improve ``Tensor::extract()``

Description

The member function Tensor::extract() could have its resilience and user interface improved regarding the lastest evolution of the class Tensor.

Proposal

Let's have a Tensor T with dimensions [3,6,6,6].

1. Currently, extracting the element [0,0,0,0] will result in an empty Tensor. With the introduction of scalar and regarding other frameworks, the first element should be extracted instead. It is now possible thanks to the introduction of scalar in Tensor class.
2. The two member functions Tensor::extract(const std::vector<std::size_t>& startCoord) and Tensor::extract(const std::vector<std::size_t>& startCoord, const std::vector<std::size_t>& dims) should be merged into a single function with a default parameter as they are very similar to avoid code duplication. Moreover, it would help with the next section.
3. Add Tensor::operator[] a variadic operator that would mimic the Tensor::extract behaviour. If we use the Numpy (so Python) interface, extracting the slice T[:2, 1:3, 4:, 5] could be done with std::pairs<std::size_t, std::size_t> and std::size_t parameters T[(0,1), (1,2), (4,5), 5]
4. By default, a dimension not mentionned in input parameters should be entirely extracted. The first extract function shows such behaviour but not the second one. It is not consistent and should be fixed.

added AddFeature label

added LanguageC++ PirorityLow StatusTODO Tensor AddUnit-test + 1 deleted label and removed AddFeature label

marked this issue as blocked by #78 (closed)

marked this issue as blocked by #84 (closed)

assigned to @pineapple

unassigned @pineapple

removed StatusTODO label