What's Changed
* New default renderer with 2-6x better performance by volfpeter in https://github.com/volfpeter/htmy/pull/25
**Full Changelog**: https://github.com/volfpeter/htmy/compare/v0.3.6...v0.4.0
Additional notes from 25
- `htmy.HTMY` should be considered deprecated and replaced by `htmy.Renderer`.
- `htmy.Renderer` is a completely new renderer implementation with 2-6x performance compared to the renderer in version `0.3.6`, largely depending of course on the use-case.
- `htmy.renderer` is now a package that offers the default `Renderer` of the lib, and also a `BaselineRenderer` (the original renderer of the lib). The reason for keeping the latter is its simplicity and easy to prove correctness. It can be used for debugging, testing, benchmarking other implementations.
Regarding performance, the improvement you can expect strongly depends on the composition and depth of your component tree. The 2-6x improvement (compared to `0.3.6`) can be expected for component trees that have a mixture of sync components, async components, and `ErrorBoundary`s. Some general recommendations:
- Sync components can be resolved faster, so (as recommended in the docs) if a component doesn't need to be async, don't make it async.
- Less nesting improves performance. If a component is mostly "boilerplate" (some HTML layout with dynamic content in a couple of places), then you could consider using `Snippet` instead with a `text_processor` to fill in the dynamic parts.
- `ErrorBoundary` components can be relatively expensive, use them only where they are necessary.
With the tips above, you can potentially expect even better performance. This will still be considerably slower than Jinja for example, but there is no free lunch. For really performance critical parts, you could still create a wrapper component that uses a different rendering engine internally and return its result as a `SafeStr`.