Re: [sage-support] Current advice for generic SageMath install advice

2023-03-27 Thread Isuru Fernando 
> I'm not sure that's quite enough. In my experience, sagelib needs
rebuilding to interface with normaliz (I think it's "make normaliz
pynormaliz" nowadays, or perhaps one needs a pip install). I would expect
that the binary distribution of sage for conda is built without
normaliz/pynormaliz support, because those are optional packages.
Installing these as prerequisites in conda wouldn't automatically activate
the interfaces in sagemath. Does conda do something to get that to work?
Has sagemath grown better at dynamically detecting libraries to interface
with?

pynormaliz interface is indeed dynamically detected by sage. You just have
to install pynormaliz.

There are a few optional libraries that are not dynamically detected and
need to be present at build time.
See https://github.com/sagemath/sage/blob/master/src/setup.py#L82-L83
Out of those, conda builds sage with bliss and sirocco, but they are not
installed by default when you
install sage. You need to install bliss and sirocco to get that
functionality.

Isuru

On Mon, Mar 27, 2023 at 8:35 PM Nils Bruin  wrote:

> On Monday, 27 March 2023 at 17:47:54 UTC-7 Isuru Fernando wrote:
>
>
> We used to have separate architecture specific builds, but `sage` is now a
> meta-package that is architecture neutral (i.e. noarch).
> So, you get sage-9.8 for all architectures. We support sage-9.8 for the
> following OS and architecture combinations
> - linux-x86_64  (glibc>=2.12, most distros after 2010)
> - linux-aarch64  (glibc>=2.17, most distros after 2014)
> - macos-x86_64  (macos>=10.9)
> - macos-arm64  (macos>=11.0)
> You can have a look at https://anaconda.org/conda-forge/sagelib to see
> the architecture-specific builds.
>
>
> Thanks! I was hoping for an answer like that.Then it sounds like conda is
> presently the best option to get up to date binary sage.
>
>
> > Also: if students want to use packages like normaliz, can they install
> those on binary installs? When I do it on source-built versions, it
> triggers extensive recompilation.
>
> With conda, you can install binary packages for normaliz and thousands of
> other packages into the same environment as sage.
>
>
> I'm not sure that's quite enough. In my experience, sagelib needs
> rebuilding to interface with normaliz (I think it's "make normaliz
> pynormaliz" nowadays, or perhaps one needs a pip install). I would expect
> that the binary distribution of sage for conda is built without
> normaliz/pynormaliz support, because those are optional packages.
> Installing these as prerequisites in conda wouldn't automatically activate
> the interfaces in sagemath. Does conda do something to get that to work?
> Has sagemath grown better at dynamically detecting libraries to interface
> with?
>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-support+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-support/5a6a27ad-750d-4bfc-86d6-a354f5e6a0e2n%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-support+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-support/CA%2B01voO6Ez3VuZ8mBHr5QYjDFANS0JVVxNnsdus4iGNuL0H-3g%40mail.gmail.com.

Re: [sage-support] Current advice for generic SageMath install advice

2023-03-27 Thread Nils Bruin 
On Monday, 27 March 2023 at 17:47:54 UTC-7 Isuru Fernando wrote:


We used to have separate architecture specific builds, but `sage` is now a 
meta-package that is architecture neutral (i.e. noarch).
So, you get sage-9.8 for all architectures. We support sage-9.8 for the 
following OS and architecture combinations
- linux-x86_64  (glibc>=2.12, most distros after 2010)
- linux-aarch64  (glibc>=2.17, most distros after 2014)
- macos-x86_64  (macos>=10.9)
- macos-arm64  (macos>=11.0)
You can have a look at https://anaconda.org/conda-forge/sagelib to see the 
architecture-specific builds.


Thanks! I was hoping for an answer like that.Then it sounds like conda is 
presently the best option to get up to date binary sage.
 

> Also: if students want to use packages like normaliz, can they install 
those on binary installs? When I do it on source-built versions, it 
triggers extensive recompilation.

With conda, you can install binary packages for normaliz and thousands of 
other packages into the same environment as sage.


I'm not sure that's quite enough. In my experience, sagelib needs 
rebuilding to interface with normaliz (I think it's "make normaliz 
pynormaliz" nowadays, or perhaps one needs a pip install). I would expect 
that the binary distribution of sage for conda is built without 
normaliz/pynormaliz support, because those are optional packages. 
Installing these as prerequisites in conda wouldn't automatically activate 
the interfaces in sagemath. Does conda do something to get that to work? 
Has sagemath grown better at dynamically detecting libraries to interface 
with?

-- 
You received this message because you are subscribed to the Google Groups 
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-support+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-support/5a6a27ad-750d-4bfc-86d6-a354f5e6a0e2n%40googlegroups.com.

Re: [sage-support] Current advice for generic SageMath install advice

2023-03-27 Thread Isuru Fernando 
Hi,

> However: https://anaconda.org/conda-forge/sage seems to indicate that
various architecture-specific builds are woefully outdated: "noarch" seems
to be on point, but the architecture-specific ones seem stuck on 9.2. Am I
reading the info wrong? Obviously I don't want to point people to 9.2
installs.

We used to have separate architecture specific builds, but `sage` is now a
meta-package that is architecture neutral (i.e. noarch).
So, you get sage-9.8 for all architectures. We support sage-9.8 for the
following OS and architecture combinations
- linux-x86_64  (glibc>=2.12, most distros after 2010)
- linux-aarch64  (glibc>=2.17, most distros after 2014)
- macos-x86_64  (macos>=10.9)
- macos-arm64  (macos>=11.0)
You can have a look at https://anaconda.org/conda-forge/sagelib to see the
architecture-specific builds.

> Also: if students want to use packages like normaliz, can they install
those on binary installs? When I do it on source-built versions, it
triggers extensive recompilation.

With conda, you can install binary packages for normaliz and thousands of
other packages into the same environment as sage.

Isuru

On Mon, Mar 27, 2023 at 7:34 PM Nils Bruin  wrote:

> What would the current (Spring 2023) easiest instructions be for people to
> install sage? I'm asking in support of an install-fest for students, so the
> objective is to have easy solutions for giving students access on whatever
> platform they have available.
> I know about cloud-based solutions, so I'll definitely point them to
> those. I'm asking for "the next step up".
>
> In the install advice I see:
> for OSX:
>  - binary build of SageMath
>  (looks like an
> excellent solution)
> - https://anaconda.org/conda-forge/sage (would be a great place to point
> students to, because it's a rich environment for computational software).
> However: https://anaconda.org/conda-forge/sage seems to indicate that
> various architecture-specific builds are woefully outdated: "noarch" seems
> to be on point, but the architecture-specific ones seem stuck on 9.2. Am I
> reading the info wrong? Obviously I don't want to point people to 9.2
> installs.
>
> for windows:
>  - OK WSL; that's great. However, it looks like Ubuntu would be the
> easiest linux distribution to get and as far as I can see, Ubuntu has 9.5
> packaged at most? That's not great either.
>  - conda: see above
>
> for linux:
>  - same thing.
>  - conda: see above.
>
> So is building from source the only way nowadays? That's sad. I'm fine
> doing that for myself, but for an installfest, that's really not feasible.
> Probably some machines will go in thermal meltdown as a result! Or should I
> just send them to 9.2 and 9.5 etc.
>
> Also: if students want to use packages like normaliz, can they install
> those on binary installs? When I do it on source-built versions, it
> triggers extensive recompilation.
>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-support+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-support/d1717bce-b359-4f83-902d-68edc2399d2dn%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-support+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-support/CA%2B01voMn7MBBzUAWEKpakdmDYL%3DFsSapb9e36p16QVMFtJQ5hw%40mail.gmail.com.

Re: [sage-support] Current advice for generic SageMath install advice

2023-03-27 Thread William Stein 
On Mon, Mar 27, 2023 at 5:34 PM Nils Bruin  wrote:

> What would the current (Spring 2023) easiest instructions be for people to
> install sage? I'm asking in support of an install-fest for students, so the
> objective is to have easy solutions for giving students access on whatever
> platform they have available.
> I know about cloud-based solutions, so I'll definitely point them to
> those. I'm asking for "the next step up".
>


I can send you a bunch of vouchers for free cocalc licenses if you want to
also give those out…


> In the install advice I see:
> for OSX:
>  - binary build of SageMath
>  (looks like an
> excellent solution)
> - https://anaconda.org/conda-forge/sage (would be a great place to point
> students to, because it's a rich environment for computational software).
> However: https://anaconda.org/conda-forge/sage seems to indicate that
> various architecture-specific builds are woefully outdated: "noarch" seems
> to be on point, but the architecture-specific ones seem stuck on 9.2. Am I
> reading the info wrong? Obviously I don't want to point people to 9.2
> installs.
>
> for windows:
>  - OK WSL; that's great. However, it looks like Ubuntu would be the
> easiest linux distribution to get and as far as I can see, Ubuntu has 9.5
> packaged at most? That's not great either.
>  - conda: see above
>
> for linux:
>  - same thing.
>  - conda: see above.
>
> So is building from source the only way nowadays? That's sad. I'm fine
> doing that for myself, but for an installfest, that's really not feasible.
> Probably some machines will go in thermal meltdown as a result! Or should I
> just send them to 9.2 and 9.5 etc.
>
> Also: if students want to use packages like normaliz, can they install
> those on binary installs? When I do it on source-built versions, it
> triggers extensive recompilation.
>
> --
> You received this message because you are subscribed to the Google Groups
> "sage-support" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sage-support+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sage-support/d1717bce-b359-4f83-902d-68edc2399d2dn%40googlegroups.com
> 
> .
>
-- 
-- William Stein

-- 
You received this message because you are subscribed to the Google Groups 
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-support+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-support/CACLE5GB11D8DQ%3DDW4YWXgiRB2vgRhp56oeCFb4QcvUqWkAuQaQ%40mail.gmail.com.

[sage-support] Current advice for generic SageMath install advice

2023-03-27 Thread Nils Bruin 
What would the current (Spring 2023) easiest instructions be for people to 
install sage? I'm asking in support of an install-fest for students, so the 
objective is to have easy solutions for giving students access on whatever 
platform they have available.
I know about cloud-based solutions, so I'll definitely point them to those. 
I'm asking for "the next step up".

In the install advice I see:
for OSX:
 - binary build of SageMath 
 (looks like an 
excellent solution)
- https://anaconda.org/conda-forge/sage (would be a great place to point 
students to, because it's a rich environment for computational software). 
However: https://anaconda.org/conda-forge/sage seems to indicate that 
various architecture-specific builds are woefully outdated: "noarch" seems 
to be on point, but the architecture-specific ones seem stuck on 9.2. Am I 
reading the info wrong? Obviously I don't want to point people to 9.2 
installs.

for windows:
 - OK WSL; that's great. However, it looks like Ubuntu would be the easiest 
linux distribution to get and as far as I can see, Ubuntu has 9.5 packaged 
at most? That's not great either.
 - conda: see above

for linux:
 - same thing.
 - conda: see above.

So is building from source the only way nowadays? That's sad. I'm fine 
doing that for myself, but for an installfest, that's really not feasible. 
Probably some machines will go in thermal meltdown as a result! Or should I 
just send them to 9.2 and 9.5 etc.

Also: if students want to use packages like normaliz, can they install 
those on binary installs? When I do it on source-built versions, it 
triggers extensive recompilation.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sage-support+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sage-support/d1717bce-b359-4f83-902d-68edc2399d2dn%40googlegroups.com.