From 50a9b44dac133e6d3250760fcebc42a92cb49590 Mon Sep 17 00:00:00 2001
From: MLopez-Ibanez <2620021+MLopez-Ibanez@users.noreply.github.com>
Date: Tue, 23 Jul 2024 19:26:08 +0100
Subject: [PATCH] Use data.table in extractElites.

---
 R/irace.R | 35 ++++++++++++++++++++++++++++++++++-
 1 file changed, 34 insertions(+), 1 deletion(-)

diff --git a/R/irace.R b/R/irace.R
index f1e2e62e..45fef379 100644
--- a/R/irace.R
+++ b/R/irace.R
@@ -423,7 +423,40 @@ allConfigurationsInit <- function(scenario)
   allConfigurations
 }
 
-extractElites <- function(elites, nbElites, debugLevel)
+## extractElites
+# Input: the configurations with the .RANK. field filled.
+#        the number of elites wished
+# Output: nbElites elites, sorted by ranks, with the weights assigned.
+extractElites <- function(configurations, nbElites, debugLevel)
+{
+  irace.assert(nbElites > 0L)
+  # Keep only alive configurations.
+  elites <- as.data.table(configurations)
+  before <- nrow(elites)
+  # Remove duplicated. Duplicated configurations may be generated, however, it
+  # is too slow to check at generation time. Nevertheless, we can check now
+  # since we typically have very few elites.
+  elites <- unique(elites, by=which(!startsWith(colnames(elites), ".")))
+  after <- nrow(elites)
+  if (debugLevel >= 2L && after < before)
+    irace.note("Dropped ", before - after, " duplicated elites.\n")
+
+  after <- min(after, nbElites)
+  .RANK. <- NULL
+  setorder(elites, .RANK.)
+  selected <- seq_len(after)
+  elites <- elites[selected, ]
+  set(elites, j = ".WEIGHT.", value = ((after + 1L) - selected) / (after * (after + 1L) / 2))
+  setDF(elites)
+  rownames(elites) <- elites[[".ID."]]
+  print(elites)
+  old <- old_extractElites(configurations, nbElites, debugLevel)
+  print(old)
+  testthat::expect_identical(elites, old)
+  elites
+}
+
+old_extractElites <- function(elites, nbElites, debugLevel)
 {
   irace.assert(nbElites > 0L)
   # Remove duplicated. Duplicated configurations may be generated, however, it