Re: storage -- unbounded_array
oswin krause <[email protected]>
| Newsgroups | gmane.comp.lib.boost.ublas |
|---|---|
| Message-ID | <[email protected]> |
Hi Alex, I assume you want to create a vector/matrix from some provided external piece of memory and not a vector/matrix using c-arrays (because the latter is implemented in uBLAS and works pretty well). So take some advise from someone who went down the first road: The storage concept is not compatible with what you want to do. It -and everything that uses it - assumes that storage can be *copied* and is owned by a *single* vector/matrix. Essentially the whole concept of implementing a vector using external storage as a vector_container is seriously flawed and it can only be safely implemented as a new vector_proxy in the style of vector_range (only that it does not access another vector but the c-array). I don't know what the current status is but i heard that people wanted to do add this kind of proxy to uBLAS. Also there are already a few projects which want to "replace some custom code preserving existing API" (for example my own). Maybe all you want to do is already done in one of these projects. Best, Oswin On 22.07.2014 16:44, Alexey Luchko wrote: > Hi folks, > > I'm trying to start with boost uBLAS. > > My intention is to replace some custom code preserving existing API. > > The first goal is to create vector and matrix from a C array. > > I've found that Storage concept offers range constructor, however both > bounded and unbounded array storages do not provide it. > http://www.boost.org/doc/libs/1_55_0/libs/numeric/ublas/doc/storage_concept.htm > http://www.boost.org/doc/libs/1_55_0/libs/numeric/ublas/doc/bounded_array.htm > http://www.boost.org/doc/libs/1_55_0/libs/numeric/ublas/doc/unbounded_array.htm > > First of all, it is disappointing. Is there a special purpose for the mismatch? > > -- > Regards, > Alex > _______________________________________________ > ublas mailing list > [email protected] > http://lists.boost.org/mailman/listinfo.cgi/ublas > Sent to: [email protected]