Support distmat as lower triangular for PAM

.github/workflows/check-sanitizers.yaml

@@ -52,8 +52,8 @@ jobs:
         run: |
           install.packages("remotes")
           library("remotes")
-          install_cran(c("Matrix", "bigmemory"))
-          install_cran(c("codetools", "cluster", "lattice", "class"), force = TRUE)
+          install_cran(c("bigmemory"))
+          install_cran(c("Matrix", "codetools", "cluster", "lattice", "class"), force = TRUE)
           update(dev_package_deps(dependencies = TRUE))
 
       - name: Check with sanitizers
@@ -62,7 +62,7 @@ jobs:
           ASAN_OPTIONS: "detect_leaks=0:detect_odr_violation=0"
         run: |
           Rdevel CMD build --no-build-vignettes --no-manual .
-          Rdevel --vanilla CMD check *.tar.gz --as-cran --no-manual
+          Rdevel --vanilla CMD check *.tar.gz --as-cran --no-manual --ignore-vignettes
         continue-on-error: true
 
       - name: Show testthat output

CHANGELOG.md

@@ -1,8 +1,9 @@
 # Changelog
 
-## Version 5.6.0
+## Version 6.0.0
 * Update Makevars for ARM version of Windows.
 * Sanitize internal usage of `do.call` to avoid huge backtraces.
+* Support lower triangular `distmat` objects for symmetric distances (#77) - breaking change.
 
 ## Version 5.5.12
 * Remove explicit C++ requirements.

DESCRIPTION

@@ -12,7 +12,7 @@ Description: Time series clustering along with optimized techniques related
     All included distance functions have custom loops optimized for the 
     calculation of cross-distance matrices, including parallelization support.
     Several cluster validity indices are included.
-Version: 5.5.12
+Version: 5.5.12.9000
 Depends:
     R (>= 3.3.0),
     methods,
@@ -47,7 +47,7 @@ Suggests:
     knitr,
     rmarkdown,
     testthat
-Date: 2024-06-22
+Date: 2024-07-01
 Author: Alexis Sarda-Espinosa
 Maintainer: Alexis Sarda <[email protected]>
 BugReports: https://github.com/asardaes/dtwclust/issues
@@ -83,7 +83,9 @@ Collate:
     'DISTANCES-sbd.R'
     'DISTANCES-sdtw.R'
     'GENERICS-cvi.R'
+    'RD-helpers.R'
     'S4-Distmat.R'
+    'S4-DistmatLowerTriangular.R'
     'S4-PairTracker.R'
     'S4-SparseDistmat.R'
     'S4-tsclustFamily.R'

NAMESPACE

@@ -6,6 +6,7 @@ S3method(as.data.frame,pairdist)
 S3method(as.matrix,crossdist)
 S3method(as.matrix,pairdist)
 S3method(base::dim,Distmat)
+S3method(base::dim,DistmatLowerTriangular)
 S3method(base::dim,SparseDistmat)
 S3method(cl_class_ids,TSClusters)
 S3method(cl_membership,TSClusters)
@@ -105,6 +106,7 @@ importFrom(ggplot2,theme_bw)
 importFrom(ggrepel,geom_label_repel)
 importFrom(graphics,plot)
 importFrom(methods,S3Part)
+importFrom(methods,as)
 importFrom(methods,callNextMethod)
 importFrom(methods,initialize)
 importFrom(methods,is)

R/CENTROIDS-dba.R

@@ -21,7 +21,7 @@
 #' @param max.iter Maximum number of iterations allowed.
 #' @param delta At iteration `i`, if `all(abs(centroid_{i}` `-` `centroid_{i-1})` `< delta)`,
 #'   convergence is assumed.
-#' @template error-check
+#' @param error.check `r roxygen_error_check_param()`
 #' @param trace If `TRUE`, the current iteration is printed to output.
 #' @param mv.ver Multivariate version to use. See below.
 #'
@@ -34,15 +34,13 @@
 #' the same result provided the elements of `X` keep the same values, although their order may
 #' change.
 #'
-#' @template rcpp-parallel
+#' `r roxygen_window_details()`
 #'
-#' @section Parallel Computing:
+#' @section `r roxygen_rcpp_parallel_section()`
 #'
 #'   This function appears to be very sensitive to numerical inaccuracies if multi-threading is used
 #'   in a **32 bit** installation. In such systems, consider limiting calculations to 1 thread.
 #'
-#' @template window
-#'
 #' @return The average time series.
 #'
 #' @section Multivariate series:

R/CENTROIDS-pam.R

@@ -3,6 +3,7 @@
 #' Extract the medoid time series based on a distance measure.
 #'
 #' @export
+#' @importFrom methods as
 #' @importFrom rlang exprs
 #' @importFrom Matrix rowSums
 #'
@@ -12,7 +13,7 @@
 #' @param ids Integer vector indicating which of the `series` should be considered.
 #' @param distmat Optionally, a pre-computed cross-distance matrix of *all* `series`.
 #' @param ... Any extra parameters for the `distance` function that may be used.
-#' @template error-check
+#' @param error.check `r roxygen_error_check_param()`
 #'
 #' @details
 #'
@@ -43,24 +44,21 @@ pam_cent <- function(series, distance, ids = seq_along(series), distmat = NULL,
         if (missing(distance))
             distance <- attr(distmat, "method")
 
-        args <- rlang::exprs(
-            distmat = distmat,
-            series = series,
-            dist_args = dots,
-            distance = distance,
-            control = partitional_control(),
-            error.check = error.check
-        )
-
         if (is.null(distmat)) {
             if (is.null(distance))
                 stop("If 'distmat' is missing, 'distance' must be provided.")
 
-            args$distmat <- NULL
+            distmat <- Distmat$new(
+                series = series,
+                dist_args = dots,
+                distance = distance,
+                control = partitional_control(),
+                error.check = error.check
+            )
+        }
+        else {
+            distmat <- methods::as(distmat, "Distmat")
         }
-
-        # S4-Distmat.R
-        distmat <- do.call(Distmat$new, args)
     }
 
     d <- distmat[ids, ids, drop = FALSE]

R/CENTROIDS-sdtw-cent.R

@@ -41,7 +41,7 @@ sdtw_cent_stats_gr <- function(ignored, .shared_env_, ...) {
 #'   for `optim`, and `...` for both).
 #' @param opts List of options to pass to `nloptr` or [stats::optim()]'s `control`. The defaults in
 #'   the function's formals are for `nloptr`, but the value will be adjusted for `optim` if needed.
-#' @template error-check
+#' @param error.check `r roxygen_error_check_param()`
 #'
 #' @details
 #'
@@ -52,9 +52,7 @@ sdtw_cent_stats_gr <- function(ignored, .shared_env_, ...) {
 #' @return The resulting centroid, with the optimization results as attributes (except for the
 #'   returned centroid).
 #'
-#' @template rcpp-parallel
-#'
-#' @section Parallel Computing:
+#' @section `r roxygen_rcpp_parallel_section()`
 #'
 #'   For unknown reasons, this function has returned different results (in the order of 1e-6) when
 #'   using multi-threading in x64 Windows installations in comparison to other environments (using

R/CENTROIDS-shape-extraction.R

@@ -14,7 +14,7 @@
 #'   the matrices in `X`. **It will be z-normalized**.
 #' @param znorm Logical flag. Should z-scores be calculated for `X` before processing?
 #' @param ... Further arguments for [zscore()].
-#' @template error-check
+#' @param error.check `r roxygen_error_check_param()`
 #'
 #' @details
 #'

R/CLUSTERING-ddist2.R

@@ -179,16 +179,19 @@ ddist2 <- function(distance, control) {
 
         if (!dist_entry$loop) {
             # CUSTOM LOOP, LET THEM HANDLE OPTIMIZATIONS
-            dm <- base::as.matrix(quoted_call(
+            dm <- quoted_call(
                 proxy::dist, x = x, y = centroids, method = distance, dots = dots
-            ))
+            )
 
             if (isTRUE(dots$pairwise)) {
                 dim(dm) <- NULL
                 return(ret(dm, class = "pairdist"))
             }
+            else if (inherits(dm, "dist")) {
+                return(ret(dm))
+            }
             else {
-                return(ret(dm, class = "crossdist"))
+                return(ret(base::as.matrix(dm), class = "crossdist"))
             }
         }
 
@@ -237,11 +240,16 @@ ddist2 <- function(distance, control) {
             }
             else if (!multiple_workers) {
                 # WHOLE SYMMETRIC DISTMAT WITHOUT CUSTOM LOOP OR USING SEQUENTIAL proxy LOOP
-                dm <- base::as.matrix(quoted_call(
+                dm <- quoted_call(
                     proxy::dist, x = x, y = NULL, method = distance, dots = dots
-                ))
+                )
 
-                return(ret(dm, class = "crossdist"))
+                if (inherits(dm, "dist")) {
+                    return(ret(dm))
+                }
+                else {
+                    return(ret(base::as.matrix(dm), class = "crossdist"))
+                }
             }
         }
 

R/CLUSTERING-tadpole.R

@@ -10,7 +10,7 @@
 #' @param window.size Window size constraint for DTW (Sakoe-Chiba). See details.
 #' @param k The number of desired clusters. Can be a vector with several values.
 #' @param dc The cutoff distance(s). Can be a vector with several values.
-#' @template error-check
+#' @param error.check `r roxygen_error_check_param()`
 #' @param lb Which lower bound to use, "lbk" for [lb_keogh()] or "lbi" for [lb_improved()].
 #' @param trace Logical flag. If `TRUE`, more output regarding the progress is printed to screen.
 #'
@@ -39,7 +39,7 @@
 #' - For multiple `dc` values, multi-processing with [foreach::foreach()].
 #' - The internal distance calculations use multi-threading with [RcppParallel::RcppParallel].
 #'
-#' @template window
+#' `r roxygen_window_details()`
 #'
 #' @return
 #'
@@ -52,7 +52,7 @@
 #' For multiple `k`/`dc` values, a list of lists is returned, each internal list having the
 #' aforementioned elements.
 #'
-#' @template rcpp-parallel
+#' @section `r roxygen_rcpp_parallel_section()`
 #'
 #' @references
 #'