More results

_benchplot/benchplot-dict.R

@@ -220,7 +220,7 @@ groupby.syntax.dict = {list(
     "sum v1:v3 by id6" = "SELECT id6, sum(v1) AS v1, sum(v2) AS v2, sum(v3) AS v3 FROM tbl GROUP BY id6",
     "median v3 sd v3 by id4 id5" = "SELECT id4, id5, quantile_cont(v3, 0.5) AS median_v3, stddev(v3) AS sd_v3 FROM tbl GROUP BY id4, id5",
     "max v1 - min v2 by id3" = "SELECT id3, max(v1)-min(v2) AS range_v1_v2 FROM tbl GROUP BY id3",
-    "largest two v3 by id6" = "SELECT id6, v3 AS largest2_v3 FROM (SELECT id6, v3, row_number() OVER (PARTITION BY id6 ORDER BY v3 DESC) AS order_v3 FROM x WHERE v3 IS NOT NULL) sub_query WHERE order_v3 <= 2",
+    "largest two v3 by id6" = "SELECT id6, unnest(max(v3, 2)) largest2_v3 FROM x WHERE v3 IS NOT NULL GROUP BY id6",
     "regression v1 v2 by id2 id4" = "SELECT id2, id4, pow(corr(v1, v2), 2) AS r2 FROM tbl GROUP BY id2, id4",
     "sum v3 count by id1:id6" = "SELECT id1, id2, id3, id4, id5, id6, sum(v3) AS v3, count(*) AS count FROM tbl GROUP BY id1, id2, id3, id4, id5, id6"
   )},
@@ -232,7 +232,7 @@ groupby.syntax.dict = {list(
     "sum v1:v3 by id6" = "SELECT id6, sum(v1) AS v1, sum(v2) AS v2, sum(v3) AS v3 FROM tbl GROUP BY id6",
     "median v3 sd v3 by id4 id5" = "SELECT id4, id5, quantile_cont(v3, 0.5) AS median_v3, stddev(v3) AS sd_v3 FROM tbl GROUP BY id4, id5",
     "max v1 - min v2 by id3" = "SELECT id3, max(v1)-min(v2) AS range_v1_v2 FROM tbl GROUP BY id3",
-    "largest two v3 by id6" = "SELECT id6, v3 AS largest2_v3 FROM (SELECT id6, v3, row_number() OVER (PARTITION BY id6 ORDER BY v3 DESC) AS order_v3 FROM x WHERE v3 IS NOT NULL) sub_query WHERE order_v3 <= 2",
+    "largest two v3 by id6" = "SELECT id6, unnest(max(v3, 2)) largest2_v3 FROM x WHERE v3 IS NOT NULL GROUP BY id6",
     "regression v1 v2 by id2 id4" = "SELECT id2, id4, pow(corr(v1, v2), 2) AS r2 FROM tbl GROUP BY id2, id4",
     "sum v3 count by id1:id6" = "SELECT id1, id2, id3, id4, id5, id6, sum(v3) AS v3, count(*) AS count FROM tbl GROUP BY id1, id2, id3, id4, id5, id6"
   )},

_control/skipped_benchmarks.csv

@@ -57,5 +57,4 @@ pandas,join,J1_1e9_NA_0_0,c6id.4xlarge
 polars,join,J1_1e9_NA_0_0,c6id.4xlarge
 pydatatable,join,J1_1e9_NA_0_0,c6id.4xlarge
 spark,join,J1_1e9_NA_0_0,c6id.4xlarge
-clickhouse,join,J1_1e9_NA_0_0,c6id.4xlarge
 

_run/run_large.sh

@@ -30,7 +30,7 @@ echo "Running all solutions on large (50GB) datasets"
 ./run.sh
 
 
-###
+##
 echo "done..."
 echo "removing data files"
 rm data/*.csv

clickhouse/VERSION

@@ -1 +1 @@
-25.1.3.23
+25.9.3.48

clickhouse/setup-clickhouse.sh

@@ -15,7 +15,8 @@ sudo service clickhouse-server start ||:
 
 # modify clickhouse settings so data is stored on the mount.
 sudo mkdir -p /var/lib/mount/clickhouse-nvme-mount/
-sudo chown clickhouse:clickhouse /var/lib/mount/clickhouse-nvme-mount
+sudo chown -R clickhouse:clickhouse /var/lib/mount/clickhouse-nvme-mount
+sudo chown -R clickhouse:clickhouse /var/lib/mount/clickhouse-nvme-mount/store
 
 # set up clickhouse tmp space
 sudo mkdir -p /var/lib/mount/clickhouse-tmp/

clickhouse/ver-clickhouse.sh

@@ -8,4 +8,4 @@ ch_installed && clickhouse-client --version-clean > clickhouse/VERSION && echo "
 if [[ $TEST_RUN != "true" ]]; then
 	sudo chown ubuntu:ubuntu clickhouse/VERSION
 	sudo chown ubuntu:ubuntu clickhouse/REVISION
-fi
+fi

collapse/setup-collapse.sh

@@ -3,6 +3,8 @@ set -e
 
 # install stable collapse
 mkdir -p ./collapse/r-collapse
-Rscript -e 'install.packages(c("Rcpp", "collapse"), lib="./collapse/r-collapse", repos = "http://cloud.r-project.org")'
+ncores=`python3 -c 'import multiprocessing as mp; print(mp.cpu_count())'`
+MAKEFLAGS="-j$ncores" Rscript -e 'install.packages(c("Rcpp", "collapse"), lib="./collapse/r-collapse", repos = "http://cloud.r-project.org")'
+
+./collapse/ver-collapse.sh
 
-./collapse/ver-collapse.sh

datafusion/VERSION

@@ -1 +1 @@
-50.0.0
+50.1.0

datatable/VERSION

@@ -1 +1 @@
-1.16.99
+1.17.99

duckdb/VERSION

@@ -1 +1 @@
-1.4.0
+1.4.1

duckdb/groupby-duckdb.R

@@ -36,7 +36,7 @@ if (on_disk) {
 
 table_type = "TEMP"
 if (machine_type == 'c6id.4xlarge' && on_disk) {
-  dbExecute(con, "pragma memory_limit='25G'")
+  dbExecute(con, "pragma memory_limit='20G'")
 }
 
 dbExecute(con, "SET enable_progress_bar = false;")
@@ -106,27 +106,25 @@ print(dbGetQuery(con, "SELECT * FROM ans WHERE ROWID > (SELECT count(*) FROM ans
 invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
 
 
-if (!(machine_type == 'c6id.4xlarge' && on_disk)) {
-  question = "sum v1 mean v3 by id3" # q3
-  t = system.time({
-    dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id3, sum(v1) AS v1, avg(v3) AS v3 FROM x GROUP BY id3", table_type))
-    print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
-  })[["elapsed"]]
-  m = memory_usage()
-  chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(v1) AS v1, sum(v3) AS v3 FROM ans"))[["elapsed"]]
-  write.log(run=1L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
-  invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
-  t = system.time({
-    dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id3, sum(v1) AS v1, avg(v3) AS v3 FROM x GROUP BY id3", table_type))
-    print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
-  })[["elapsed"]]
-  m = memory_usage()
-  chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(v1) AS v1, sum(v3) AS v3 FROM ans"))[["elapsed"]]
-  write.log(run=2L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
-  print(dbGetQuery(con, "SELECT * FROM ans LIMIT 3"))                                      ## head
-  print(dbGetQuery(con, "SELECT * FROM ans WHERE ROWID > (SELECT count(*) FROM ans) - 4")) ## tail
-  invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
-}
+question = "sum v1 mean v3 by id3" # q3
+t = system.time({
+dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id3, sum(v1) AS v1, avg(v3) AS v3 FROM x GROUP BY id3", table_type))
+print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
+})[["elapsed"]]
+m = memory_usage()
+chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(v1) AS v1, sum(v3) AS v3 FROM ans"))[["elapsed"]]
+write.log(run=1L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
+invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
+t = system.time({
+dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id3, sum(v1) AS v1, avg(v3) AS v3 FROM x GROUP BY id3", table_type))
+print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
+})[["elapsed"]]
+m = memory_usage()
+chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(v1) AS v1, sum(v3) AS v3 FROM ans"))[["elapsed"]]
+write.log(run=2L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
+print(dbGetQuery(con, "SELECT * FROM ans LIMIT 3"))                                      ## head
+print(dbGetQuery(con, "SELECT * FROM ans WHERE ROWID > (SELECT count(*) FROM ans) - 4")) ## tail
+invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
 
 question = "mean v1:v3 by id4" # q4
 t = system.time({
@@ -209,27 +207,25 @@ print(dbGetQuery(con, "SELECT * FROM ans WHERE ROWID > (SELECT count(*) FROM ans
 invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
 
 
-if (!(machine_type == 'c6id.4xlarge' && on_disk)) {
-  question = "largest two v3 by id6" # q8
-  t = system.time({
-    dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id6, unnest(max(v3, 2)) largest2_v3 FROM x WHERE v3 IS NOT NULL GROUP BY id6", table_type))
-    print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
-  })[["elapsed"]]
-  m = memory_usage()
-  chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(largest2_v3) AS largest2_v3 FROM ans"))[["elapsed"]]
-  write.log(run=1L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
-  invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
-  t = system.time({
-    dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id6, unnest(max(v3, 2)) largest2_v3 FROM x WHERE v3 IS NOT NULL GROUP BY id6", table_type))
-    print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
-  })[["elapsed"]]
-  m = memory_usage()
-  chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(largest2_v3) AS largest2_v3 FROM ans"))[["elapsed"]]
-  write.log(run=2L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
-  print(dbGetQuery(con, "SELECT * FROM ans LIMIT 3"))                                      ## head
-  print(dbGetQuery(con, "SELECT * FROM ans WHERE ROWID > (SELECT count(*) FROM ans) - 4")) ## tail
-  invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
-}
+question = "largest two v3 by id6" # q8
+t = system.time({
+dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id6, unnest(max(v3, 2)) largest2_v3 FROM x WHERE v3 IS NOT NULL GROUP BY id6", table_type))
+print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
+})[["elapsed"]]
+m = memory_usage()
+chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(largest2_v3) AS largest2_v3 FROM ans"))[["elapsed"]]
+write.log(run=1L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
+invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
+t = system.time({
+dbExecute(con, sprintf("CREATE %s TABLE ans AS SELECT id6, unnest(max(v3, 2)) largest2_v3 FROM x WHERE v3 IS NOT NULL GROUP BY id6", table_type))
+print(c(nr<-dbGetQuery(con, "SELECT count(*) AS cnt FROM ans")$cnt, nc<-ncol(dbGetQuery(con, "SELECT * FROM ans LIMIT 0"))))
+})[["elapsed"]]
+m = memory_usage()
+chkt = system.time(chk<-dbGetQuery(con, "SELECT sum(largest2_v3) AS largest2_v3 FROM ans"))[["elapsed"]]
+write.log(run=2L, task=task, data=data_name, in_rows=in_nr, question=question, out_rows=nr, out_cols=nc, solution=solution, version=ver, git=git, fun=fun, time_sec=t, mem_gb=m, cache=cache, chk=make_chk(chk), chk_time_sec=chkt, on_disk=on_disk, machine_type=machine_type)
+print(dbGetQuery(con, "SELECT * FROM ans LIMIT 3"))                                      ## head
+print(dbGetQuery(con, "SELECT * FROM ans WHERE ROWID > (SELECT count(*) FROM ans) - 4")) ## tail
+invisible(dbExecute(con, "DROP TABLE IF EXISTS ans"))
 
 question = "regression v1 v2 by id2 id4" # q9
 t = system.time({