diff --git a/trainbenchmark/trainbenchmark-reporting/individual.R b/trainbenchmark/trainbenchmark-reporting/individual.R
index 075533c2bec65d6f536e226b1a213b95d2fab6e9..9d9a1f67a5f3c3949cd8c37a4cde2b466006cdf2 100644
--- a/trainbenchmark/trainbenchmark-reporting/individual.R
+++ b/trainbenchmark/trainbenchmark-reporting/individual.R
@@ -29,153 +29,156 @@ sizes[["Repair"]] = c("8k", "15k", "33k", "66k", "135k", "271k", "566k", "1.1M",
toolList = read.csv("tool-list.csv")
-# load the data
-tsvs = list.files("../results/merged/individual", pattern = "times-.*\\.csv", full.names = T, recursive = T)
-
-l = lapply(tsvs, read.csv)
-times = rbindlist(l)
-
-# preprocess the data
-times$Tool = factor(times$Tool, levels = toolList$Tool)
-keep_descriptions_first_char(times)
-
-times$Model = gsub("\\D+", "", times$Model)
-times$Model = as.numeric(times$Model)
-times$Time = times$Time / 10^6
-# make the phases a factor with a fixed set of values to help dcasting
-# (e.g. Batch measurements do not have Transformation and Recheck attributes,
-# hence accessing the "Transformation" attribute would throw an error)
-times$Phase = factor(times$Phase, levels = c("Read", "Check", "Transformation", "Recheck"))
-
-times.wide = dcast(data = times,
- formula = Tool + Workload + Description + Model + Run ~ Phase,
- value.var = "Time",
- drop = T,
- fun.aggregate = mean
-)
-
-# calculate aggregated values
-times.derived = times.wide
-times.derived$Read.and.Check = times.derived$Read + times.derived$Check
-times.derived$Transformation.and.Recheck = times.derived$Transformation + times.derived$Recheck
-
-# calculate the median value of runs
-times.aggregated.runs = ddply(
- .data = times.derived,
- .variables = c("Tool", "Workload", "Description", "Model"),
- .fun = colwise(median),
- .progress = "text"
-)
-# drop the "Run" column
-times.aggregated.runs = subset(times.aggregated.runs, select = -c(Run))
-
-times.processed = melt(
- data = times.aggregated.runs,
- id.vars = c("Tool", "Workload", "Description", "Model"),
- measure.vars = phases,
- variable.name = "Phase",
- value.name = "Time"
-)
-
-# beautify plotted record:
-# 1. change dots to spaces
-# 2. make sure that the phases are still factors
-times.plot = times.processed
-times.plot$Phase = gsub('\\.', ' ', times.plot$Phase)
-times.plot$Phase = factor(times.plot$Phase, levels = phasesPrettified)
-times.plot$Workload = factor(times.plot$Workload, levels = workloads)
-
-### line charts
-for (phase in phasesPrettified) {
- phase.filename = gsub(' ', '-', phase)
- workloadSizes = sizes[["Repair"]]
-
- # filter the dataframe to the current phase
- df = times.plot[times.plot$Phase == phase, ]
-
- # do not visualize empty data sets
- if (nrow(df) == 0) {
- print(paste("No rows to visualize for phase", phase))
- next
- }
-
- # x axis labels
- xbreaks = unique(df$Model)
- currentWorkloadSizes = head(workloadSizes, n=length(xbreaks))
- xlabels = paste(xbreaks, "\n", currentWorkloadSizes, sep = "")
-
- # drop every other models size
- maxLabel = max(log2(max(df$Model)), 2)
- if (maxLabel %% 2) {
- start = 3
- } else {
- start = 2
- }
- filter = seq(start, maxLabel, by=2)
-
- xlabels[filter] = ""
-
- # y axis labels
- yaxis = nice_y_axis()
- ybreaks = yaxis$ybreaks
- yminor_breaks = yaxis$yminor_breaks
- ylabels = yaxis$ylabels
-
- for (cpage in 1:6) {
- p = ggplot(df) + #na.omit(df)) +
- aes(x = as.factor(Model), y = Time) +
- labs(title = paste("Individual query execution time,", phase, "phase"), x = "Model size\n#Elements", y = "Execution times [ms]") +
- geom_point(aes(col = Tool, shape = Tool), size = 2.0) +
- # scale_shape_manual(values = seq(0, 15)) +
- # scale_shape_manual(#values = c(0, 15, 1, 16, 2, 17, 4, 7),
- # # labels=c(
- # # "Name Lookup ", "Name Lookup (Incremental)", # http://www.sthda.com/english/wiki/r-plot-pch-symbols-the-different-point-shapes-available-in-r - JastAdd NameLookup
- # # "Intrinsic References", "Intrinsic References (Incremental)", # Tud-cyan - JastAdd Optimized
- # # "Grammar Extension", "Grammar Extension (Incremental)", # HKS65 (green) - JastAdd Specialized
- # # "TinkerGraph", # HKS92 (grey) - TinkerGraph
- # # "Viatra (Incremental)", # HKS33 (violet) - VIATRA
- # # "(none)"
- # # )
- # ) +
- # scale_colour_manual(
- # values = c(
- # "#EE7F00", "#EE7F00", # HKS07 (orange) - JastAdd NameLookup
- # "#009EE0", "#009EE0", # Tud-cyan - JastAdd Optimized
- # "#6AB023", "#6AB023", # HKS65 (green) - JastAdd Specialized
- # "#727879", # HKS92 (grey) - TinkerGraph
- # "#93107E", # HKS33 (violet) - VIATRA
- # "#F0E442", # Yellow
- # "#CC79A7"),
- # labels=c(
- # "Name Lookup ", "Name Lookup (Incremental)", # HKS07 (orange) - JastAdd NameLookup
- # "Intrinsic References", "Intrinsic References (Incremental)", # Tud-cyan - JastAdd Optimized
- # "Grammar Extension", "Grammar Extension (Incremental)", # HKS65 (green) - JastAdd Specialized
- # "TinkerGraph", # HKS92 (grey) - TinkerGraph
- # "Viatra (Incremental)", # HKS33 (violet) - VIATRA
- # "(none)"
- # )) +
- geom_line(aes(col = Tool, group = Tool), size = 0.5) +
- scale_x_discrete(breaks = xbreaks, labels = xlabels) +
- scale_y_log10(breaks = ybreaks, minor_breaks=yminor_breaks, labels = ylabels, limits = c(2e-3,8e4), expand = c(0, 0)) +
- facet_grid_paginate(~ Workload, nrow=1, ncol = 1, page=cpage, scale = "free") +
- guides(color = guide_legend(ncol = 5)) +
- theme_bw() +
- theme(
- plot.title = element_text(hjust = 0.5),
- # text = element_text(family="Open Sans", size = 10),
- # legend.key = element_blank(),
- legend.title = element_blank(),
- legend.position = "bottom",
- axis.text = element_text(size = 16)
- #, panel.grid.minor = element_blank()
+for (scenario in c("inject", "repair")) {
+
+ # load the data
+ tsvs = list.files("../results/merged/individual", pattern = paste("times-.*", scenario, ".*\\.csv", sep=""), full.names = T, recursive = T)
+
+ l = lapply(tsvs, read.csv)
+ times = rbindlist(l)
+
+ # preprocess the data
+ times$Tool = factor(times$Tool, levels = toolList$Tool)
+ keep_descriptions_first_char(times)
+
+ times$Model = gsub("\\D+", "", times$Model)
+ times$Model = as.numeric(times$Model)
+ times$Time = times$Time / 10^6
+ # make the phases a factor with a fixed set of values to help dcasting
+ # (e.g. Batch measurements do not have Transformation and Recheck attributes,
+ # hence accessing the "Transformation" attribute would throw an error)
+ times$Phase = factor(times$Phase, levels = c("Read", "Check", "Transformation", "Recheck"))
+
+ times.wide = dcast(data = times,
+ formula = Tool + Workload + Description + Model + Run ~ Phase,
+ value.var = "Time",
+ drop = T,
+ fun.aggregate = mean
+ )
+
+ # calculate aggregated values
+ times.derived = times.wide
+ times.derived$Read.and.Check = times.derived$Read + times.derived$Check
+ times.derived$Transformation.and.Recheck = times.derived$Transformation + times.derived$Recheck
+
+ # calculate the median value of runs
+ times.aggregated.runs = ddply(
+ .data = times.derived,
+ .variables = c("Tool", "Workload", "Description", "Model"),
+ .fun = colwise(median),
+ .progress = "text"
+ )
+ # drop the "Run" column
+ times.aggregated.runs = subset(times.aggregated.runs, select = -c(Run))
+
+ times.processed = melt(
+ data = times.aggregated.runs,
+ id.vars = c("Tool", "Workload", "Description", "Model"),
+ measure.vars = phases,
+ variable.name = "Phase",
+ value.name = "Time"
+ )
+
+ # beautify plotted record:
+ # 1. change dots to spaces
+ # 2. make sure that the phases are still factors
+ times.plot = times.processed
+ times.plot$Phase = gsub('\\.', ' ', times.plot$Phase)
+ times.plot$Phase = factor(times.plot$Phase, levels = phasesPrettified)
+ times.plot$Workload = factor(times.plot$Workload, levels = workloads)
+
+ ### line charts
+ for (phase in phasesPrettified) {
+ phase.filename = gsub(' ', '-', phase)
+ workloadSizes = sizes[["Repair"]]
+
+ # filter the dataframe to the current phase
+ df = times.plot[times.plot$Phase == phase, ]
+
+ # do not visualize empty data sets
+ if (nrow(df) == 0) {
+ print(paste("No rows to visualize for phase", phase))
+ next
+ }
+
+ # x axis labels
+ xbreaks = unique(df$Model)
+ currentWorkloadSizes = head(workloadSizes, n=length(xbreaks))
+ xlabels = paste(xbreaks, "\n", currentWorkloadSizes, sep = "")
+
+ # drop every other models size
+ maxLabel = max(log2(max(df$Model)), 2)
+ if (maxLabel %% 2) {
+ start = 3
+ } else {
+ start = 2
+ }
+ filter = seq(start, maxLabel, by=2)
+
+ xlabels[filter] = ""
+
+ # y axis labels
+ yaxis = nice_y_axis()
+ ybreaks = yaxis$ybreaks
+ yminor_breaks = yaxis$yminor_breaks
+ ylabels = yaxis$ylabels
+
+ for (cpage in 1:6) {
+ p = ggplot(df) + #na.omit(df)) +
+ aes(x = as.factor(Model), y = Time) +
+ labs(title = paste("Individual query execution time,", phase, "phase"), x = "Model size\n#Elements", y = "Execution times [ms]") +
+ geom_point(aes(col = Tool, shape = Tool), size = 2.0) +
+ # scale_shape_manual(values = seq(0, 15)) +
+ # scale_shape_manual(#values = c(0, 15, 1, 16, 2, 17, 4, 7),
+ # # labels=c(
+ # # "Name Lookup ", "Name Lookup (Incremental)", # http://www.sthda.com/english/wiki/r-plot-pch-symbols-the-different-point-shapes-available-in-r - JastAdd NameLookup
+ # # "Intrinsic References", "Intrinsic References (Incremental)", # Tud-cyan - JastAdd Optimized
+ # # "Grammar Extension", "Grammar Extension (Incremental)", # HKS65 (green) - JastAdd Specialized
+ # # "TinkerGraph", # HKS92 (grey) - TinkerGraph
+ # # "Viatra (Incremental)", # HKS33 (violet) - VIATRA
+ # # "(none)"
+ # # )
+ # ) +
+ # scale_colour_manual(
+ # values = c(
+ # "#EE7F00", "#EE7F00", # HKS07 (orange) - JastAdd NameLookup
+ # "#009EE0", "#009EE0", # Tud-cyan - JastAdd Optimized
+ # "#6AB023", "#6AB023", # HKS65 (green) - JastAdd Specialized
+ # "#727879", # HKS92 (grey) - TinkerGraph
+ # "#93107E", # HKS33 (violet) - VIATRA
+ # "#F0E442", # Yellow
+ # "#CC79A7"),
+ # labels=c(
+ # "Name Lookup ", "Name Lookup (Incremental)", # HKS07 (orange) - JastAdd NameLookup
+ # "Intrinsic References", "Intrinsic References (Incremental)", # Tud-cyan - JastAdd Optimized
+ # "Grammar Extension", "Grammar Extension (Incremental)", # HKS65 (green) - JastAdd Specialized
+ # "TinkerGraph", # HKS92 (grey) - TinkerGraph
+ # "Viatra (Incremental)", # HKS33 (violet) - VIATRA
+ # "(none)"
+ # )) +
+ geom_line(aes(col = Tool, group = Tool), size = 0.5) +
+ scale_x_discrete(breaks = xbreaks, labels = xlabels) +
+ scale_y_log10(breaks = ybreaks, minor_breaks=yminor_breaks, labels = ylabels, limits = c(2e-3,8e4), expand = c(0, 0)) +
+ facet_grid_paginate(~ Workload, nrow=1, ncol = 1, page=cpage, scale = "free") +
+ guides(color = guide_legend(ncol = 5)) +
+ theme_bw() +
+ theme(
+ plot.title = element_text(hjust = 0.5),
+ # text = element_text(family="Open Sans", size = 10),
+ # legend.key = element_blank(),
+ legend.title = element_blank(),
+ legend.position = "bottom",
+ axis.text = element_text(size = 16)
+ #, panel.grid.minor = element_blank()
+ )
+ print(p)
+ ggsave(
+ plot = p,
+ filename = paste("../diagrams/recent/", scenario, "/", phase.filename, "-", workloads[cpage], "-", scenario, ".pdf", sep=""),
+ width = 500, height = 300, units = "mm"
)
- print(p)
- ggsave(
- plot = p,
- filename = paste("../diagrams/recent/", phase.filename, "-",workloads[cpage], ".pdf", sep=""),
- width = 500, height = 300, units = "mm"
- )
- embed_fonts(paste("../diagrams/recent/", phase.filename, "-",workloads[cpage], ".pdf", sep=""), outfile=paste("../diagrams/recent/", phase.filename, "-",workloads[cpage], "-embed", ".pdf", sep=""))
- }
+ embed_fonts(paste("../diagrams/recent/", scenario, "/", phase.filename, "-", workloads[cpage], "-", scenario, ".pdf", sep=""), outfile=paste("../diagrams/recent/", scenario, "/", phase.filename, "-",workloads[cpage], "-", scenario, "-embed", ".pdf", sep=""))
+ }
-}
+ }
+}
\ No newline at end of file