RSX Working Example

A Rails 7.1 app that uses RSX — JSX-style templates in Ruby — to render a catalog of Ruby on Rails courses. The same catalog is also implemented in isolated html.erb views so the two stacks can be compared side by side.

The Gemfile pins rsx-rb to 0.1.0.

Setup

Ruby 3.3.0.

bin/setup
bin/dev

bin/dev starts Puma and the Tailwind watcher. Then open http://127.0.0.1:3000.

Routes

Two isolated stacks. They share no views, components, or controllers.

	RSX	ERB
Home	/	/erb
Course table	/courses	/erb/courses
Course detail	/courses/:id	/erb/courses/:id

Example detail page: /courses/learn-ruby-on-rails-full-course vs /erb/courses/learn-ruby-on-rails-full-course.

Where the code lives

	RSX	ERB
Controllers	app/controllers/	app/controllers/erb/
Views	app/views/pages, app/views/courses (*.html.rsx)	app/views/erb/pages, app/views/erb/courses (*.html.erb)
Components	app/components/*.rsx	app/views/erb/components/_*.html.erb
Layout	app/views/layouts/application.html.erb	app/views/erb/layouts/_document.html.erb
Helpers	—	app/helpers/erb/

Both controllers pass the same hardcoded course list (name, hours, YouTube URL, channel, thumbnail) into a nested table: catalog → table → row → thumbnail / meta / duration / watch.

Tests

bin/rails test

That runs:

• Functional coverage for the RSX pages (/, /courses, /courses/:id)
• Functional coverage for the ERB pages (/erb, /erb/courses, /erb/courses/:id)
• A performance suite that prints an RSX vs ERB comparison to the terminal

Sample size defaults to 60 timed iterations after 8 warmups. Override if you want a longer or shorter run:

PERF_ITERATIONS=120 PERF_WARMUP=15 bin/rails test

RSX vs ERB performance

Ratio is ERB / RSX. Below 1.00 means ERB was faster; above 1.00 means RSX was faster. Allocs are Ruby objects per call. The missing-course path is dominated by Rails exception handling, not the template engine, so it is excluded from the “Rendering + requests” line.

Numbers will move with hardware, Ruby, and PERF_ITERATIONS. Re-run bin/rails test for a fresh table.

rsx-rb 0.1.0

==============================================================================
RSX vs ERB performance comparison
Iterations: 60 timed (8 warmup)  |  times are per-call averages
==============================================================================
Scenario                     RSX ms     ERB ms      Ratio      Winner
------------------------------------------------------------------------------
GET home                      2.134      3.941      1.85x         RSX
GET courses index             1.961      4.832      2.46x         RSX
GET course show               1.852      3.651      1.97x         RSX
GET missing course          239.970    251.033      1.05x         tie
render course table           0.590      2.386      4.04x         RSX
render course row             0.383      0.458      1.20x         RSX
render course detail          0.528      1.637      3.10x         RSX
render empty table            0.317      0.163      0.52x         ERB
render duration badges        0.348      0.662      1.91x         RSX
------------------------------------------------------------------------------
Rendering + requests          8.113     17.731      2.19x
Allocs / call                 10369      17785      1.72x
All scenarios               248.082    268.764      1.08x

Ratio is ERB / RSX. Below 1.00 means ERB was faster; above 1.00 means RSX was faster.
Allocs are Ruby objects per call. Override with PERF_ITERATIONS and PERF_WARMUP.
==============================================================================

On this machine, RSX 0.1.0 was faster on the real page and component renders (about 2.19× on the rendering + request total, and 1.72× fewer allocations).

rsx-rb 0.2.1

==============================================================================
RSX vs ERB performance comparison
Iterations: 60 timed (8 warmup)  |  times are per-call averages
==============================================================================
Scenario                     RSX ms     ERB ms      Ratio      Winner
------------------------------------------------------------------------------
GET home                      1.900      3.775      1.99x         RSX
GET courses index             1.886      4.602      2.44x         RSX
GET course show               1.940      3.444      1.77x         RSX
GET missing course          241.206    246.333      1.02x         tie
render course table           0.610      2.630      4.31x         RSX
render course row             0.395      0.502      1.27x         RSX
render course detail          0.538      1.591      2.96x         RSX
render empty table            0.311      0.162      0.52x         ERB
render duration badges        0.335      0.631      1.88x         RSX
------------------------------------------------------------------------------
Rendering + requests          7.915     17.336      2.19x
Allocs / call                 10479      17785      1.70x
All scenarios               249.121    263.669      1.06x

Ratio is ERB / RSX. Below 1.00 means ERB was faster; above 1.00 means RSX was faster.
Allocs are Ruby objects per call. Override with PERF_ITERATIONS and PERF_WARMUP.
==============================================================================

0.2.1 keeps the same 2.19× rendering + request advantage vs ERB, with slightly lower RSX wall time on that total (7.915 ms vs 8.113 ms in 0.1.0). Allocations are about 1.70× fewer than ERB.