Allow computing estimators directly on a PMCSimulation

docs/src/index.md

@@ -78,7 +78,7 @@ julia> using Revise
 Rimu offers a number of tools for representing Hamiltonians (see
 [`Hamiltonians`](@ref)) and state vectors / wave functions
 (see [`DictVectors`](@ref))
-as well as algorithms to find the ground state, e.g. [`lomc!`](@ref).
+as well as algorithms to find the ground state, e.g. [`ProjectorMonteCarloProblem`](@ref), [`ExactDiagonalizationProblem`](@ref).
 
 ## Scripts
 
@@ -99,7 +99,7 @@ distribute work by making use of MPI, or "message passing interface". For exampl
 ```
 > julia scripts/BHM-example.jl
 ```
-will run on one processor with the main `lomc!()` computation (i.e. after
+will run on one processor with the main computation (i.e. after
 package loading and compilation) completing in 2.69 seconds.
 
 Running
@@ -110,10 +110,10 @@ on the same hardware makes use of 4 cores and the main part completes in 1.04
 seconds, a speedup factor of 2.6. This seems reasonable, given that extra work
 needs to be done for communicating between different processes.
 
-Using MPI parallelism with `Rimu` is easy. Enabling MPI for use in [`lomc!()`](@ref) is
-enabled automatically if [`PDVec`](@ref) is used to store a vector. In that case, data will
-be stored in a distributed fashion among the MPI ranks and only communicated between ranks
-when necessary. Additional MPI-related functionality is provided by the module [`RMPI`](@ref
+Using MPI parallelism with `Rimu` is easy. Enabling MPI enabled automatically if
+[`PDVec`](@ref) is used to store a vector. In that case, data will be stored in a
+distributed fashion among the MPI ranks and only communicated between ranks when
+necessary. Additional MPI-related functionality is provided by the module [`RMPI`](@ref
 Rimu.RMPI).
 
 ## Compatibility

docs/src/statstools.md

@@ -18,11 +18,11 @@ sample means is done with [`ratio_of_means`](@ref), where error propagation
 of correlated uncertainties is done with the help of the package
 [`MonteCarloMeasurements`](https://github.com/baggepinnen/MonteCarloMeasurements.jl).
 
-Many convenience functions are implemented for directly analysing data
-obtained from [`lomc!`](@ref) as a `DataFrame`. See, e.g.,
-[`shift_estimator`](@ref) and [`projected_energy`](@ref). Asymptotically
-unbiased estimators are implemented as [`mixed_estimator`](@ref),
-[`growth_estimator`](@ref) and [`rayleigh_replica_estimator`](@ref).
+Many convenience functions are implemented for directly analysing data obtained from
+[`solve`](@ref CommonSolve.solve(::ProjectorMonteCarloProblem)) as a `DataFrame`. See, e.g.,
+[`shift_estimator`](@ref) and [`projected_energy`](@ref). Asymptotically unbiased estimators
+are implemented as [`mixed_estimator`](@ref), [`growth_estimator`](@ref) and
+[`rayleigh_replica_estimator`](@ref).
 
 ## Exported
 ```@autodocs

src/StatsTools/StatsTools.jl

@@ -22,7 +22,7 @@ Tools for the statistical analysis of Monte Carlo data.
 """
 module StatsTools
 
-using DataFrames: DataFrames, DataFrame
+using DataFrames: DataFrames, DataFrame, metadata
 using Distributions: Distributions, Chisq, Distribution, MvNormal, Normal,
     cquantile, var
 using LinearAlgebra: LinearAlgebra, diag, norm

src/StatsTools/fidelity.jl

@@ -1,12 +1,12 @@
 """
     replica_fidelity(df::DataFrame; p_field = :hproj, skip = 0)
+    replica_fidelity(sim::PMCSimulation; kwargs...)
+
 Compute the fidelity of the average coefficient vector and the projector defined in
-`p_field` from the result of replica [`lomc!`](@ref Main.lomc!) passed as  argument `df`,
-using replicas `_1` and `_2`.
-Calls [`ratio_of_means()`](@ref) to perform a blocking analysis
-on a ratio of the means of separate time series and returns a
-[`RatioBlockingResult`](@ref).
-The first `skip` steps in the time series are skipped.
+`p_field` from the [`PMCSimulation`](@ref Main.Rimu.PMCSimulation) or `DataFrame` returned
+by solve, using replicas `_1` and `_2`. Calls [`ratio_of_means`](@ref) to perform a
+blocking analysis on a ratio of the means of separate time series and returns a
+[`RatioBlockingResult`](@ref). The first `skip` steps in the time series are skipped.
 
 The fidelity of states `|ψ⟩` and `|ϕ⟩` is defined as
 ```math
@@ -22,7 +22,8 @@ where `v` is the projector specified by `p_field`, which is assumed to be normal
 unity with the two-norm (i.e. `v⋅v == 1`), and ``\\mathbf{c}_1`` and ``\\mathbf{c}_2``
 are two replica coefficient vectors.
 """
-function replica_fidelity(df::DataFrame; p_field = :hproj, skip = 0,  args...)
+function replica_fidelity(sim; p_field = :hproj, skip = 0,  args...)
+    df = DataFrame(sim)
     p_field_1 = Symbol(p_field, :_1)
     p_field_2 = Symbol(p_field, :_2)
     fid_num = conj(getproperty(df, p_field_1)) .* getproperty(df, p_field_2)

src/StatsTools/growth_witness.jl

@@ -1,9 +1,6 @@
-# growth_witness()
-# smoothen()
-
 """
-    growth_witness(shift::AbstractArray, norm::AbstractArray, dt, [b]; skip=0) -> g
-    growth_witness(df::DataFrame, [b]; skip=0) -> g
+    growth_witness(shift::AbstractArray, norm::AbstractArray, dt, [b]; skip=0)
+
 Compute the growth witness
 ```math
 G^{(n)} = S^{(n)} - \\frac{\\vert\\mathbf{c}^{(n+1)}\\vert -
@@ -36,14 +33,21 @@ function growth_witness(shift::AbstractArray, norm::AbstractArray, dt, b; kwargs
 end
 """
     growth_witness(df::DataFrame, [b]; shift=:shift, norm=:norm, dτ=df.dτ[end], skip=0)
-Calculate the growth witness directly from a `DataFrame` returned by
-[`lomc!`](@ref Main.lomc!). The keyword arguments `shift` and `norm`
-can be used to change the names of the relevant columns.
+    growth_witness(sim::PMCSimulation, [b]; kwargs...)
+
+Calculate the growth witness directly from the result (`DataFrame` or
+[`PMCSimulation`](@ref Main.Rimu.PMCSimulation)) of
+[`solve`](@ref CommonSolve.solve(::ProjectorMonteCarloProblem))ing a
+[`ProjectorMonteCarloProblem`](@ref Main.ProjectorMonteCarloProblem). The keyword arguments
+`shift` and `norm` can be used to change the names of the relevant columns.
 """
 function growth_witness(
-    df::DataFrame, b=Val(0);
-    shift=:shift, norm=:norm, dτ=df.dτ[end], kwargs...
+    sim, b=Val(0);
+    shift=:shift, norm=:norm, dτ=nothing, kwargs...
 )
+    df = DataFrame(sim)
+    dτ = determine_constant_time_step(df)
+
     shift_vec = getproperty(df, Symbol(shift))
     norm_vec = getproperty(df, Symbol(norm))
     return growth_witness(shift_vec, norm_vec, dτ, b; kwargs...)