precompilation

Project.toml

@@ -1,8 +1,10 @@
 name = "Unfold"
 uuid = "181c99d8-e21b-4ff3-b70b-c233eddec679"
+
 authors = ["Benedikt Ehinger <[email protected]>"]
 version = "0.8.8"
 
+
 [deps]
 BSplineKit = "093aae92-e908-43d7-9660-e50ee39d5a0a"
 CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
@@ -24,6 +26,7 @@ MLBase = "f0e99cf1-93fa-52ec-9ecc-5026115318e0"
 Missings = "e1d29d7a-bbdc-5cf2-9ac0-f12de2c33e28"
 OrderedCollections = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
 PooledArrays = "2dfb63ee-cc39-5dd5-95bd-886bf059d720"
+PrecompileTools = "aea7be01-6a6a-4083-8856-8a6e6704d82a"
 ProgressMeter = "92933f4c-e287-5a05-a399-4b506db050ca"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 RobustModels = "d6ea1423-9682-4bbd-952f-b1577cbf8c98"
@@ -35,6 +38,7 @@ StatsAPI = "82ae8749-77ed-4fe6-ae5f-f523153014b0"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 StatsFuns = "4c63d2b9-4356-54db-8cca-17b64c39e42c"
 StatsModels = "3eaba693-59b7-5ba5-a881-562e759f1c8d"
+Suppressor = "fd094767-a336-5f1f-9728-57cf17d0bbfb"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 Term = "22787eb5-b846-44ae-b979-8e399b8463ab"
 TimerOutputs = "a759f4b9-e2f1-59dc-863e-4aeb61b1ea8f"
@@ -73,6 +77,7 @@ MLBase = "0.9"
 Missings = "1"
 OrderedCollections = "1"
 PooledArrays = "1"
+PrecompileTools = "1"
 ProgressMeter = "1"
 Random = "1"
 RobustModels = "0.4,0.5,0.6"
@@ -84,6 +89,7 @@ StatsAPI = "1"
 StatsBase = "0.33,0.34"
 StatsFuns = "1"
 StatsModels = "0.6,0.7"
+Suppressor = "0.2"
 Tables = "1"
 Term = "2"
 TimerOutputs = "0.5"

src/Unfold.jl

@@ -31,6 +31,8 @@ using TypedTables # DataFrames loose the pooled array, so we have to do it diffe
 
 using Interpolations # for FIR duration scaling
 using ImageTransformations # for FIR duration scaling
+
+using Suppressor
 #using Tullio
 #using BSplineKit # for spline predictors
 
@@ -70,6 +72,8 @@ include("solver/main.jl")
 include("solver/solvers.jl")
 include("solver/prepare.jl")
 
+include("precompile.jl")
+
 #include("plot.jl") # don't include for now
 export fit, fit!, designmatrix!
 export firbasis, hrfbasis

src/precompile.jl

@@ -0,0 +1,48 @@
+using PrecompileTools: @setup_workload, @compile_workload
+
+
+
+
+@setup_workload begin
+    data = rand(Float64, 2, 100)
+    evts = DataFrame(
+        :latency => (range(10, 70, length = 3)),
+        :condition => [:a, :a, :b],
+        :continuous => 1:3,
+    )
+
+
+    @compile_workload begin
+
+        # cut the data into epochs
+        data_epochs, t = Unfold.epoch(data = data, tbl = evts, τ = (-0.1, 0.1), sfreq = 10) # channel x timesteps x trials
+        f = @formula 0 ~ 1 + condition + continuous # note the formulas left side is `0 ~ ` for technical reasons`
+
+
+        basisfunction = firbasis(τ = (-0.1, 0.1), sfreq = 10)
+        bf_vec = [Any => (f, basisfunction)]
+
+
+        for m in [
+            @suppress(fit(UnfoldModel, f, evts, data_epochs, t)),
+            @suppress(fit(UnfoldModel, [Any => (f, t)], evts, data_epochs)),
+            @suppress(
+                fit(
+                    UnfoldModel,
+                    [Any => (f, t)],
+                    Unfold.drop_missing_epochs(evts, data_epochs)[1],
+                    Unfold.drop_missing_epochs(evts, data_epochs)[2],
+                )
+            ),
+            @suppress(fit(UnfoldModel, bf_vec, evts, data)),
+        ]
+            coef(m)
+            coeftable(m)
+            effects(Dict(:condition => [:a, :b]), m)
+            effects(Dict(:continuous => 1:2), m)
+            effects(Dict(:continuous => collect(1:2)), m)
+
+        end
+
+    end
+end

src/utilities.jl

@@ -137,6 +137,7 @@ function equalize_size(
     _get_data(data::AbstractVector, n) = @view data[1:n]
 
     n = min(size(X, 1), size(data)[end])
+
     X_v = X[1:n, :] # with view we got performance problems downstream, hotfixing this here
     data_v = _get_data(data, n)
     return X_v, data_v

test/designmatrix.jl

@@ -1,5 +1,5 @@
 ##---
-
+using DataFrames, StatsModels
 tbl = DataFrame([1 4]', [:latency])
 X = ones(size(tbl))
 shouldBeNeg = zeros(4, 4)

test/predict.jl

@@ -214,6 +214,6 @@ end
     _r2 = Unfold.r2(m_reshape, data_reshape)
     @test length(_r2) == 1
     _r2 = Unfold.r2(m_e_reshape, data_e_reshape)
-    @test length(_r2) == 45
+    @test size(_r2) == (1, size(data_e_reshape, 2))
 
 end