> > > AA no longer available at CTFE without object.d/newaa.d. There was some support for this, but no more. Is this a problem? > > > > I don't...

I looked at all issues labeled AA, these are the ones that should be fixed by this PR: #17455 Reimplement associative arrays on runtime #17412 Associative array opApply is not...

> #17235 [AA] Associative array with enum array keys is slow When trying the benchmark of #17235 I noticed one possible downside: with non-optimized builds: using AAs extensively will result...

> Roughly how much? When I compare the test from the report (on `immutable(char)`, not the slow enum versiob) with dmd-2.111, it takes indeed almost twice as much time: dmd-2.111:...

I have increased the loop count in https://github.com/dlang/dmd/issues/17235 by a factor 10 to improve profiling and added some `pragma(inline, true)` after applying https://github.com/dlang/dmd/pull/21524. This reduces the gap for debug builds...

> but expression precedence information might be lost lost if operator translated to call. This is fixed by `expPrecedence` that uses the precedence of the added `loweredFrom` expression of the...