remove remaining un-needed dependencies (#93)

DESCRIPTION

@@ -3,7 +3,7 @@ Title: Multivariate (Dynamic) Generalized Additive Models
 Version: 1.1.4
 Date: 2024-09-05
 Authors@R: person("Nicholas J", "Clark", , "[email protected]", role = c("aut", "cre"), comment = c(ORCID = "0000-0001-7131-3301"))
-Description: Fit Bayesian Dynamic Generalized Additive Models to sets of time series. Users can build dynamic nonlinear State-Space models that can incorporate semiparametric effects in observation and process components, using a wide range of observation families. Estimation is performed using Markov Chain Monte Carlo with Hamiltonian Monte Carlo in the software 'Stan'. References: Clark & Wells (2022) <doi:10.1111/2041-210X.13974>.
+Description: Fit Bayesian Dynamic Generalized Additive Models to multivariate observations. Users can build nonlinear State-Space models that can incorporate semiparametric effects in observation and process components, using a wide range of observation families. Estimation is performed using Markov Chain Monte Carlo with Hamiltonian Monte Carlo in the software 'Stan'. References: Clark & Wells (2022) <doi:10.1111/2041-210X.13974>.
 URL: https://github.com/nicholasjclark/mvgam, https://nicholasjclark.github.io/mvgam/
 BugReports: https://github.com/nicholasjclark/mvgam/issues
 License: MIT + file LICENSE
@@ -23,10 +23,8 @@ Imports:
     bayesplot (>= 1.5.0),
     ggplot2 (>= 2.0.0),
     parallel,
-    pbapply,
     mvnfast,
     purrr,
-    zoo,
     dplyr,
     magrittr,
     rlang

NEWS.md

@@ -8,6 +8,7 @@
 * Added a `fevd.mvgam()` method to compute forecast error variance decompositions from models fit with Vector Autoregressive dynamics (#21 and #76)
 
 ## Bug fixes
+* Not necessarily a "bug fix", but this update removes several dependencies to lighten installation and improve efficiency of the workflow (#93)
 * Fixed a minor bug in the way `trend_map` recognises levels of the `series` factor
 * Bug fix to ensure `lfo_cv` recognises the actual times in `time`, just in case the user supplies data that doesn't start at `t = 1`. Also updated documentation to better reflect this
 * Bug fix to ensure `update.mvgam` captures any `knots` or `trend_knots` arguments that were passed to the original model call

R/evaluate_mvgams.R

@@ -340,8 +340,8 @@ roll_eval_mvgam = function(object,
   validate_pos_integer(fc_horizon)
 
   # Generate time variable from training data
-  if(class(object$obs_data)[1] == 'list'){
-    all_timepoints <- (data.frame(time = object$obs_data$time)  %>%
+  if(inherits(object$obs_data, 'list')){
+    all_timepoints <- (data.frame(time = object$obs_data$index..time..index) %>%
                          dplyr::select(time) %>%
                          dplyr::distinct() %>%
                          dplyr::arrange(time) %>%
@@ -350,9 +350,9 @@ roll_eval_mvgam = function(object,
 
   } else {
     all_timepoints <- (object$obs_data %>%
-                         dplyr::select(time) %>%
+                         dplyr::select(index..time..index) %>%
                          dplyr::distinct() %>%
-                         dplyr::arrange(time) %>%
+                         dplyr::arrange(index..time..index) %>%
                          dplyr::mutate(time = dplyr::row_number())) %>%
       dplyr::pull(time)
   }
@@ -393,14 +393,17 @@ roll_eval_mvgam = function(object,
   clusterEvalQ(cl, library(mgcv))
   clusterEvalQ(cl, library(rstan))
   clusterEvalQ(cl, library(dplyr))
-  clusterExport(cl = cl,
-                          unclass(lsf.str(envir = asNamespace("mvgam"),
-                                          all = T)),
-                          envir = as.environment(asNamespace("mvgam"))
-  )
-
-  pbapply::pboptions(type = "none")
-  evals <- pbapply::pblapply(evaluation_seq, function(timepoint){
+
+  # Grab internal functions to export to each worker
+  funs_list <- c('eval_mvgam')
+  attr(funs_list, 'envir') <- as.environment(asNamespace("mvgam"))
+  attr(funs_list, 'mode') <- 'function'
+
+  parallel::clusterExport(cl = cl,
+                          funs_list,
+                          envir = as.environment(asNamespace("mvgam")))
+
+  evals <- parallel::parLapply(cl = cl, evaluation_seq, function(timepoint){
     eval_mvgam(object = object,
                n_samples = n_samples,
                n_cores = 1,
@@ -409,8 +412,7 @@ roll_eval_mvgam = function(object,
                score = score,
                log = log,
                weights = weights)
-  },
-  cl = cl)
+  })
   stopCluster(cl)
 
   # Take sum of score at each evaluation point for multivariate models

R/forecast.mvgam.R

@@ -858,14 +858,21 @@ forecast_draws = function(object,
                                     'obs_uncertainty',
                                     'time_dis'),
                             envir = environment())
+
+    # Grab internal functions to export to each worker
+    funs_list <- c('extract_general_trend_pars',
+                   'linkfun',
+                   'forecast_trend',
+                   'extract_series_trend_pars',
+                   'mvgam_predict')
+    attr(funs_list, 'envir') <- as.environment(asNamespace("mvgam"))
+    attr(funs_list, 'mode') <- 'function'
+
     parallel::clusterExport(cl = cl,
-                            unclass(lsf.str(envir = asNamespace("mvgam"),
-                                            all = T)),
+                            funs_list,
                             envir = as.environment(asNamespace("mvgam")))
 
-    pbapply::pboptions(type = "none")
-
-    fc_preds <- pbapply::pblapply(seq_len(dim(betas)[1]), function(i){
+    fc_preds <- parallel::parLapply(cl = cl, seq_len(dim(betas)[1]), function(i){
       # Sample index
       samp_index <- i
 
@@ -1023,7 +1030,7 @@ forecast_draws = function(object,
         })
       }
       out
-    }, cl = cl)
+    })
     stopCluster(cl)
   }