Back to TIM Core implementation
The Pastel library defines the type aliases integer and real to work as abstractions
for the corresponding mathematical concepts. In TIM, instead of making the underlying
computation type a template parameter, we have chosen to use the real type throughout.
We believe this simplifies the library while not making it substantially less generic.
The real type can be chosen library-wide a float, a double, or a long double based
on the desired configuration. For the precompiled software we used the double type.
The integer type is an alias for the int type.
TIM uses iterator ranges to abstract away the differences
between different kinds of containers, as well as to make up for
an efficient implementation. An iterator range is essentially
a triple (b, e, d) where b and e form an iterator range [b, e[ and
d is the distance between e and b. Apart from making the interface
of functions compact, iterator ranges also offer a form of concept-based
overloading, compared to non-restricted templates when using only iterators.
Whenever a template function parametrizes an iterator range, the types
of the iterators are named in the form DereferencedType_Iterator. Here is
an example function template declaration from TIM:
template <
ranges::forward_range Signal_Range,
typename Real_OutputIterator>
integer temporalDifferentialEntropyKl(
const Signal_Range& signalSet,
integer timeWindowRadius,
Real_OutputIterator result,
integer kNearest = 1);
Iterator ranges do not always make sense for output. For example,
if the output of the function is varying in number of writes, as with
the above function, then no fixed-size iterator range does. Therefore,
outputs are always specified with a single output iterator. Whenever a
template function parametrizes an output iterator, the type is named in the
form DereferencedType_OutputIterator.
You can read more about special iterators and iterator ranges from the Pastel library’s documentation.
Simply imports the type definitions of the Pastel library