*This package is obsolete and archived. Please use DifferentialEquations.jl.*

An implementation of the well-known Runge-Kutta-Fehlberg time integration method of 4th and 5th order (RKF45). The algorithm integrates differential equations of the form:

```
dx / dt = f[x](t)
```

Notably, `f`

can be either a function or a functional of `x`

. This is useful for certain types of partial differential equations (e.g. the heat equation).

You import the package as usual:

```
using RungeKuttaFehlberg
```

The package exports exactly one function

```
rkf45_step(f, x, t, tolerance, dt[, error, safety])
```

which returns `dx`

and `dt`

as a tuple.
Most arguments should be self-explanatory but more detailed documentation is included in the package.

Most likely you will iterate over `rkf45_step()`

and sum up `dx`

and `dt`

.
The algorithm will run most efficiently if you pass the last return value for `dt`

back into `rkf45_step()`

at the next iteration.

The r.h.s. function `f()`

must take exactly two arguments, `x`

and `t`

. Currently, there is no way to pass additional parameters to `f()`

.
However, you can easily define an intermediate function which contains the values of each parameter and then pass it to `rkf45_step()`

.

RKF45 evaluates `f()`

*at least* 6 times during each step, so optimizing `f()`

can increase performance a lot.

I am currently hosting this in a separate package, but I am open to suggestions w.r.t. inclusion in a package for time integration methods.