Merge pull request #100 from Byrth/enhancement/reduce_cd_dependence

Switch to absolute path
StanJulia · Aug 22, 2020 · 3b6d5ad · 3b6d5ad
2 parents 98d5754 + 8212224
commit 3b6d5ad
Show file tree

Hide file tree

Showing 10 changed files with 130 additions and 127 deletions.
diff --git a/examples/BernoulliOptimize/bernoulli_optimize.jl b/examples/BernoulliOptimize/bernoulli_optimize.jl
@@ -30,7 +30,7 @@ cd(ProjDir) #do
     println()
     display(optim)
     println()
-    println("Test round.(mean(optim[1][\"theta\"]), digits=1) ≈ 0.3")
+    println("Test round.(mean(optim[\"theta\"]), digits=1) ≈ 0.3")
     @test round.(mean(optim["theta"]), digits=1) ≈ 0.3
   end
 

diff --git a/src/main/stancode.jl b/src/main/stancode.jl
@@ -96,11 +96,12 @@ function stan(
   @assert isdir(ProjDir) "Incorrect ProjDir specified: $(ProjDir)"
   @assert isdir(model.tmpdir) "$(model.tmpdir) not created"
 
-  cd(model.tmpdir) do 
-    isfile("$(model.name)_build.log") && rm("$(model.name)_build.log")
-    isfile("$(model.name)_make.log") && rm("$(model.name)_make.log")
-    isfile("$(model.name)_run.log") && rm("$(model.name)_run.log")
-  end
+  absolute_tempdir_path = cd(pwd,model.tmpdir)
+  path_prefix = joinpath(splitpath(absolute_tempdir_path)..., model.name)
+  model.object_file = path_prefix
+  isfile("$(path_prefix)_build.log") && rm("$(path_prefix)_build.log")
+  isfile("$(path_prefix)_make.log") && rm("$(path_prefix)_make.log")
+  isfile("$(path_prefix)_run.log") && rm("$(path_prefix)_run.log")
 
   cd(CmdStanDir) do
     local tmpmodelname::String
@@ -126,138 +127,136 @@ function stan(
     end
   end
 
-  cd(model.tmpdir) do  
-    if data != Nothing && (typeof(data) <: AbstractString || check_dct_type(data))
-      if typeof(data) <: AbstractString
+  if data != Nothing && (typeof(data) <: AbstractString || check_dct_type(data))
+    if typeof(data) <: AbstractString
+      for i in 1:model.nchains
+        cp(data, "$(path_prefix)_$(i).data.R", force=true)
+      end   
+    else
+      if typeof(data) <: Array && length(data) == model.nchains
         for i in 1:model.nchains
-          cp(data, "$(model.name)_$(i).data.R", force=true)
-        end   
+          if length(keys(data[i])) > 0
+            update_R_file("$(path_prefix)_$(i).data.R", data[i])
+          end
+        end
       else
-        if typeof(data) <: Array && length(data) == model.nchains
+        if typeof(data) <: Array
           for i in 1:model.nchains
-            if length(keys(data[i])) > 0
-              update_R_file("$(model.name)_$(i).data.R", data[i])
+            if length(keys(data[1])) > 0
+              update_R_file("$(path_prefix)_$(i).data.R", data[1])
             end
           end
         else
-          if typeof(data) <: Array
-            for i in 1:model.nchains
-              if length(keys(data[1])) > 0
-                update_R_file("$(model.name)_$(i).data.R", data[1])
-              end
-            end
-          else
-            for i in 1:model.nchains
-              if length(keys(data)) > 0
-                update_R_file("$(model.name)_$(i).data.R", data)
-              end
+          for i in 1:model.nchains
+            if length(keys(data)) > 0
+              update_R_file("$(path_prefix)_$(i).data.R", data)
             end
           end
         end
       end
     end
-
-    if init != Nothing && (typeof(init) <: AbstractString || check_dct_type(init))
-      if typeof(init) <: AbstractString
+  end
+
+  if init != Nothing && (typeof(init) <: AbstractString || check_dct_type(init))
+    if typeof(init) <: AbstractString
+      for i in 1:model.nchains
+        cp(init, "$(path_prefix)_$(i).init.R", force=true)
+      end   
+    else
+      if typeof(init) <: Array && length(init) == model.nchains
         for i in 1:model.nchains
-          cp(init, "$(model.name)_$(i).init.R", force=true)
-        end   
+          if length(keys(init[i])) > 0
+            update_R_file("$(path_prefix)_$(i).init.R", init[i])
+          end
+        end
       else
-        if typeof(init) <: Array && length(init) == model.nchains
+        if typeof(init) <: Array
           for i in 1:model.nchains
-            if length(keys(init[i])) > 0
-              update_R_file("$(model.name)_$(i).init.R", init[i])
+            if length(keys(init[1])) > 0
+              update_R_file("$(path_prefix)_$(i).init.R", init[1])
             end
           end
         else
-          if typeof(init) <: Array
-            for i in 1:model.nchains
-              if length(keys(init[1])) > 0
-                update_R_file("$(model.name)_$(i).init.R", init[1])
-              end
-            end
-          else
-            for i in 1:model.nchains
-              if length(keys(init)) > 0
-                update_R_file("$(model.name)_$(i).init.R", init)
-              end
+          for i in 1:model.nchains
+            if length(keys(init)) > 0
+              update_R_file("$(path_prefix)_$(i).init.R", init)
             end
           end
         end
       end
     end
+  end
+
+  for i in 1:model.nchains
+    model.id = i
+    model.data_file ="$(path_prefix)_$(i).data.R"
+    if init != Nothing
+        model.init_file = "$(path_prefix)_$(i).init.R"
+    end
+    if isa(model.method, Sample)
+      model.output.file = path_prefix*"_samples_$(i).csv"
+      isfile(model.output.file) && rm(model.output.file)
+      if diagnostics
+        model.output.diagnostic_file = path_prefix*"_diagnostics_$(i).csv"
+        isfile(model.output.diagnostic_file) && rm(model.output.diagnostic_file)
+      end
+    elseif isa(model.method, Optimize)
+      model.output.file = path_prefix*"_optimize_$(i).csv"
+      isfile(model.output.file) && rm(model.output.file)
+    elseif isa(model.method, Variational)
+      model.output.file = path_prefix*"_variational_$(i).csv"
+      isfile(model.output.file) && rm(model.output.file)
+    elseif isa(model.method, Diagnose)
+      model.output.file = path_prefix*"_diagnose_$(i).csv"
+      isfile(model.output.file) && rm(model.output.file)
+    end
+    model.command[i] = cmdline(model)
+  end
+
+  try
+    if file_run_log
+      run(pipeline(par(model.command), stdout="$(path_prefix)_run.log"))
+    else
+      run(par(model.command))
+    end
+  catch e
+    println("\nAn error occurred while running the previously compiled Stan program.\n")
+    print("Please check the contents of file $(model.name)_run.log and the")
+    println("'command' field in the Stanmodel, e.g. stanmodel.command.\n")
+    error("Return code = -5")
+  end
+
+  local samplefiles = String[]
+  local ftype
+  # local cnames = String[]
+
+  if typeof(model.method) in [Sample, Variational]
+    if isa(model.method, Sample)
+      ftype = diagnostics ? "diagnostics" : "samples"
+    else
+      ftype = lowercase(string(typeof(model.method)))
+    end
 
     for i in 1:model.nchains
-      model.id = i
-      model.data_file ="$(model.name)_$(i).data.R"
-      if init != Nothing
-         model.init_file = "$(model.name)_$(i).init.R"
-      end
-      if isa(model.method, Sample)
-        model.output.file = model.name*"_samples_$(i).csv"
-        isfile("$(model.name)_samples_$(i).csv") && rm("$(model.name)_samples_$(i).csv")
-        if diagnostics
-          model.output.diagnostic_file = model.name*"_diagnostics_$(i).csv"
-          isfile("$(model.name)_diagnostics_$(i).csv") && rm("$(model.name)_diagnostics_$(i).csv")
-        end
-      elseif isa(model.method, Optimize)
-        isfile("$(model.name)_optimize_$(i).csv") && rm("$(model.name)_optimize_$(i).csv")
-        model.output.file = model.name*"_optimize_$(i).csv"
-      elseif isa(model.method, Variational)
-        isfile("$(model.name)_variational_$(i).csv") && rm("$(model.name)_variational_$(i).csv")
-        model.output.file = model.name*"_variational_$(i).csv"
-      elseif isa(model.method, Diagnose)
-        isfile("$(model.name)_diagnose_$(i).csv") && rm("$(model.name)_diagnose_$(i).csv")
-        model.output.file = model.name*"_diagnose_$(i).csv"
-      end
-      model.command[i] = cmdline(model)
+      push!(samplefiles, "$(path_prefix)_$(ftype)_$(i).csv")
     end
 
-    try
-      if file_run_log
-        run(pipeline(par(model.command), stdout="$(model.name)_run.log"))
-      else
-        run(par(model.command))
-      end
-    catch e
-      println("\nAn error occurred while running the previously compiled Stan program.\n")
-      print("Please check the contents of file $(model.name)_run.log and the")
-      println("'command' field in the Stanmodel, e.g. stanmodel.command.\n")
-      error("Return code = -5")
+    if summary
+      stan_summary(model, par(samplefiles), CmdStanDir=CmdStanDir)
     end
 
-    local samplefiles = String[]
-    local ftype
-    # local cnames = String[]
+    (res, cnames) = read_samples(model, diagnostics)
 
-    if typeof(model.method) in [Sample, Variational]
-      if isa(model.method, Sample)
-        ftype = diagnostics ? "diagnostics" : "samples"
-      else
-        ftype = lowercase(string(typeof(model.method)))
-      end
-
-      for i in 1:model.nchains
-        push!(samplefiles, "$(model.name)_$(ftype)_$(i).csv")
-      end
-
-      if summary
-        stan_summary(model, par(samplefiles), CmdStanDir=CmdStanDir)
-      end
-
-      (res, cnames) = read_samples(model, diagnostics)
-
-    elseif isa(model.method, Optimize)
-      res, cnames = read_optimize(model)
+  elseif isa(model.method, Optimize)
+    res, cnames = read_optimize(model)
 
-    elseif isa(model.method, Diagnose)
-      res, cnames = read_diagnose(model)
+  elseif isa(model.method, Diagnose)
+    res, cnames = read_diagnose(model)
 
-    else
-      println("\nAn unknown method is specified in the call to stan().")
-      error("Return code = -10")
-    end
-  end # cd()
+  else
+    println("\nAn unknown method is specified in the call to stan().")
+    error("Return code = -10")
+  end
 
   if model.output_format != :array
     start_sample = 1

diff --git a/src/main/stanmodel.jl b/src/main/stanmodel.jl
@@ -41,6 +41,7 @@ mutable struct Stanmodel
   monitors::Vector{String}
   data_file::String
   command::Vector{Base.AbstractCmd}
+  object_file::String
   method::Method
   random::Random
   init_file::String
@@ -155,6 +156,7 @@ function Stanmodel(
 
   id::Int=0
   data_file::String=""
+  object_file::String=""
   init_file::String=""
   cmdarray = fill(``, nchains)
 
@@ -173,7 +175,7 @@ function Stanmodel(
   Stanmodel(name, nchains, 
     num_warmup, num_samples, thin,
     id, model, model_file, monitors,
-    data_file, cmdarray, method, random,
+    data_file, cmdarray, object_file, method, random,
     init_file, output, printsummary, pdir, tmpdir, output_format);
 end
 

diff --git a/src/utilities/create_cmd_line.jl b/src/utilities/create_cmd_line.jl
@@ -23,7 +23,7 @@ function cmdline(m)
   cmd = ``
   if isa(m, Stanmodel)
     # Handle the model name field for unix and windows
-    cmd = @static Sys.isunix() ? `./$(getfield(m, :name))` : `cmd /c $(getfield(m, :name)).exe`
+    cmd = @static Sys.isunix() ? `$(getfield(m, :object_file))` : `cmd /c $(getfield(m, :object_file)).exe`
 
     # Method (sample, optimize, variational and diagnose) specific portion of the model
     cmd = `$cmd $(cmdline(getfield(m, :method)))`

diff --git a/src/utilities/read_diagnose.jl b/src/utilities/read_diagnose.jl
@@ -27,15 +27,15 @@ function read_diagnose(model::Stanmodel)
   local sstr
 
   for i in 1:model.nchains
-    if isfile("$(model.name)_$(res_type)_$(i).csv")
+    if isfile("$(model.object_file)_$(res_type)_$(i).csv")
 
       ## A result type file for chain i is present ##
 
       if i == 1
 
         # Extract cmdstan version
 
-        str = read("$(model.name)_$(res_type)_$(i).csv", String)
+        str = read("$(model.object_file)_$(res_type)_$(i).csv", String)
         sstr = split(str)
         tdict[:stan_version] = "$(parse(Int, sstr[4])).$(parse(Int, sstr[8])).$(parse(Int, sstr[12]))"
       end

diff --git a/src/utilities/read_optimize.jl b/src/utilities/read_optimize.jl
@@ -28,12 +28,12 @@ function read_optimize(model::Stanmodel)
   tdict = Dict()
 
   for i in 1:model.nchains
-    if isfile("$(model.name)_$(res_type)_$(i).csv")
+    if isfile("$(model.object_file)_$(res_type)_$(i).csv")
 
       # A result type file for chain i is present ##
-      instream = open("$(model.name)_$(res_type)_$(i).csv")
+      instream = open("$(model.object_file)_$(res_type)_$(i).csv")
       if i == 1
-        open("$(model.name)_$(res_type)_$(i).csv") do instream 
+        open("$(model.object_file)_$(res_type)_$(i).csv") do instream 
           str = read(instream, String)
           sstr = split(str)
           tdict[:stan_major_version] = [parse(Int, sstr[4])]
@@ -43,7 +43,7 @@ function read_optimize(model::Stanmodel)
       end
 
       # After reopening the file, skip all comment lines
-      open("$(model.name)_$(res_type)_$(i).csv") do instream
+      open("$(model.object_file)_$(res_type)_$(i).csv") do instream
         skipchars(isspace, instream, linecomment='#')
         line = Unicode.normalize(readline(instream), newline2lf=true)
 

diff --git a/src/utilities/read_samples.jl b/src/utilities/read_samples.jl
@@ -50,11 +50,13 @@ function read_samples(m::Stanmodel, diagnostics=false, warmup_samples=false)
   end
 
   # Read .csv files created by each chain
-
+
+  cnames = String[]
+  a3d = Dict{String,Any}("a" => nothing)
   for i in 1:m.nchains
-    if isfile("$(m.name)_$(ftype)_$(i).csv")
+    if isfile("$(m.object_file)_$(ftype)_$(i).csv")
 
-      instream = open("$(m.name)_$(ftype)_$(i).csv")
+      instream = open("$(m.object_file)_$(ftype)_$(i).csv")
 
       # Skip initial set of commented lines, e.g. containing
       # cmdstan version info, etc.
@@ -78,8 +80,8 @@ function read_samples(m::Stanmodel, diagnostics=false, warmup_samples=false)
       # Preserve cnames and create a3d
 
       if i == 1
-        global cnames = convert.(String, idx[indvec])
-        global a3d = fill(0.0, noofsamples, length(indvec), m.nchains)
+        append!(cnames, convert.(String, idx[indvec]))
+        a3d["a"] = fill(0.0, noofsamples, length(indvec), m.nchains)
       end
 
       # Read in the samples for all chains
@@ -93,12 +95,12 @@ function read_samples(m::Stanmodel, diagnostics=false, warmup_samples=false)
         else
           flds = parse.(Float64, split(strip(line), ","))
           flds = reshape(flds[indvec], 1, length(indvec))
-          a3d[j,:,i] = flds
+          a3d["a"][j,:,i] = flds
         end
       end   # read in samples  
     end   # select next file if it exists
   end   # loop over all chains
 
-  (a3d, cnames)
+  (a3d["a"], cnames)
 
 end   # end of read_samples
diff --git a/src/utilities/read_summary.jl b/src/utilities/read_summary.jl
@@ -19,7 +19,7 @@ read_summary(m::Stanmodel)
 """
 function read_summary(m::Stanmodel)
 
-  fname = "$(m.tmpdir)/$(m.name)_summary.csv"
+  fname = "$(m.object_file)_summary.csv"
   !isfile(fname) && stan_summary(m)
 
   df = CSV.read(fname, DataFrame; delim=",", comment="#")