| Re: [eigen] documentation |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
On 26.10.2014 23:46, Mathieu Dutour wrote:
I have many questions, first of all on the documentation.
1> Is there somewhere a full documentation in .pdf of
the software? That would help quite a lot.
We don't generate that automatically, but with some effort you can
generate that using the (existing) doxygen documentation (see below).
2> Related to that, when I do "cmake doc" in the top
directory, I get following error
For cmake you need to create a separate (build) directory, go into that
directory, then type
$ cmake {PATH_TO_EIGEN}
afterward, from that build directory you call
$ make doc
This should automatically generate html documentation (the same as
available via our homepage). For latex/pdf documentation look into
eigen/doc/Doxyfile.in for GENERATE_LATEX and set it to YES (and run
`make doc` again). This should also solve your 3rd issue.
4> Is it possible to have a matrix whose size naturally grows?
I mean that in the enumration rays are continuously created and
need to be appended to the list (i.e. the matrix). Is it possible
to do that in eigen?
Similar things have been requested from time to time, this was never
considered a primary use-case of Eigen (it might come in future versions
-- but most likely not yet in 3.3).
You can use the conservativeResize function which keeps the existing
elements of your matrix valid, but that is not very efficient (every
resizing requires re-allocating and copying of the entire matrix).
A (slightly) better idea is to start with a matrix big enough to hold
the expected number of elements, then fill it column-by-column (or
row-by-row) and address the currently valid elements with
storage_matrix.leftCols(numberOfValidElements)
Alternatively, store your data in a std::vector and access it using
Eigen::Map.
5> Is there a way to have some number theoretic functions in
eigen? For example finding the integrl nullspace of a matrix?
AFAIK, we don't provide anything at the moment (in fact, all
decomposition lead to erroneous results in most cases for integral types).
But you can use Eigen for basic integer matrix operations
(multiplication and addition, accessing sub-blocks, etc) if that helps
you implementing anything in that direction.
If someone has a mature enough implementation of higher level integral
operations, we can debate about adding it as an unsupported module.
Christoph
--
----------------------------------------------
Dipl.-Inf., Dipl.-Math. Christoph Hertzberg
Cartesium 0.049
Universität Bremen
Enrique-Schmidt-Straße 5
28359 Bremen
Tel: +49 (421) 218-64252
----------------------------------------------