> For the complete documentation index, see [llms.txt](https://doc.hfsm.dev/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.hfsm.dev/general-information/development-principles.md).

# Development Principles

## High Performance

[HFSM2](https://github.com/andrew-gresyk/HFSM2) uses template meta-programming to generate **optimal code** for a given FSM topology.

It does not use performance-prohibitive dynamic allocation, virtual methods, complex algorithms.

The structural information is **statically encoded** into the FSM instance **type**, which removes the need for complex lookups, and internal bookkeeping logic can always **access** the required information **directly**.

## Small Footprint

With memory access being the performance bottleneck on modern architectures, small data structures **fit better** in memory **caches**, not only resulting in **better performance** for the FSM instance itself, and also leaving more **resources available** to be used by the other parts of the host application.

Additionally, low memory requirements make [HFSM2](https://github.com/andrew-gresyk/HFSM2) a **good fit** for the memory-constrained environments including **embedded** systems and **robotics**.

## Powerful Feature Set

[HFSM2](https://github.com/andrew-gresyk/HFSM2) evolution is driven by the **development experience** rather than dry mathematical theory.

The tools provided by the library are not restricted to conform to an external standard, and the ideas for the **new features** come from from both the **solutions** to the **typical** problems encountered in development, and **competing** decision making **methodologies**, including behavior trees, utility theory and AI planners.

## No Artificial Limitations

[HFSM2](https://github.com/andrew-gresyk/HFSM2) will not hold you back following questionable conventions.\
It's goal is to **empower the developer** instead.

It does not support any artificial standards (including UML), but rather **provides tools** that make sense in a given setting.

There are no superfluous **transition tables**, beyond the one **defined implicitly** in your client code.