forder3 (#3162)

R/data.table.R

@@ -1464,11 +1464,13 @@ chmatch2 <- function(x, table, nomatch=NA_integer_) {
       if (!bysameorder && identical(byindex,FALSE)) {
         if (verbose) {last.started.at=proc.time();cat("Finding groups using forderv ... ");flush.console()}
         o__ = forderv(byval, sort=!missing(keyby), retGrp=TRUE)
-        # The sort= argument is called sortStr at C level. It's just about saving the sort of unique strings at
-        # C level for efficiency (cgroup vs csort) when by= not keyby=. All other types are always sorted. Getting
-        # orginal order below is the part that retains original order. Passing sort=TRUE here always won't change any
-        # result at all (tested and confirmed to double check), it'll just make by= slower when there's a large
-        # number of unique strings. It must be TRUE when keyby= though, since the key is just marked afterwards.
+        # The sort= argument is called sortGroups at C level. It's primarily for saving the sort of unique strings at
+        # C level for efficiency when by= not keyby=. Other types also retain appearance order, but at byte level to
+        # minimize data movement and benefit from skipping subgroups which happen to be grouped but not sorted. This byte
+        # appearance order is not the same as the order of group values within by= columns, so the 2nd forder below is
+        # still needed to get the group appearance order. Always passing sort=TRUE above won't change any result at all
+        # (tested and confirmed), it'll just make by= slower. It must be TRUE when keyby= though since the key is just
+        # marked afterwards.
         # forderv() returns empty integer() if already ordered to save allocating 1:xnrow
         bysameorder = orderedirows && !length(o__)
         if (verbose) {

R/timetaken.R

@@ -1,17 +1,15 @@
 timetaken <- function(started.at)
 {
-   if (!inherits(started.at,"proc_time")) stop("Use started.at=proc.time() (faster) not Sys.time() (POSIXt and slow)")
-   secs = proc.time()[3L] - started.at[3L]
-   mins = as.integer(secs) %/% 60L
-   hrs = mins %/% 60L
-   days = hrs %/% 24L
-   mins = mins - hrs * 60L
-   hrs = hrs - days * 24L
-   if (secs > 60.0) {
-     res = if (days>=1L) paste0(days," day", if (days>1L) "s " else " ") else ""
-     paste0(res,sprintf("%02d:%02d:%02d", hrs, mins, as.integer(secs) %% 60L))
-   } else {
-     sprintf(if (secs >= 10.0) "%.1fsec" else "%.3fsec", secs)
-   }
+  if (!inherits(started.at,"proc_time")) stop("Use started.at=proc.time() not Sys.time() (POSIXt and slow)")  # nocov
+  format = function(secs) {
+    if (secs > 60.0) {
+      secs = as.integer(secs)
+      sprintf("%02d:%02d:%02d", secs%/%3600L, (secs%/%60L)%%60L, secs%%60L)
+    } else {
+      sprintf(if (secs >= 10.0) "%.1fs" else "%.3fs", secs)
+    }
+  }
+  tt = proc.time()-started.at  # diff all 3 times
+  paste0(format(tt[3L])," elapsed (", format(tt[1L]), " cpu)")
 }
 

cc.R

@@ -37,7 +37,7 @@ sourceDir <- function(path=getwd(), trace = TRUE, ...) {
   if(trace) cat("\n")
 }
 
-cc = function(test=TRUE, clean=FALSE, debug=FALSE, cc_dir=Sys.getenv("CC_DIR"), CC="gcc") {
+cc = function(test=TRUE, clean=FALSE, debug=FALSE, omp=!debug, cc_dir=Sys.getenv("CC_DIR"), CC="gcc") {
   stopifnot(is.character(CC), length(CC)==1L, !is.na(CC), nzchar(CC))
   gc()
 
@@ -59,10 +59,11 @@ cc = function(test=TRUE, clean=FALSE, debug=FALSE, cc_dir=Sys.getenv("CC_DIR"),
   setwd(file.path(cc_dir,"src"))
   cat(getwd(),"\n")
   if (clean) system("rm *.o *.so")
+  OMP = if (omp) "" else "no-"
   if (debug) {
-    ret = system(sprintf("MAKEFLAGS='-j CC=%s PKG_CFLAGS=-fno-openmp CFLAGS=-std=c99\\ -O0\\ -ggdb\\ -pedantic' R CMD SHLIB -d -o data.table.so *.c", CC))
+    ret = system(sprintf("MAKEFLAGS='-j CC=%s PKG_CFLAGS=-f%sopenmp CFLAGS=-std=c99\\ -O0\\ -ggdb\\ -pedantic' R CMD SHLIB -d -o data.table.so *.c", CC, OMP))
   } else {
-    ret = system(sprintf("MAKEFLAGS='-j CC=%s CFLAGS=-fopenmp\\ -std=c99\\ -O3\\ -pipe\\ -Wall\\ -pedantic' R CMD SHLIB -o data.table.so *.c", CC))
+    ret = system(sprintf("MAKEFLAGS='-j CC=%s CFLAGS=-f%sopenmp\\ -std=c99\\ -O3\\ -pipe\\ -Wall\\ -pedantic' R CMD SHLIB -o data.table.so *.c", CC, OMP))
     # TODO add -Wextra too?
   }
   if (ret) return()
@@ -86,5 +87,5 @@ cc = function(test=TRUE, clean=FALSE, debug=FALSE, cc_dir=Sys.getenv("CC_DIR"),
   invisible()
 }
 
-dd = function()cc(FALSE,debug=TRUE,clean=TRUE)
+dd = function(omp=FALSE)cc(FALSE,debug=TRUE,omp=omp,clean=TRUE)