Re: [sage-devel] Single instance of R only
On Mon, Dec 24, 2018 at 2:01 PM Timo Kaufmann wrote:
>>
>> When I last looked, I think the rpy2 interface happens entirely at the C
>> library level, and there is only one process involved.
>
>
> Yes, that is true. Working at the C level has a lot of advantages, but there
> is only one process involved. It may be possible to work around this either
> by introducing multiple R processes upstream or by doing some multiprocessing
> hackery.
>
>>
>> It's really unfortunately if somebody just deleted the pexpect interface and
>> replaced it with rpy2. Instead, I would think both should be supported,
>> since they have complementary functionality (pros and cons to each)... Oh
>> well.
>
>
> That is exactly what I did in https://trac.sagemath.org/ticket/26596. The
> reason for it was that the pexpect interface was very hard to maintain and
> near impossible to update to R 3.5. While implementing the rpy2 interface I
> also noticed that it was broken in various ways, e.g.:
>
> - `r('NaN').sage()`, `r('Inf').sage()`, `r('3i + 4').sage()` all threw errors
> because parsing failed
> - errors were ignored entirely due to faulty pattern matching (leading to
> some errors in sage code)
> - `r.help` would block indefinitely when multiple help pages are available
>
> And I would guess there are more. That is why I don't think "maintaining
> both" is a viable option. The repl isn't meant to be parsed and if we can get
> someone else to maintain the interface for us (rpy2), that is great.
Just to be clear, I fully support your choice.By "both should be
supported, since they have complementary functionality (pros and cons
to each)... Oh well." I mean "I personally think both should be
supported. However, since I do not have the time or resources right
now to do so, then whoever does have those resources (you) should
decide." Thank you for taking the time to explain your choices
above, and more about the issues.
-- William
--
William (http://wstein.org)
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] Single instance of R only
>
> When I last looked, I think the rpy2 interface happens entirely at the C
> library level, and there is only one process involved.
>
Yes, that is true. Working at the C level has a lot of advantages, but
there is only one process involved. It may be possible to work around this
either by introducing multiple R processes upstream or by doing some
multiprocessing hackery.
> It's really unfortunately if somebody just deleted the pexpect interface
> and replaced it with rpy2. Instead, I would think both should be
> supported, since they have complementary functionality (pros and cons to
> each)... Oh well.
>
That is exactly what I did in https://trac.sagemath.org/ticket/26596. The
reason for it was that the pexpect interface was very hard to maintain and
near impossible to update to R 3.5. While implementing the rpy2 interface I
also noticed that it was broken in various ways, e.g.:
- `r('NaN').sage()`, `r('Inf').sage()`, `r('3i + 4').sage()` all threw
errors because parsing failed
- errors were ignored entirely due to faulty pattern matching (leading to
some errors in sage code)
- `r.help` would block indefinitely when multiple help pages are available
And I would guess there are more. That is why I don't think "maintaining
both" is a viable option. The repl isn't meant to be parsed and if we can
get someone else to maintain the interface for us (rpy2), that is great.
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] Single instance of R only
On Mon, Dec 24, 2018, 7:21 AM Andrey Novoseltsev On Sunday, 23 December 2018 19:23:27 UTC-7, William wrote: >> >> I didn't have anything to do with the change, but for what it is >> worth, it was definitely not my intention that calling R() twice only >> creates one interface. >> When I designed and implemented the interfaces, my intention was that >> it be possible to create multiple interfaces to multiple copies of R >> (or to any >> other pexpect interface). >> >> -- William >> >> > Well, that was also my impression given how Interface class starts: > > class Interface(WithEqualityById, ParentWithBase): > """ > Interface interface object. > > .. NOTE:: > > Two interfaces compare equal if and only if they are identical > objects (this is a critical constraint so that caching of > representations of objects in interfaces works > correctly). Otherwise they are never equal. > """ > > The change for R() happened since it is no longer a pexpect interface, but > rpy2 one. I have no idea how exactly it works and if it is possible to use > it to create multiple instances or namespaces with it, but perhaps > knowledgeable people should look into it. > When I last looked, I think the rpy2 interface happens entirely at the C library level, and there is only one process involved. It's really unfortunately if somebody just deleted the pexpect interface and replaced it with rpy2. Instead, I would think both should be supported, since they have complementary functionality (pros and cons to each)... Oh well. -- William > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
Re: [sage-devel] Single instance of R only
On Sunday, 23 December 2018 19:23:27 UTC-7, William wrote: > > I didn't have anything to do with the change, but for what it is > worth, it was definitely not my intention that calling R() twice only > creates one interface. > When I designed and implemented the interfaces, my intention was that > it be possible to create multiple interfaces to multiple copies of R > (or to any > other pexpect interface). > > -- William > > Well, that was also my impression given how Interface class starts: class Interface(WithEqualityById, ParentWithBase): """ Interface interface object. .. NOTE:: Two interfaces compare equal if and only if they are identical objects (this is a critical constraint so that caching of representations of objects in interfaces works correctly). Otherwise they are never equal. """ The change for R() happened since it is no longer a pexpect interface, but rpy2 one. I have no idea how exactly it works and if it is possible to use it to create multiple instances or namespaces with it, but perhaps knowledgeable people should look into it. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
