better compatibility with RegressionFormulae.jl

NEWS.md

@@ -1,3 +1,7 @@
+MixedModels v4.12.0 Release Notes
+==============================
+* The pirated method `Base.:/(a::AbstractTerm, b::AbstractTerm)` is no longer defined. This does not impact the use of `/` as a nesting term in `@formula` within MixedModels, only the programmatic runtime construction of formula, e.g. `term(:a) / term(:b)`. If you require `Base.:/`, then [`RegressionFormulae.jl`](https://github.com/kleinschmidt/RegressionFormulae.jl) provides this method. (Avoiding method redefinition when using `RegressionFormulae.jl` is the motivating reason for this change.) [#677]
+
 MixedModels v4.11.0 Release Notes
 ==============================
 * `raneftables` returns a `NamedTuple` where the names are the grouping factor names and the values are some `Tables.jl`-compatible type.  Currently this type is a `DictTable` from `TypedTables.jl`. [#634]
@@ -412,3 +416,4 @@ Package dependencies
 [#667]: https://github.com/JuliaStats/MixedModels.jl/issues/667
 [#674]: https://github.com/JuliaStats/MixedModels.jl/issues/674
 [#676]: https://github.com/JuliaStats/MixedModels.jl/issues/676
+[#677]: https://github.com/JuliaStats/MixedModels.jl/issues/677

Project.toml

@@ -1,7 +1,7 @@
 name = "MixedModels"
 uuid = "ff71e718-51f3-5ec2-a782-8ffcbfa3c316"
 author = ["Phillip Alday <me@phillipalday.com>", "Douglas Bates <dmbates@gmail.com>", "Jose Bayoan Santiago Calderon <jbs3hp@virginia.edu>"]
-version = "4.11.0"
+version = "4.12.0"
 
 [deps]
 Arrow = "69666777-d1a9-59fb-9406-91d4454c9d45"

src/randomeffectsterm.jl

@@ -139,8 +139,10 @@ function _ranef_refs(
     return refs, uniques
 end
 
-# TODO: split this off into a RegressionFormula package?
-Base.:/(a::AbstractTerm, b::AbstractTerm) = a + a & b
+# TODO: remove all of this and either
+# - require users to use RegressionFormulae.jl
+# - add a dependency on RegressionFormulae.jl
+
 function StatsModels.apply_schema(
     t::FunctionTerm{typeof(/)}, sch::StatsModels.FullRank, Mod::Type{<:MixedModel}
 )

test/FactorReTerm.jl

@@ -245,31 +245,11 @@ end
         @test modelcols(f2.rhs, d2) == [ones(20) d2.b .== :Y (d2.b .== :X).*d2.a (d2.b .== :Y).*d2.a]
         @test coefnames(f2.rhs) == ["(Intercept)", "b: Y", "b: X & a", "b: Y & a"]
 
-        # runtime
-        fr2 = term(1) + term(:b) / term(:a)
-        @test length(fr2) == 3
-        @test fr2[1] isa ConstantTerm
-        @test fr2[2] isa Term
-        @test fr2[3] isa InteractionTerm
-        frf2 = apply_schema(term(0) ~ fr2, schema(d2), MixedModel)
-        @test modelcols(frf2.rhs, d2) == [ones(20) d2.b .== :Y (d2.b .== :X).*d2.a (d2.b .== :Y).*d2.a]
-        @test coefnames(frf2.rhs) == ["(Intercept)", "b: Y", "b: X & a", "b: Y & a"]
-
         # check promotion
         f3 = apply_schema(@formula(0 ~ 0 + b/a), schema(d2), MixedModel)
         @test modelcols(f3.rhs, d2) == [d2.b .== :X d2.b .== :Y (d2.b .== :X).*d2.a (d2.b .== :Y).*d2.a]
         @test coefnames(f3.rhs) == ["b: X", "b: Y", "b: X & a", "b: Y & a"]
 
-        # runtime:
-        fr3 = term(0) + term(:b) / term(:a)
-        @test length(fr3) == 3
-        @test fr3[1] isa ConstantTerm
-        @test fr3[2] isa Term
-        @test fr3[3] isa InteractionTerm
-        ffr3 = apply_schema(term(0) ~ fr3, schema(d2), MixedModel)
-        @test modelcols(ffr3.rhs, d2) == [d2.b .== :X d2.b .== :Y (d2.b .== :X).*d2.a (d2.b .== :Y).*d2.a]
-        @test coefnames(ffr3.rhs) == ["b: X", "b: Y", "b: X & a", "b: Y & a"]
-
         # errors for continuous grouping
         @test_throws ArgumentError apply_schema(@formula(0 ~ 1 + a/b), schema(d2), MixedModel)
 
@@ -280,12 +260,11 @@ end
         psts = dataset("pastes")
         m = fit(MixedModel, @formula(strength ~ 1 + (1|batch/cask)), psts; progress=false)
         m2 = fit(MixedModel, @formula(strength ~ 1 + (1|batch) + (1|batch&cask)), psts; progress=false)
-        mr = fit(MixedModel, term(:strength) ~ term(1) + (term(1)|term(:batch)/term(:cask)), psts; progress=false)
         m2r = fit(MixedModel, term(:strength) ~ term(1) + (term(1)|term(:batch)) + (term(1)|term(:batch)&term(:cask)), psts; progress=false)
 
-        @test fnames(m) == fnames(m2) == fnames(mr) == fnames(m2r) == (Symbol("batch & cask"), :batch)
-        @test m.λ == m2.λ == mr.λ == m2r.λ
-        @test deviance(m) == deviance(m2) == deviance(mr) == deviance(m2r)
+        @test fnames(m) == fnames(m2) == fnames(m2r) == (Symbol("batch & cask"), :batch)
+        @test m.λ == m2.λ == m2r.λ
+        @test deviance(m) == deviance(m2) == deviance(m2r)
     end
 
     @testset "multiple terms with same grouping" begin