feat: add benchmark script

scripts/run_bench.sh

@@ -0,0 +1,19 @@
+#!/usr/bin/env bash
+# Compare runtime of rdfpipe vs rdfpipe-rs
+# hyperfine is the only dependency (besides rdfpipe and rdfpipe-rs)
+set -euo pipefail
+
+# File path to a (large) ntriples RDF dataset
+DATASET="$1"
+RDFPIPE_PY="rdfpipe"
+RDFPIPE_RS="./target/release/rdfpipe-rs"
+
+# Run both implementations with different number of triples
+# timings are saved in timings.csv
+hyperfine \
+    --warmup 1 \
+    -L N 1,2,3,4,5,10,15,20,50 \
+    -L FMT ttl,xml \
+    --export-csv timings.csv \
+    "head -n {N}000 ${DATASET} | ${RDFPIPE_PY} -i nt -o {FMT} - > /dev/null" \
+    "head -n {N}000 ${DATASET} | ${RDFPIPE_RS} -i nt -o {FMT} - > /dev/null"

scripts/viz_bench.R

@@ -0,0 +1,32 @@
+# Visualization of timings for rdfpipe vs rdfpipe-rs
+# tidyverse>=1.1.3 is the only dependency
+
+library(tidyverse)
+
+timings <- read_csv("timings.csv")
+
+timings <- timings %>%
+    rename(
+        tool = command,
+        thousand_lines = parameter_N,
+        fmt = parameter_FMT
+    ) %>%
+    mutate(tool = case_when(
+        str_detect(tool, "rdfpipe-rs") ~ "rdfpipe-rs",
+        TRUE ~ "rdfpipe"
+    )) %>%
+    select(tool, mean, fmt, stddev, thousand_lines) %>%
+    arrange(thousand_lines, tool)
+
+ggplot(timings, aes(x = thousand_lines, y = log10(mean), color = tool)) +
+    geom_line() +
+    xlab("Thousands of lines parsed") +
+    ylab("Log10 time (seconds)") +
+    theme_bw(base_size = 18) +
+    coord_fixed(ratio = 10) +
+    facet_grid(~fmt, labeller = labeller(
+        fmt = c(
+            "ttl" = "ntriples -> turtle",
+            "xml" = "ntriples -> xml"
+        )
+    ))