Replaces |b| by the ordered canonical basis of the vector space $V$ it spans. Flags in |t| the set of coordinate positions associated to |b|.
What is flagged in |t| is the set $J$ in described in the comment for |normalSpanAdd| below. For any $j$, only |b[j]| has a nonzero bit at the position $j'$ of set bit number |j| of |t|; consequently, for any 
, the coordinate of |b[j]| in $v$ is $v[j']$.
This function works essentially be repeatedly calling |normalSpanAdd| for the vectors of |b|, replacing |b| by the resulting canonical basis at the end. However, the selected coordiante positions |f| do not come out increasingly this way, so we have to sort the canonical basis by leading bit position. This amounts to setting $a'[k]=a[p(k)]$ where 
is the result of sorting ![$f[0]\ldots,f[l-1]$](form_21.png)
with $l=f.size()=a.size()$.
Definition at line 700 of file bitvector_def.h.
References atlas::polynomials::compare(), normalSpanAdd(), atlas::bitset::BitSet< n >::reset(), atlas::bitset::BitSet< n >::set(), and atlas::bitset::BitSet< n >::swap().
Referenced by firstSolution(), and atlas::bitvector::BitMatrix< dim >::kernel().
such that the standard basis vectors $e_i$ for 
generate a complementary subspace $e_I$ to $V$ (one can find such an $I$ by repeatedly throwing in $e_i$s linearly independent to $V$ and previously chosen ones). The normal basis of $V$ corresponding to $I$ is obtained by projecting the $e_j$ for $j$ in the complement $J$ of $I$ onto $V$ along $e_I$ (i.e., according to the direct sum decompostion 
). This can be visualised by viewing $V$ as the function-graph of a linear map from $k^J$ to $k^I$; then the normal basis is the lift to $V$ of the standard basis of $k^J$. We define the canonical basis of $V$ be the normal basis for the complement $I$ of the lexicographically minimal possible set $J$ (lexicographic for the increasing sequences representing the subsets; in fact $I$ is lexicographically maximal since complementation reverses this ordering one fixed-size subsets). One can find this $J$ by repeatedly choosing the smallest index such that the projection from $V$ defined by extracting the coordinates at the selected indices remains surjective.