update version, minor change in plotting progress bar & typos

R/DESCRIPTION

@@ -1,7 +1,7 @@
 Package: Robyn
 Type: Package
 Title: Automated Marketing Mix Modeling (MMM) Open Source Beta Project from Facebook Marketing Science 
-Version: 3.4.7
+Version: 3.4.8
 Authors@R: c(
     person("Gufeng", "Zhou", , "gufeng@fb.com", c("aut")),
     person("Leonel", "Sentana", , "leonelsentana@fb.com", c("aut")),

R/R/imports.R

@@ -54,7 +54,7 @@ dt_vars <- c(
   "optmResponseUnitTotalLift", "optmSpendUnit", "optmSpendUnitTotalDelta", "param",
   "perc", "percentage", "pos", "predicted", "refreshStatus", "response", "rn", "robynPareto",
   "roi", "roi_mean", "roi_total", "rsq_lm", "rsq_nls", "rsq_train", "s0", "scale_shape_halflife",
-  "season", "shape", "solID", "spend", "spend_share", "spend_share_refresh",
+  "season", "shape", "solID", "spend", "spend_share", "spend_share_refresh", "sid",
   "theta", "theta_halflife", "total_spend", "trend", "trial", "type", "value", "variable",
   "weekday", "x", "xDecompAgg", "xDecompMeanNon0", "xDecompMeanNon0Perc",
   "xDecompMeanNon0PercRF", "xDecompMeanNon0RF", "xDecompPerc", "xDecompPercRF", "y", "yhat",

R/R/model.R

@@ -155,7 +155,7 @@ robyn_run <- function(InputCollect,
         "hyperparameters & 10-fold ridge x-validation on 1 core (Windows fallback)"
       ))
     }
-    
+
     # ng_collect <- list()
     model_output_collect <- list()
 
@@ -468,15 +468,15 @@ robyn_run <- function(InputCollect,
   # ggplot doesn't work with process forking on MacOS
   # however it works fine on Linux and Windows
   parallel_plotting <- Sys.info()["sysname"] != "Darwin"
-  
+
   if (plot_pareto) {
     if (parallel_plotting) {
       message(paste(">>> Plotting", num_pareto123, "Pareto optimum models on", InputCollect$cores, "cores..."))
     } else {
       message(paste(">>> Plotting", num_pareto123, "Pareto optimum models on 1 core (MacOS fallback)..."))
     }
   }
-  
+
   if (parallel_plotting) {
     registerDoParallel(InputCollect$cores)
   } else {
@@ -857,6 +857,11 @@ robyn_run <- function(InputCollect,
         dt_transformSaturationSpendReverse[, (InputCollect$organic_vars) := NA]
       }
 
+      if (!parallel_plotting) {
+        cnt <- cnt + 1
+        setTxtProgressBar(pbplot, cnt)
+      }
+
       return(list(
         mediaVecCollect = rbind(
           dt_transformPlot[, ":="(type = "rawMedia", solID = sid)],
@@ -872,8 +877,10 @@ robyn_run <- function(InputCollect,
       ))
     } # end solution loop
 
-    cnt <- cnt + length(uniqueSol)
-    setTxtProgressBar(pbplot, cnt)
+    if (parallel_plotting) {
+      cnt <- cnt + length(uniqueSol)
+      setTxtProgressBar(pbplot, cnt)
+    }
 
     # append parallel run results
     mediaVecCollect <- append(mediaVecCollect, lapply(parallelResult, function (x) x$mediaVecCollect))
@@ -1118,7 +1125,7 @@ robyn_mmm <- function(hyper_collect,
 
   # enable parallelisation of main modelling loop for MacOS and Linux only
   parallel_processing <- .Platform$OS.type == "unix"
-  
+
   # create cluster before big for-loop to minimize overhead for parallel backend registering
   if (parallel_processing) {
     registerDoParallel(InputCollect$cores)

demo/demo.R

@@ -4,7 +4,7 @@
 # LICENSE file in the root directory of this source tree.
 
 #############################################################################################
-####################         Facebook MMM Open Source - Robyn 3.4.7    ######################
+####################         Facebook MMM Open Source - Robyn 3.4.8    ######################
 ####################                    Quick guide                   #######################
 #############################################################################################
 
@@ -129,7 +129,7 @@ InputCollect <- robyn_inputs(
 
   ,nevergrad_algo = "TwoPointsDE" # recommended algorithm for Nevergrad, the gradient-free
   # optimisation library https://facebookresearch.github.io/nevergrad/index.html
-  ,trials = 5 # number of allowed iterations per trial. 5 is recommended without calibration,
+  ,trials = 5 # number of allowed trials. 5 is recommended without calibration,
   # 10 with calibration.
 
   # Time estimation: with geometric adstock, 2000 iterations * 5 trials