Opened 5 years ago
Closed 7 weeks ago
#20739 closed enhancement (duplicate)
Make snappy a pip package
Reported by: | vdelecroix | Owned by: | |
---|---|---|---|
Priority: | major | Milestone: | sage-duplicate/invalid/wontfix |
Component: | packages: optional | Keywords: | days74, snappy |
Cc: | culler, mgoerner, dunfield, mmarco, tscrim, slelievre | Merged in: | |
Authors: | Reviewers: | Matthias Koeppe | |
Report Upstream: | N/A | Work issues: | |
Branch: | Commit: | ||
Dependencies: | Stopgaps: |
Description (last modified by )
Make SnapPy an "official" optional package. With the branch apply you just need to run
sage -i snappy
Sadly, the options such as --user
are not transmitted to pip
. So for a user wide installation, one still need to use pip version
sage -pip install snappy --user
In a follow-up ticket, we could ensure compatibility of the Sage Link class and the SnapPy
link class via some doctests.
Note: a Docker image for Sage + SnapPy + Regina + PHCPack + more is available:
- Docker image: https://hub.docker.com/r/computop/sage/
- code repository: https://bitbucket.org/t3m/sagedocker
- dockerfile: https://bitbucket.org/t3m/sagedocker/src/tip/sage/Dockerfile
This ticket is superseded by #31180.
Change History (16)
comment:1 Changed 5 years ago by
- Branch set to public/20739
- Commit set to be1b40d99510bbbeeacc60b66fd0fcf8e8a3a03f
- Description modified (diff)
comment:2 Changed 5 years ago by
- Description modified (diff)
comment:3 Changed 5 years ago by
Nice!
How exactly do you imagine the integration with our Link class?
comment:4 Changed 5 years ago by
As a first step, Nathan tought us the sage_link
method
sage: sage: import snappy sage: L = snappy.Link('K10n11') sage: K = L.sage_link() sage: K.alexander_polynomial() -t^-3 + 5*t^-2 - 9*t^-1 + 11 - 9*t + 5*t^2 - t^3 sage: L.alexander_polynomial() t^6 - 5*t^5 + 9*t^4 - 11*t^3 + 9*t^2 - 5*t + 1
And there is a nice conversion from Sage link to SnapPy? handled automatically
sage: B = BraidGroup(8) sage: K = Link(B([-1, -1, -1, -2, 1, -2, 3, -2, 3])) sage: L = snappy.Link(K)
This might already help for:
- double checking the outputs (from both sides)
- comparing the implementations
To enhance it further we could make work the natural conversion the other way around
sage: L = snappy.Link('K10n11') sage: Link(L)
comment:5 follow-up: ↓ 6 Changed 5 years ago by
Something that would be good to have upstream is a _sage_
method as that is the standard way to call something back to something Sage knows about. In any case, our Link
code should just call the appropriate method in the SnapPy Link
(and as a bonus, we can pass strings to SnapPy and convert those back to Sage so the knot names work).
As for passing the --user
option, we should talk to Jeroen or Volker about that.
comment:6 in reply to: ↑ 5 Changed 5 years ago by
Replying to tscrim:
Something that would be good to have upstream is a
_sage_
method as that is the standard way to call something back to something Sage knows about.
I just added the _sage_
method to Spherogram, essentially as an alias for the sage_link
method mentioned above.
comment:7 Changed 5 years ago by
- Reviewers set to Jeroen Demeyer
- Status changed from new to needs_review
- Summary changed from Make snappy an optional package to Make snappy a pip package
comment:8 Changed 5 years ago by
- Status changed from needs_review to positive_review
comment:9 Changed 5 years ago by
- Description modified (diff)
comment:10 Changed 5 years ago by
- Status changed from positive_review to needs_work
I'm concerned that pip is not being passed the flag --no-binary :all:
(which is equivalent to the old --no-use-wheel
). On OS X, the binary wheel we post on PyPI is built against a "Frameworks" Python rather than the "plain Unix" Python that ships with Sage. I couldn't reproduce this when I tried right now, but I have observed in the past where "sage -pip install snappy" on OS X results in installing the binary wheel from PyPI; this results in a fatal linking error when trying to load the module.
This is not a problem on Linux simply because we currently don't ship binary wheels for Linux --- the standard for that was only recently finalized --- but that will change soon.
comment:11 Changed 5 years ago by
- Cc culler mgoerner added
comment:12 Changed 5 years ago by
Please take note of this SMC issue and my proposal there. The problem is that snappy
is also the name of a Python module for the compression library python-snappy, which is used in the very popular machine-learning package tensorflow.
I propose there to use relative imports to avoid this clash.
comment:13 Changed 3 years ago by
- Cc slelievre added
- Description modified (diff)
- Keywords snappy added
Adding the following note to the ticket description, for reference.
Note: a Docker image for Sage + SnapPy + Regina + PHCPack + more is available:
- Docker image: https://hub.docker.com/r/computop/sage/
- code repository: https://bitbucket.org/t3m/sagedocker
- dockerfile: https://bitbucket.org/t3m/sagedocker/src/tip/sage/Dockerfile
comment:14 Changed 7 weeks ago by
- Description modified (diff)
- Milestone changed from sage-7.3 to sage-duplicate/invalid/wontfix
Probably superseded by #31180.
If you agree, remove the branch and set the milestone to "duplicate".
comment:15 Changed 7 weeks ago by
- Reviewers changed from Jeroen Demeyer to Matthias Koeppe
- Status changed from needs_work to positive_review
comment:16 Changed 7 weeks ago by
- Branch public/20739 deleted
- Commit be1b40d99510bbbeeacc60b66fd0fcf8e8a3a03f deleted
- Resolution set to duplicate
- Status changed from positive_review to closed
New commits:
Trac 20739: snappy as an "official" package