Benchmarks

Citry is built to render fast, and the repository ships a benchmark that keeps that claim honest. It renders one large, realistic page and compares citry against django-components, plain Django templates, and Jinja2.

The scenario is a full page with about 325 component instances and roughly 205 KB of HTML in a single render. In the chart below, lower bars mean faster renders.

Versus django-components

django-components is the fair comparison, because both it and citry pay the full cost of a component lifecycle on every render: constructing components, resolving slots, and collecting JS and CSS dependencies. On this page citry comes out ahead across the board.

Versus bare template engines

Plain Django and Jinja2 render no components at all, so they skip the work citry does on every render. That head start shows up in the raw numbers, but the gap is smaller than it sounds: citry's repeat render is only about 1.3x the time of a bare Django template, even though Django is doing none of the component work.

Jinja2, the no-component baseline

Jinja2 is the fast baseline with no component model, where each citry component becomes a precompiled macro. That makes it the quickest engine to start up and the quickest once warm, since a warm render just runs precompiled macro code. It pays for that on the first render, where its whole macro library compiles at once.

Reading these numbers

These are relative numbers from a single machine, not absolute guarantees.

• "Citry renders this page N times faster than django-components here" is a fair reading.
• "A render takes X milliseconds, so my page will take X" is not: a real page has a different mix of components, templates, and data.
• Never compare numbers across machines, runs, or build profiles.

Reproduce it

The full methodology, the exact engine versions, and step-by-step instructions to run the comparison yourself live in the benchmarks README: benchmarks/README.md.

One trap worth repeating from there: the Rust extension must be built in release mode before measuring. A debug build makes citry's Rust-backed paths many times slower and invalidates every citry number.

Related pages

• Performance covers Const, the opt-in optimization that renders a template's unchanging parts once and reuses them across renders.