The source of the bytes can still be files.  Just a separate step.  As a 
library, both options can be available. 






On Thursday, September 24, 2020, 01:03:42 p.m. EDT, 'Zhihao Yuan' via Source 
<[email protected]> wrote: 





On Thu, Sep 24, 2020 at 11:24 AM 'Pascal Jasmin' via Source <
[email protected]> wrote:

> Even if patch weren't accepted for some reason, is there a library version
> that can be called from J?


There is a C++ library called "cnpy" but of
relately low quality.  The npy parser offered
in this patch is as solid as that in NumPy.
Xtensor-io is header-only and requires more
dependencies.


> Less, but still, useful would be conversion to/from 3!:1 format.
>
> For consistency with 3!:1, likely should be 2 monadic foreigns (if
> encoding does not require a dyadic control parameter) that operate on
> strings rather than files.
>

We want to avoid exchanging data in J
byte strings, because the npy files can
be giga bytes large.


-- 
Zhihao
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to