Merge branch 'feature/AllOverlaps-excited-states' of https://github.c…

…om/joachimbrand/Rimu.jl into feature/AllOverlaps-excited-states

Project.toml

@@ -67,7 +67,7 @@ FFTW = "1"
 Folds = "0.2"
 HypergeometricFunctions = "0.3"
 IterativeSolvers = "0.9"
-KrylovKit = "0.6, 0.7, 0.8"
+KrylovKit = "0.6, 0.7, 0.8, 0.9"
 MPI = "0.20"
 MacroTools = "0.5"
 Measurements = "2"

src/StatsTools/growth_witness.jl

@@ -51,7 +51,7 @@ function growth_witness(
     shift=:shift, norm=:norm, time_step=nothing, kwargs...
 )
     df = DataFrame(sim)
-    time_step = determine_constant_time_step(df)
+    time_step = isnothing(time_step) ? determine_constant_time_step(df) : time_step
 
     shift_vec = getproperty(df, Symbol(shift))
     norm_vec = getproperty(df, Symbol(norm))

src/StatsTools/reweighting.jl

@@ -22,7 +22,7 @@ function determine_constant_time_step(df)
         elseif hasproperty(df, "time_step_1")
             return df.time_step_1[end]
         else
-            throw(ArgumentError("Time step not found in `df`"))
+            throw(ArgumentError("key `\"time_step\"` not found in `df` metadata"))
         end
     else
         throw(ArgumentError("Time step not constant"))
@@ -221,6 +221,7 @@ Returns a `NamedTuple` with the fields
 * `threading = Threads.nthreads() > 1`: if `false` a progress meter is displayed
 * `shift_name = :shift` name of column in `df` with shift data
 * `norm_name = :norm` name of column in `df` with walkernumber data
+* `time_step = determine_constant_time_step(df)` the time step
 * `warn = true` whether to log warning messages when blocking fails or denominators are
   small
 
@@ -244,14 +245,15 @@ function growth_estimator_analysis(
     threading=Threads.nthreads() > 1,
     shift_name=:shift,
     norm_name=:norm,
+    time_step=nothing,
     warn=true,
     kwargs...
 )
     df = DataFrame(sim)
     shift_v = Vector(getproperty(df, Symbol(shift_name))) # casting to `Vector` to make SIMD loops efficient
     norm_v = Vector(getproperty(df, Symbol(norm_name)))
     num_reps = length(filter(startswith("norm"), names(df)))
-    time_step = determine_constant_time_step(df)
+    time_step = isnothing(time_step) ? determine_constant_time_step(df) : time_step
     se = blocking_analysis(shift_v; skip)
     E_r = se.mean
     correlation_estimate = 2^(se.k - 1)
@@ -390,6 +392,7 @@ Returns a `NamedTuple` with the fields
 * `shift_name = :shift` name of column in `df` with shift data
 * `hproj_name = :hproj` name of column in `df` with operator overlap data
 * `vproj_name = :vproj` name of column in `df` with projector overlap data
+* `time_step = determine_constant_time_step(df)` the time step
 * `warn = true` whether to log warning messages when blocking fails or denominators are small
 
 ## Example
@@ -414,14 +417,15 @@ function mixed_estimator_analysis(
     hproj_name=:hproj,
     vproj_name=:vproj,
     warn=true,
+    time_step=nothing,
     kwargs...
 )
     shift_v = Vector(getproperty(df, Symbol(shift_name))) # casting to `Vector` to make SIMD loops efficient
     hproj_v = Vector(getproperty(df, Symbol(hproj_name)))
     vproj_v = Vector(getproperty(df, Symbol(vproj_name)))
     num_reps = length(filter(startswith("norm"), names(df)))
 
-    time_step = determine_constant_time_step(df)
+    time_step = isnothing(time_step) ? determine_constant_time_step(df) : time_step
     se = blocking_analysis(shift_v; skip)
     E_r = se.mean
     correlation_estimate = 2^(se.k - 1)
@@ -560,11 +564,12 @@ function rayleigh_replica_estimator(
     h=0,
     skip=0,
     Anorm=1,
+    time_step=nothing,
     kwargs...
 )
     df = DataFrame(sim)
     num_reps = parse(Int, metadata(df, "num_replicas"))
-    time_step = determine_constant_time_step(df)
+    time_step = isnothing(time_step) ? determine_constant_time_step(df) : time_step
     T = eltype(df[!, Symbol(shift_name, "_r1s1")])
     shift_v = Vector{T}[]
     for a in 1:num_reps
@@ -632,11 +637,12 @@ function rayleigh_replica_estimator_analysis(
     spectral_state=1,
     Anorm=1,
     warn=true,
+    time_step=nothing,
     kwargs...
 )
     df = DataFrame(sim)
     num_reps = parse(Int, metadata(df, "num_replicas"))
-    time_step = determine_constant_time_step(df)
+    time_step = isnothing(time_step) ? determine_constant_time_step(df) : time_step
     # estimate the correlation time by blocking the shift data
     T = eltype(df[!, Symbol(shift_name, "_r1s1")])
     shift_v = Vector{T}[]