Algebraic polyhedra with Normaliz / e-antic

[mkoeppe]

Implements polyhedra over embedded algebraic number fields.

• Preliminary setup within sage: install e-antic, upgrade normaliz/pynormaliz (see #27682)

• the input is allowed to come from various fields, which will be coerced to a number field first via #20181:

          sage: x = polygen(ZZ); P = Polyhedron(vertices=[[sqrt(2)], [AA.polynomial_root(x^3-2, RIF(0,3))]], backend='normaliz')
  

• See doctests in src/sage/geometry/polyhedron/backend_normaliz.py and src/sage/geometry/polyhedron/library.py

• For debugging help, we can write out Normaliz input files and PyNormaliz? function calls as a side effect by passing verbose=True.

[vdelecroix]

prerequisite: package ​antic (that has no release yet).

[mkoeppe]

Last 10 new commits:

​117e428	Upgrade normaliz to 3.5.3
​7057ef4	Merge branch to get docstring adaptation
​b6ae686	Merge branch 'develop' of git://trac.sagemath.org/sage into t/25090/pynormaliz2
​eaedcc2	Update PyNormaliz to 1.14
​4dbdc7d	Updated the Cone calls format
​1dcef4c	_init_from_normaliz_data: New, use it in _init_from_*representation
​fcd8101	Updated pynormaliz to 1.15
​19d8539	Polyhedron_normaliz: In verbose mode, write out normaliz format files
​4cfcd14	_normaliz_format: Fix last change
​ea7a6b8	Polyhedron_normaliz: Prepare classes for work over field extensions

[mkoeppe]

Could someone who has worked on / has opinions on the Polyhedron constructor code help here?

What should we do for polyhedra over AA such as the following:

sage: P = polytopes.regular_polygon(3)
sage: P.base_ring()
Algebraic Real Field

If backend=normaliz is requested, should we cast to the appropriate number field via #20181? Or should it be an error?

[jipilab]

What should we do for polyhedra over AA such as the following:

sage: P = polytopes.regular_polygon(3)
sage: P.base_ring()
Algebraic Real Field

If backend=normaliz is requested, should we cast to the appropriate number field via #20181? Or should it be an error?

I would say the cleanest thing to do once #20181 is done will be to use it! This probably will also simplify the constructor for field.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​26dc3e4	Added option for embedded NumberField
​db778ee	Made it possible to deal with complex number fields
​13d27d4	Merge remote-tracking branch 'trac/public/algnum_to_numberfield' into t/25097/public/25097/qnormaliz-algebraic

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​5af97b3

Polyhedron_normaliz: Separate constructors for QQ, general fields (no cnaceling)

[mkoeppe]

I put some preliminary code in to do the coercion to number fields inside the Polyhedron_normaliz implementation, using number_field_elements_from_algebraics.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​0f325c0	Polyhedron: Allow SR input for normaliz
​aa22d44	Polyhedron_normaliz._nmz_result: New, use it instead of directly calling PyNormaliz.NmzResult
​7e4f1ac	Polyhedron_normaliz: _init_from_Vrepresentation: Coerce to numberfield
​0ba9f6b	Polyhedron_normaliz: Pass number field to QNormaliz
​d2aee02	_init_from_normaliz_data: Separate implementations for QNormaliz, Normaliz

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​2bb8958	_number_field_elements_from_algebraics_list_of_lists: New, refactor Polyhedron_normaliz using it
​a9045bf	Added tests and made them pass
​09df646	Merge remote-tracking branch 'trac/public/algnum_to_numberfield' into t/25097/public/25097/qnormaliz-algebraic
​01d7cfc	lists of lists of lists of algebraic numbers, actually

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​d7ecb03

Temporary build scripts for normaliz, pynormaliz, pyqnormaliz

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​8f675ad	Add pyqnormaliz
​79bda61	Polyhedron_normaliz: Update for pyqnormaliz

[jipilab]

A note: I am trying out the method in the ticket description and I get a ImportError: No module named PyQNormaliz_cpp when I try with an example.

What should I do?

[jipilab]

Replying to jipilab:

A note: I am trying out the method in the ticket description and I get a ImportError: No module named PyQNormaliz_cpp when I try with an example.

What should I do?

Coming from the installation of normaliz through sage:

[normaliz] Installing ARB...
[normaliz] curl: (1) Protocol "https" not supported or disabled in libcurl
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

[jipilab]

Replying to jipilab:

Replying to jipilab:

A note: I am trying out the method in the ticket description and I get a ImportError: No module named PyQNormaliz_cpp when I try with an example.

What should I do?

Coming from the installation of normaliz through sage:

[normaliz] Installing ARB...
[normaliz] curl: (1) Protocol "https" not supported or disabled in libcurl
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

To fix this, I had to force install curl inside a sage shell. On a debian-sid distribution, it seems that it was not necessary.

[jipilab]

After installing autotools and automake I get the following error message:

[normaliz] Installing E-ANTIC...
[normaliz] From https://github.com/mkoeppe/e-antic
[normaliz]  * branch            winfried   -> FETCH_HEAD
[normaliz] HEAD is now at 9de1e6f... renf_elem_class.get_den: New
[normaliz] aclocal: warning: couldn't open directory 'm4': No such file or directory
[normaliz] configure.ac:9: error: possibly undefined macro: AC_PROG_LIBTOOL
[normaliz]       If this token and others are legitimate, please use m4_pattern_allow.
[normaliz]       See the Autoconf documentation.
[normaliz] autoreconf: /usr/bin/autoconf failed with exit status: 1
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

Then, after installing libtool (sage should complain about that not installed?!) then I get:

[normaliz] make[2]: Leaving directory '/home/jplabbe/sage/local/enfnormaliz/E-ANTIC_source/e-antic'
[normaliz] configure: error: cannot find install-sh, install.sh, or shtool in "." "./.." "./../.."
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

[jipilab]

... Problem solved by gh-sebasguts: there was an old config file lying around. Deleting it solved the issue.

[moritz]

works flawlessly for me!

[gh-TimmyChan]

Replying to jipilab:

Replying to jipilab:

Replying to jipilab:

A note: I am trying out the method in the ticket description and I get a ImportError: No module named PyQNormaliz_cpp when I try with an example.

What should I do?

Coming from the installation of normaliz through sage:

[normaliz] Installing ARB...
[normaliz] curl: (1) Protocol "https" not supported or disabled in libcurl
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

To fix this, I had to force install curl inside a sage shell. On a debian-sid distribution, it seems that it was not necessary.

Same issue in Ubuntu 16.04LTS, doing sage -f curl in a sage shell does not work. Purging curl and reinstalling also didn't work. In order to enable ssl for curl, used the command sudo apt-get install libssl-dev. After that I no longer have the error curl: (1) Protocol "https" not supported...

[gh-TimmyChan]

Replying to jipilab:

After installing autotools and automake I get the following error message:

[normaliz] Installing E-ANTIC...
[normaliz] From https://github.com/mkoeppe/e-antic
[normaliz]  * branch            winfried   -> FETCH_HEAD
[normaliz] HEAD is now at 9de1e6f... renf_elem_class.get_den: New
[normaliz] aclocal: warning: couldn't open directory 'm4': No such file or directory
[normaliz] configure.ac:9: error: possibly undefined macro: AC_PROG_LIBTOOL
[normaliz]       If this token and others are legitimate, please use m4_pattern_allow.
[normaliz]       See the Autoconf documentation.
[normaliz] autoreconf: /usr/bin/autoconf failed with exit status: 1
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

Then, after installing libtool (sage should complain about that not installed?!) then I get:

[normaliz] make[2]: Leaving directory '/home/jplabbe/sage/local/enfnormaliz/E-ANTIC_source/e-antic'
[normaliz] configure: error: cannot find install-sh, install.sh, or shtool in "." "./.." "./../.."
Makefile:2080: recipe for target 'normaliz' failed
make[1]: *** [normaliz] Error 1
make[1]: Leaving directory '/home/jplabbe/sage/build/make'

Installed autotools-dev and automake and libtool and installation completed on Ubuntu 16.04LTS.

[jipilab]

There is a link error in my installation:

sage: import PyQNormaliz_cpp
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-10-4c34a88d49d7> in <module>()
----> 1 import PyQNormaliz_cpp

ImportError: /home/jplabbe/sage/local/enfnormaliz/lib/libantic.so.0: undefined symbol: _fmpq_poly_resultant_div

This can be solved by typing

export LD_LIBRARY_PATH=your_sage_path/local/enfnormaliz/lib/:$LD_LIBRARY_PATH

before starting sage in the terminal...

[gh-sebasguts]

While this solution works for jipilab, it does not for gh-TimmyChan?. Apparently his Ubuntu 16.04 honors the -rpath setting from the PyQNormaliz install script.

The problem is that the -rpath will be set by python itself, and it is not possible to insert anything before that path.

[gh-TimmyChan]

Replying to gh-sebasguts:

While this solution works for jipilab, it does not for gh-TimmyChan?. Apparently his Ubuntu 16.04 honors the -rpath setting from the PyQNormaliz install script.

The problem is that the -rpath will be set by python itself, and it is not possible to insert anything before that path.

Temporarily, import PyQNormaliz_cpp can work by removing 'libflint.so.13 from ~/sage/local/lib/` to some other directory.

[Winfried]

Since version 3.6.0 QNormaliz and PyQNormaliz are included in the Normaliz distribution.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​855a2c0	build/pkgs/normaliz/spkg-install: Only warn if git fetch/pull fails (airplane mode)
​5829bc0	src/sage/geometry/polyhedron/backend_normaliz.py: Add #optional - pyqnormaliz
​de64a4a	build/pkgs/normaliz/dependencies: Depend on gc (dependency of our flint copy)

[git]

Branch pushed to git repo; I updated commit sha1. Last 10 new commits:

​976b93f	Update normaliz to 3.6.3, pynormaliz to 1.19
​e36973a	build/pkgs/normaliz/spkg-install: Update comment
​98d5bd3	Merge remote-tracking branch 'trac/develop' into t/25090/public/pynormaliz1.17_normaliz3.6.1
​aa71e25	Merge remote-tracking branch 'trac/develop' into t/25090/public/pynormaliz1.17_normaliz3.6.1
​7907f7a	Polyhedron_normaliz: Fix: INPUT: items generally do not end with a period
​1e20ef2	Polyhedron_normaliz._make_normaliz_cone: Add doctest
​f3d11a3	Polyhedron_normaliz._init_from_normaliz_data: Add doctest
​36d5473	src/sage/geometry/polyhedron/backend_normaliz.py: Add # optional - pynormaliz
​e852ffa	Merge branch 't/25090/public/pynormaliz1.17_normaliz3.6.1' into t/25097/public/25097/qnormaliz-algebraic
​855391c	normaliz, pynormaliz, pyqnormaliz: Update build scripts to use releases

[mkoeppe]

Updated this branch to use the current released versions of Normaliz from #25090 and PyQNormaliz from pypi.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​e34e10b

Polyhedron_normaliz._nmz_result: Use the PyQNormaliz conversion handlers

[mkoeppe]

Far from "needs review", but needs testing / debugging. (I'm currently getting a segfault when running the testsuite for backend_normaliz.py.)

[Winfried]

Could you give more specific information?

[mkoeppe]

sage: x = polygen(ZZ); P = Polyhedron(vertices=[[sqrt(2)], [AA.polynomial_root(x^3-2, RIF(0,3))]], backend='normaliz', ver
....: bose=True)
# ----8<---- Equivalent QNormaliz input file ---8<----
amb_space 1
subspace 0
vertices 2
 (a^3) 1
 (a^2) 1
number_field min_poly (a^6 - 2) embedding [1.122462048309373 +/- 5.38e-16]
cone 0
# ----8<-------------------8<-------------------8<----
# Calling PyQNormaliz_cpp.NmzCone(**{'subspace': [], 'vertices': [[[[0L, 1L], [0L, 1L], [0L, 1L], [1L, 1L], [0L, 1L], [0L, 1L]], 1L], [[[0L, 1L], [0L, 1L], [1L, 1L], [0L, 1L], [0L, 1L], [0L, 1L]], 1L]], 'number_field': 'min_poly (a^6 - 2) embedding [1.122462048309373 +/- 5.38e-16]', 'cone': []})
------------------------------------------------------------------------
(no backtrace available)
------------------------------------------------------------------------
Unhandled SIGSEGV: A segmentation fault occurred.
This probably occurred because a *compiled* module has a bug
in it and is not properly wrapped with sig_on(), sig_off().
Python will now terminate.
------------------------------------------------------------------------
Segmentation fault: 11

[Winfried]

Normaliz 3.6.3 (and presumably already 3.6.0) computes the polytope without any problems, provided the number_field follows immediately after amb_space. (One could think of relaxing this condition.) Output:

Real embedded number field:
min_poly (a^6 - 2) embedding [1.1225 +/- 4.57e-5]

2 vertices of polyhedron
0 extreme rays of recession cone
2 support hyperplanes of polyhedron (homogenized)

embedding dimension = 2
affine dimension of the polyhedron = 1 (maximal)
rank of recession cone = 0

dehomogenization:
0 1 

***********************************************************************

2 vertices of polyhedron:
 (a^2 ~ 1.2599) 1
 (a^3 ~ 1.4142) 1

0 extreme rays of recession cone:

2 support hyperplanes of polyhedron (homogenized):
 (-1/2*a^3 ~ -0.7071)  1
   (1/2*a^4 ~ 0.7937) -1

[mkoeppe]

Replying to Winfried:

Normaliz 3.6.3 (and presumably already 3.6.0) computes the polytope without any problems,

Thanks for testing. Yes, I think the segfault is coming from somewhere in the interfacing.

provided the number_field follows immediately after amb_space. (One could think of relaxing this condition.)

No, I don't think that's necessary -- I just need to fix the normaliz file writer in my sage code to know about this.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​b6a1665

Polyhedron_normaliz._nmz_result: Fix up NumberfieldElementHandler

[mkoeppe]

x = polygen(ZZ); P = Polyhedron(vertices=[[sqrt(2)], [AA.polynomial_root(x^3-2, RIF(0,3))]], backend='normaliz', verbose=True); P._Hrepresentation
.... segfault ...

Some memory corruption involving the gc, it seems

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ec5f5c9

Polyhedron_normaliz._normaliz_format: Update code that puts 'number_field' right after 'amb_space'

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​9669a6f

Polyhedron_normaliz: If base_ring is already a number field, use it

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​2b149b5	Polyhedron_normaliz._is_zero etc.: Add, copied from Polyhedron_field
​3b454e2	polytopes library: Add tests with backend='normaliz'

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​46dfa08	Add missing build/pkgs/pyqnormaliz/checksums.ini file
​8b4decc	Experimental upgrade of FLINT to 2.5.2+trunk-2018-09-18-e12c83de3e100a113dbc030894d89f67dfe83cd4; remove patches
​2e3fe7d	Merge branch 't/25171/experimental_flint_upgrade_package_tracking_trunk' into t/25097/public/25097/qnormaliz-algebraic
​41f6f98	build/{normaliz,pyqnormaliz}/spkg-install: Use FLINT from #25171, use sage arb, get rid of enfnormaliz/ install hierarchy
​13089b4	build/pkgs/flint/checksums.ini: Update for using github release tarball
​eb6646e	Merge branch 't/25171/experimental_flint_upgrade_package_tracking_trunk' into t/25097/public/25097/qnormaliz-algebraic

[mkoeppe]

Replying to gh-sebasguts:

While this solution works for jipilab, it does not for gh-TimmyChan?. Apparently his Ubuntu 16.04 honors the -rpath setting from the PyQNormaliz install script.

The problem is that the -rpath will be set by python itself, and it is not possible to insert anything before that path.

With the latest commits, I have removed the separate flint installation and this ticket should be easier to build. Please test

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​1ccf632	polytopes.snub_cube: Add keyword arg verbose
​25e64fc	polytopes.snub_dodecahedron: Add keyword arg verbose, mark pyqnormaliz doctest 'not tested - very long time'
​7dfd8a5	polytopes.six_hundred_cell: Fix wrong doctest
​95b5bdd	polytopes.snub_cube: Add keywords exact, base_ring. Add pyqnormaliz doctest

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​1b977dc	Polyhedron_normaliz: Fix a doctest
​4eaec75	Polyhedron_normaliz: Sort keys in verbose output for predictable doctests
​7749ace	Polyhedron_normaliz: Guard against deadly exceptions in *_handler
​a692e85	Polyhedron_QQ_normaliz._init_from_normaliz_data: Add doctests
​3040b2d	Polyhedron_normaliz: Refactor using format_number_field_data, passing normaliz_field through _init_from_normaliz_data
​d56e0d4	Polyhedron_normaliz: Refactor, passing normaliz_field through _init_from_normaliz_cone.
​85166c3	Polyhedron_normaliz: New init arg, normaliz_field. Fix _init_from_normaliz_data doctest
​bb0f73d	Polyhedron_normaliz: Extend doctest
​3d8d078	Polyhedron_normaliz: Switch between Normaliz, QNormaliz based on normaliz_field, rather than class

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ad26e47

PyQNormaliz: Update to 1.2

[mkoeppe]

PyQNormaliz 1.3 is out - ​https://files.pythonhosted.org/packages/f9/3a/4ffbb22390023f7f8331309c51a73063ccee465df307e898456fbcf05356/PyQNormaliz-1.3.tar.gz

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​725ea5b

Merge branch 'trac/develop' into t/25097/public/25097/qnormaliz-algebraic

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​3881007	PyQNormaliz: Update to 1.3
​bd3c990	sage.geometry.polyhedron.constructor: Update error message in doctest

[vdelecroix]

I think we should decouple: upgrading Normaliz and interfacing Normaliz.

[vdelecroix]

I opened #27682

[jipilab]

Replying to vdelecroix:

I think we should decouple: upgrading Normaliz and interfacing Normaliz.

+1

[vdelecroix]

I think it would be good to split this ticket into two tickets:

1. Make backend='normaliz' accepts input being in a common embedded real number field. That is the following should work

   sage: K.<a> = QuadraticField(2, embedding=AA(2)**(1/2))
   sage: Polyhedron([[0,0],[1,a],[0,2*a],[-1,-a]], base_ring=K, backend='normaliz')
   

   And we should also make it so that backend='normaliz' is actually automatically chosen if available

   sage: Polyhedron([[0,0],[1,a],[0,2*a],[-1,-a]], base_ring=K)   # should use normaliz
   

   This would be enough to generate any of the regular polytope examples exactly (including the 600 cells) that could go in the SageMath database polytopes..
2. (follow-up ticket) When the input are elements in AA, or symbolic algebraic, etc, try to figure out how to convert stuff into an embedded number field

[git]

Branch pushed to git repo; I updated commit sha1. Last 10 new commits:

​61e4df9	Fix docstring even more
​9255f2c	pyflakes
​9a06e8e	Merge branch 'public/algnum_to_numberfield' of git://trac.sagemath.org/sage into t/20181/public/algnum_to_numberfield
​f8a2d52	Merge branch 't/20181/public/algnum_to_numberfield' into t/25097/public/25097/qnormaliz-algebraic
​46fcf67	Merge tag '8.8.beta3' into t/25171/experimental_flint_upgrade_package_tracking_trunk
​b320692	Merge branch 't/25171/experimental_flint_upgrade_package_tracking_trunk' into t/25097/public/25097/qnormaliz-algebraic
​76d7db1	Add patch
​c3fd47f	Merge branch 't/25171/experimental_flint_upgrade_package_tracking_trunk' into t/25097/public/25097/qnormaliz-algebraic
​0c93e30	Add FLINT patch install_name_tool_destdir.patch
​2f3c42c	Merge branch 't/25171/experimental_flint_upgrade_package_tracking_trunk' into t/25097/public/25097/qnormaliz-algebraic

[mkoeppe]

I have updated this ticket for 8.8.beta3. Still using the old Normaliz/QNormaliz/PyNormaliz/PyQNormaliz etc., but merged in #20181. 3 doctest failures.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​7c3d64c

Fix up for exact snub cube construction

[jipilab]

Replying to git:

Branch pushed to git repo; I updated commit sha1. New commits:

​7c3d64c

Fix up for exact snub cube construction

Just a small heads-up about #26340 that will conflict with this line.

[mkoeppe]

#27716 has all the backend work that does not depend on the algebraic stuff.

[mkoeppe]

I will next redo this ticket on top of #27716, #26340, #27682, #20181.

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. Last 10 new commits:

​57a9d10	add exact option to snub_cube
​8703f7f	fix snub cube example in polyhedra_tutorial.rst
​7618ea9	package e-antic
​d752708	upgrade normaliz
​4eb02f8	fix doctest in backend_normaliz.py
​56a903e	Merge branches 't/27716/public/normaliz-backend-refactoring', 't/26340/public/26340', 't/27682/27682' and 't/20181/public/algnum_to_numberfield' into t/25097/public/normaliz-3.7.0-algebraic
​915cb46	backend_normaliz: algebraic polyhedra
​3a6e39d	Add library polytope tests for algebraic polytopes
​c7aa2ef	change use of PyQNormaliz to PyNormaliz
​ec695ac	Use _number_field_triple instead of _format_number_field_data

[mkoeppe]

Replying to mkoeppe:

I will next redo this ticket on top of #27716, #26340, #27682, #20181.

Done. Replaced the branch. Many doctests fail. Needs fixing: ​https://github.com/Normaliz/PyNormaliz/issues/37

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​9732605	Fix doctests
​b7c73d1	nfelem_handler: Protect against coordinate vectors with too many elements. Handle unhandled rationals
​a7e3af3	Fixup handling of number_field in normaliz data

[mkoeppe]

All tests work now except for one that needs a fix for this: ​https://github.com/Normaliz/PyNormaliz/issues/41

[mkoeppe]

With Normaliz master (unreleased), all tests pass except for the broken snub cube example from #26340.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​3ce403d	Change optional - pyqnormaliz to pynormaliz
​b370268	pyflakes fixes
​328a04a	_nmz_result: Add doctest
​ee5ff41	Mark newest doctest # optional - pynormaliz
​b60f653	Merge branch 't/27716/public/normaliz-backend-refactoring' into t/25097/public/normaliz-3.7.0-algebraic
​05867b4	Merge tag '8.8.beta3' into t/26340/public/26340
​39c3c67	Fix construction of snub_cube data
​f52802f	Merge branch 't/26340/public/26340' into t/25097/public/normaliz-3.7.0-algebraic
​19bb5d3	Add snub_cube test with normaliz

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​e70d4a5	snub_cube: Change default back to exact=False
​706e796	Merge branch 't/26340/public/26340' into t/25097/public/normaliz-3.7.0-algebraic

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​13d8c44	docstring fix
​e3bbecd	Merge branch 't/26340/public/26340' into t/25097/public/normaliz-3.7.0-algebraic

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​fb459b4	Update e-antic to 0.1.3b0 and normaliz to 3.7.1
​3ad6905	Update pynormaliz to 2.1

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​eaa4584	package e-antic 0.1.3b0
​e5c669d	upgrade normaliz
​14d6c83	Update normaliz to 3.7.1, pynormaliz to 2.1
​64f1f19	fix doctest in backend_normaliz.py
​59e069b	Merge branch 't/27682/27682' into t/25097/public/normaliz-3.7.0-algebraic
​24a8746	Update error message in doctest for new PyNormaliz

[mkoeppe]

Ready for review. Builds and tests OK on macOS Mojave.

[jipilab]

Builds on debian stretch.

$ sage -t sage/src/sage/geometry/polyhedron/*.py --optional=bliss,dochtml,gfortran,lrslib,memlimit,mpir,ninja_build,normaliz,pynormaliz,python2,sage,topcom
too few successful tests, not using stored timings
Git branch: 25097
Using --optional=bliss,dochtml,gfortran,lrslib,memlimit,mpir,ninja_build,normaliz,pynormaliz,python2,sage,topcom
Doctesting 25 files.
----------------------------------------------------------------------
All tests passed!
----------------------------------------------------------------------

[git]

Branch pushed to git repo; I updated commit sha1. Last 10 new commits:

​160dd88	Merge branch 't/27682/27682' into t/25091/public/some_normaliz_features
​78c2de4	Merge branch 't/27716/public/normaliz-backend-refactoring' into t/25091/public/some_normaliz_features
​41bd968	Fix doctest error message - NormalizError
​7662077	Fixup merge
​885ef9e	Use EhrhartQuasiPolynomial
​313fee1	Fix doctest of hilbert_series
​7905903	Pass grading as a matrix
​9249dda	Use _nmz_result instead of calling NmzResult directly
​ec96615	Merge branch 't/25091/public/some_normaliz_features' into t/25097/public/normaliz-3.7.0-algebraic
​f232c6c	_nmz_result already constructs rationals

[mkoeppe]

Merged #25091, works fine after minor changes.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​5d92502	Add checks for real embedded base rings, improve error messages
​6e2991b	backend_normaliz: Prepare refactoring of V, H data conversion
​812395c	Refactor using Polyhedron_normaliz._compute_nmz_data_lists_and_field

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​3d949a2	Upgrade normaliz to 3.7.2, pynormaliz to 2.2
​c1169c3	Merge branch 't/27731/public/27731' into t/25097/public/normaliz-3.7.0-algebraic
​8aee20f	snub_dodecahedron: Enable doctest
​1dd5b3c	snub_dodecahedron: Update tests, doc string

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ac45511

Fix for tuples in H and V representations

[jipilab]

An error in convex hull?

[mkoeppe]

That error was caused by using an old PyNormaliz?.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​b488911	Add build/pkgs/pynormaliz/spkg-check
​5d01388	Upgrade pynormaliz to 2.4
​2c3ea3c	Merge branch 't/27731/public/27731' into t/25097/public/normaliz-3.7.0-algebraic

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​a4e5977	Upgrade pynormaliz to 2.5
​59f320b	Merge branch 't/27731/public/27731' into t/25097/public/normaliz-3.7.0-algebraic

[git]

Branch pushed to git repo; I updated commit sha1. Last 10 new commits:

​59cc7e0	Fix Error messages
​a4f01d4	Fixed typo and added require package
​fbd3dce	removed is_package_installed
​8ceee84	Change TypeError for RuntimeError
​0da067f	Fixed doctests and added author name
​d7122d7	pyflakes
​63a80b6	Forgot an import
​d8716e3	Refactor NmzResult
​08b315b	Fixed infinite recursion
​ec2646d	Merge branch 't/25091/public/some_normaliz_features' into t/25097/public/normaliz-3.7.0-algebraic

[jipilab]

Here is an error happening while coercing the values back into the given base ring (UCF in this case):

sage: W = CoxeterGroup(['I',8])
sage: n = W.one().canonical_matrix().rank()
sage: weights = W.fundamental_weights()
sage: point = [ZZ.one()] * n
sage: apex = sum(point[i-1] * weights[i] for i in weights.keys())
sage: non_zero_index = list(apex).index(filter(lambda x: x!=0, apex)[0])
sage: apex = (QQ(1)/apex[non_zero_index]) * apex
sage: apex.set_immutable()
sage: vertices = set()
sage: br = apex.base_ring()
sage: for w in W:
....:     new_point = w * apex
....:     new_point.set_immutable()
....:     vertices.add(new_point)
....:     
sage: Polyhedron(vertices=vertices, backend='normaliz', base_ring=br,verbose=True)
Traceback (most recent call last):
...
TypeError: -1 of type <type 'sage.rings.number_field.number_field_element.NumberFieldElement_absolute'> not valid to initialize an element of the universal cyclotomic field

[mkoeppe]

Unfortunately both of the following conversions fail:

UniversalCyclotomicField()(AA(-1))
TypeError: -1 of type <class 'sage.rings.qqbar.AlgebraicReal'> not valid to initialize an element of the universal cyclotomic field

and

sage: K.<sqrt2> = QuadraticField(2)
sage: UniversalCyclotomicField()(K(-1))
TypeError: -1 of type <type 'sage.rings.number_field.number_field_element_quadratic.NumberFieldElement_quadratic'> not valid to initialize an element of the universal cyclotomic field

[mkoeppe]

This should be a separate ticket

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​292e01f	backend_normaliz: py3 fix
​d44a00d	backend_normaliz: Canonicalize sort order of fields in normaliz files
​dd9a81b	Merge tag '8.8.beta4' into t/27731/public/27731
​9b3a5ba	Fix error message in doctest for new pynormaliz version
​0a8c6f8	Merge branch 't/27782/1_pynormaliz_doctest_failing_in_src_sage_geometry_polyhedron_backend_normaliz_py' into t/27731/public/27731
​978765b	Merge branch 't/27731/public/27731' into t/25097/public/normaliz-3.7.0-algebraic

[jipilab]

There are python3 issues, pyflakes conventions issues, ::: issues, and 2 missing doctests in backend_normaliz.py, see patchbot for more details.

From a recent bot, there are the following files that have failing doctests (seems like the optional hashtag is missing):

sage -t --long src/sage/geometry/polyhedron/backend_normaliz.py  # 11 doctests failed
sage -t --long src/doc/en/thematic_tutorials/geometry/polyhedra_tutorial.rst  # 2 doctests failed

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​da05706	Fix forgotten optional package test
​ec949bd	Merge branch 't/25091/public/some_normaliz_features' into t/25097/public/normaliz-3.7.0-algebraic
​a8f307e	Merge tag '8.8.beta5' into t/25097/public/normaliz-3.7.0-algebraic

[mkoeppe]

Merged in latest version of #25091 and current beta.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ca7c25c	sage/geometry/polyhedron/parent.py: Change wording of error message
​7a0886a	Fix doctest
​ed6fcbf	Update polyhedron tutorial

[mkoeppe]

Setting it back to needs_review to get a fresh patchbot run.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​4bd7635

Add more optional - pynormaliz

[mkoeppe]

Replying to jipilab:

Here is an error happening while coercing the values back into the given base ring (UCF in this case):

Should we actually allow a complex field such as UCF as the base ring of a polyhedron?

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​0ba5515	Polyhedron_normaliz._convert_to_pynormaliz: Rename from _convert_to_Qnormaliz, make static, add doctest, use int() instead of long()
​b484af7	Update doctests for int/long, py3

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​99e12c1

Polyhedron_normaliz._number_field_triple: Add doctest

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​4d82f5f

_number_field_elements_from_algebraics_list_of_lists_of_lists: Remove debugging code

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ef280d1	Fix triple_colon
​d03c531	py3: Change iteritems to items
​fcf2c35	py3: Get rid of has_key; avoid unnecessary copy

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​abb7b02

More optional - pynormaliz

[jipilab]

Replying to mkoeppe:

Replying to jipilab:

Here is an error happening while coercing the values back into the given base ring (UCF in this case):

Should we actually allow a complex field such as UCF as the base ring of a polyhedron?

Not sure. I remember I was not so happy about allowing real algebraic numbers that have a complex parent into polyhedron: it makes things quite intricate in designing the constructor.

The situation before this ticket was not allowing anything inside of SymbolicRing for example, and this would also do the same for UCF is the backend field was not given (or perhaps there was an effort to check if one can coerce it in AA)... Now, since we accept things that are in the symbolic ring if it is possible to coerce it into an embedded number field... I feel like it would be nice to accept elements of UCF if they are real... I remember I had to tweak #20181 so that it would manage to coerce real algebraic numbers even though they are elements of a complex ring.

That said, I would say that the answer should be 'No' (as far as I understand, complex polyhedra are not really well-defined, or are they?). So, this would be an exception to the idea "give back a common parent to the elements given by the user and use it as base ring". I guess the rule should be bound to ultimately "stay real".

So there could be a check to see if the user provided elements from a complex ring (although they are indeed real) and do the appropriate steps to give a real base ring. I do not know how hard implementing this here.

[jipilab]

There are two doctest failing in the tutorials (coming from changing an error message):

----------------------------------------------------------------------
sage -t --long src/sage/sat/boolean_polynomials.py  # 1 doctest failed
sage -t --long src/sage/graphs/generic_graph.py  # 1 doctest failed
sage -t --long src/doc/en/thematic_tutorials/geometry/polyhedra_tutorial.rst  # 2 doctests failed
----------------------------------------------------------------------

The first two might be spurious errors, but maybe they come from employing polyhedra...

[jipilab]

Replying to jipilab:

There are two doctest failing in the tutorials (coming from changing an error message):

----------------------------------------------------------------------
sage -t --long src/sage/sat/boolean_polynomials.py  # 1 doctest failed
sage -t --long src/sage/graphs/generic_graph.py  # 1 doctest failed
sage -t --long src/doc/en/thematic_tutorials/geometry/polyhedra_tutorial.rst  # 2 doctests failed
----------------------------------------------------------------------

The first two might be spurious errors, but maybe they come from employing polyhedra...

Actually, there is some dirt in the traceback for the errors on the branch. Most probably this is the cause of the breaking doctests...

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​ad4d739	Merge tag '8.8.beta6' into t/25097/public/normaliz-3.7.0-algebraic
​ca71cc2	polyhedra_tutorial: Fix error messages in doctests

[mkoeppe]

Fixed the doctest and merged in the newest beta.

[jipilab]

There is 1 doctest failing because of #21161. :-/

[jipilab]

Now with #27965, we may get a conflict in snub dodecahedron. I was not quick enough to warn about it on the ticket #27965 about the current one...

[jipilab]

#27965 should be merged into this ticket to avoid yet another conflict.

The other ticket allows to "whitelist" geometry for py3, probably asap for it to be in the next release.

... that said, I would also like to see this ticket in the next release so that we can move on to further enrich normaliz' interfacing...

[embray]

Tickets still needing working or clarification should be moved to the next release milestone at the soonest (please feel free to revert if you think the ticket is close to being resolved).

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​28b8dbd	Merge tag '8.8.rc1' into t/25097/public/normaliz-3.7.0-algebraic
​6fcdf05	trac 27965: py3 fix in geometry/polyhedron/library.py
​3d46bb2	Merge branch 't/27965/27965' into t/25097/public/normaliz-3.7.0-algebraic
​4258704	Fix doctests for changed embedded numberfield repr

[jipilab]

Looks good to me now. I would say this is ready and I would put it as positive review, with Vincent also as a reviewer too.

@Vincent: Let us know if you agree. It would be good to know if there are last things to be aware of.

[jipilab]

ping!

Any last comments Vincent?

[chapoton]

doctest failure in /library with py3

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​23b1cbe	Merge tag '8.9.beta1' into t/25097/public/normaliz-3.7.0-algebraic
​f94bf37	snub_dodecahedron: Disable doctests for broken floating point computation

[chapoton]

The last changes looks reasonable to me.

[jipilab]

It looks good to me too. I set it to positive review.

Just a FYI: if there are any other comments, next week is the Sage days in Bonn and vdelecroix and me are going to be available to work on follow-ups if need be.