Alexander Plavin

> I expect from y = set(x, f, val) that it satisfies the lens laws Yeah, that part is clear – but I don't think lens laws specify what happens...

I'm not too familiar with optimal precompilation setups, but personally won't be against a PR that provides measurable improvements to TTFX. Please, just ensure that precompile times don't become unreasonably...

Indeed, `delete(x, @optic _["a"][∗])` could remove all elements of `x["a"]` so that `x["a"]` exists but is an empty collection. And `delete(x, @optic _["a"][∗][∗])` would keep all `x["a"][1]`, `x["a"][2]`, ... leaving...

Probably `const AllLensTypes = Union{PropertyLens, IndexLens, Function}` is a reasonable approximation. Out of curiosity, do you have an example where you really need to constrain the type like this?

This limit style does `SELECT TOP n` rendering, right? I think it was introduced and popularized by Microsoft as part of their T-SQL dialect. Would probably be more understandable to...

Oh, didn't know those differences! I think I only used TOP selection without any offsets, just to get a few rows and see if they look reasonable – before loading...

Btw, this kind of stuff works with AccessorsExtra – not with constants like `true` but with actual functions on each side: ```julia julia> using AccessorsExtra julia> @o _.a && _.b...

Currently: yes, it pirates some internal function that does actual `@o` parsing. But piracy isn't required to achieve the same end result: AccessorsExtra can in principle define and export its...

Agree, it would be both more familiar to many readers and less cluttered to show units once per axes – https://github.com/MakieOrg/Makie.jl/issues/3890.

Note that `obj1` and `objs` are not fully symmetric in such a function, they cannot in general be swapped. In particular, `obj1` should always define the resulting type. For `Properties()`,...