Hi!
I added:
serialization_context.register_type(
np.matrix, 'np.matrix',
custom_serializer=_serialize_numpy_array_list,
custom_deserializer=_deserialize_numpy_array_list)
But it did not help. Probably also because np.matrix is a subclas of
np.ndarray anyway. So no change here.
An interesting fact is that this worked in older versions of numpy,
but stopped in numpy 1.15.2. It works with numpy 1.14.3. So it is them
changing something.
Mitar
On Sat, Mar 30, 2019 at 3:34 PM Philipp Moritz <[email protected]> wrote:
>
> Hey Mitar,
>
> It might be as simple as adding a handler here:
> https://github.com/apache/arrow/blob/master/python/pyarrow/serialization.py#L300
>
> Do you want to try that?
>
> -- Philipp.
>
> On Sat, Mar 30, 2019 at 3:22 PM Mitar <[email protected]> wrote:
>
> > Hi!
> >
> > I do not know where to start looking into this? Not sure if I have
> > enough knowledge about arrow to be able to make a PR.
> >
> >
> > Miar
> >
> > On Sat, Mar 30, 2019 at 3:17 PM Wes McKinney <[email protected]> wrote:
> > >
> > > hi Mitar,
> > >
> > > I see you reported the issue on October 2 and no one has volunteered
> > > to fix it yet. Are you up to submit a PR?
> > >
> > > Thanks
> > > Wes
> > >
> > > On Sat, Mar 30, 2019 at 5:14 PM Mitar <[email protected]> wrote:
> > > >
> > > > Hi!
> > > >
> > > > It seems numpy's matrix is not supported in recent versions of pyarrow:
> > > >
> > > > https://issues.apache.org/jira/browse/ARROW-3399
> > > >
> > > > Any ideas why this would be happening?
> > > >
> > > >
> > > > Mitar
> > > >
> > > > --
> > > > http://mitar.tnode.com/
> > > > https://twitter.com/mitar_m
> >
> >
> >
> > --
> > http://mitar.tnode.com/
> > https://twitter.com/mitar_m
> >
--
http://mitar.tnode.com/
https://twitter.com/mitar_m
