pg1003/golomb

A C++ library to compress integer values using Exponential Golomb Encoding

golomb's Issues

pg::golomb::encode unsafe output iterator

The encode function has a generic type OutputIt.
The name suggests it should be an output iterator.
Maybe a concept/requirement can be used to force the OutputIt to at least be an output iterator of correct type (by using std::output_iterator)
However, since the size of the output is unknown, there is a risk of running out of space in the container.

In the examplecode test.cpp, std::back_inserters are used everywhere. This is always* safe.
However, suppose the vector result was preallocated or was an std::array instead, then it would be unsafe since it might occur that the result takes more than the preallocated amount of elements.
There is no way to know beforehand, and there is no way to test currently, since inside the function, the original container cannot be accessed.

Two suggested solutions:

• Make the parameter a reference to a container and use the std::back_inserter within the API, so e.g. line 201:

encoder< OutputIt, OutputDataT > e( output, k );

Turns into

encoder< std::back_inserter<OutputContainer>, OutputDataT > e( std::back_inserter<outputcontainer>, k );

, where outputcontainer is required to be some container.
This is still unsafe, since a std::array might not have the correct size either, and the std::back_inserter will fail at some point

• return a container by moving it out of the encode function. In this way, the correct size is guaranteed.
  The type of container could be specified by template parameter (with requirement of being a container with back_inserter)