Performance

go-ruby-csv/csv is the pure-Go (CGO-free) library that rbgo binds for Ruby's csv. This page records real measured benchmarks of the library against the reference Ruby runtimes — no estimates, no placeholders.

Library-level benchmark (Go API vs runtimes) — 2026-07-03

This section measures the pure-Go library directly, through its Go API — not the rbgo interpreter. It isolates the library primitive from Ruby-interpreter dispatch, answering the parity question head-on: is the pure-Go implementation as fast as the reference runtime's own csv? The same workload, same inputs, same iteration counts run through the Go library and through each reference runtime's stdlib; outputs were checked identical to MRI before any timing.

• Host: Apple M4 Max (Mac16,5, arm64), macOS 26.5.1 — date 2026-07-03.
• Runtimes: Go 1.26.4 · MRI ruby 4.0.5 +PRISM · MRI + YJIT · JRuby 10.1.0.0 (OpenJDK 25) · TruffleRuby 34.0.1 (GraalVM CE Native).
• Method: each process runs 3 untimed warm-up passes, then 25 timed passes of a fixed inner loop, timed with a monotonic clock; the best pass is reported as ns/op (lower is better). vs MRI < 1.00× means faster than MRI. Interpreter start-up is outside the timed region, so these are operation costs, not ruby file.rb process costs.

generate-200x6

Runtime	ns/op	vs MRI
go-ruby (pure Go)	40826.8	0.19×
MRI	218082.0	1.00×
MRI + YJIT	106010.0	0.49×
JRuby	90318.4	0.41×
TruffleRuby	35323.4	0.16×

parse-200x6

Runtime	ns/op	vs MRI
go-ruby (pure Go)	76053.2	0.26×
MRI	292822.0	1.00×
MRI + YJIT	176362.0	0.60×
JRuby	1780914.8	6.08×
TruffleRuby	1068829.3	3.65×

The pure-Go library is ~4–5× faster than MRI on both parse (0.26×) and generate (0.19×). MRI's csv is implemented in Ruby, so a compiled pure-Go parser/generator wins outright; TruffleRuby is competitive on generate but its cold parse loop is an outlier.

Reproduce

The harness is committed under benchmarks/: a self-contained Go driver (go/, pins the published library via go.mod), the equivalent ruby/csv.rb workload, and run.sh. Run bash benchmarks/run.sh; env OUTER/WARM tune the pass budget and RUBY/JRUBY/TRUFFLERUBY select the runtime binaries.

Warm-up budget & noise — honest framing

Numbers reflect a fixed warm-process budget (3 warm-up + 25 timed passes in one process). The JVM/GraalVM JITs (JRuby, TruffleRuby) may need a larger warm-up to reach steady state, so their columns can understate peak throughput — most visibly TruffleRuby on the shortest loops (a few cold-JIT outliers are noted in the text). Sub-microsecond rows carry the most relative noise; treat those ratios as order-of-magnitude. Every number here is a real measured value from the dated run above — nothing is fabricated, estimated, or cherry-picked. The go-ruby column is the pure-Go library; every other column is that interpreter's own stdlib doing the equivalent work.