Fix bug due to a call to SSLContext() in src/sage/graphs/isgci.py

[dcoudert]

sage: graph_classes._download_db()
---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-1-3914f28ce325> in <module>
----> 1 graph_classes._download_db()

/home/dcoudert/sage/local/var/lib/sage/venv-python3.10/lib64/python3.10/site-packages/sage/graphs/isgci.py in _download_db(self)
    829         import tempfile
    830         u = urlopen('https://www.graphclasses.org/data.zip',
--> 831                     context=SSLContext())
    832         with tempfile.NamedTemporaryFile(suffix=".zip") as f:
    833             f.write(u.read())

NameError: name 'SSLContext' is not defined

This bug was fixed in #33771, replacing the call to SSLContext() in src/sage/graphs/isgci.py with a call to default_context(), but this changes has been reverted in #33829.

---

Dependency on #34079 since it modifies this file.

CC: @fchapoton @dimpase @seblabbe

Component: graph theory

Author: David Coudert

Branch: 2302af6

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/34211

[dcoudert]

Branch: public/graphs/34211_bug_in_isgci

[dcoudert]

Commit: 32f05b0

[dcoudert]

New commits:

b6db584	trac #34079: pycodestyle cleanup
b91db9e	trac #34079: Merged with 9.7.beta5
b55efb1	trac #34211: replace call to SSLContext with default_context
32f05b0	trac #34211: some care

[dcoudert]

comment:2

On the way, I found a new bug: If I do graph_classes._download_db() in a sage console, and then ./sage -t src/sage/graphs/isgci.py, I get errors like:

sage -t --warn-long 77.5 --random-seed=217922271289552033231623053257548326399 src/sage/graphs/isgci.py
**********************************************************************
File "src/sage/graphs/isgci.py", line 53, in sage.graphs.isgci
Failed example:
    Chordal.description()
Expected:
    Class of graphs : Chordal
    -------------------------
    id                             :  gc_32
    name                           :  chordal
    type                           :  base
    <BLANKLINE>
    Problems :
    -----------
    3-Colourability                :  Linear
    Clique                         :  Polynomial
    Clique cover                   :  Polynomial
    Cliquewidth                    :  Unbounded
    Cliquewidth expression         :  NP-complete
    Colourability                  :  Linear
    Cutwidth                       :  NP-complete
    Domination                     :  NP-complete
    Feedback vertex set            :  Polynomial
    Hamiltonian cycle              :  NP-complete
    Hamiltonian path               :  NP-complete
    Independent set                :  Linear
    Maximum bisection              :  Unknown
    Maximum cut                    :  NP-complete
    Minimum bisection              :  Unknown
    Recognition                    :  Linear
    Treewidth                      :  Polynomial
    Weighted clique                :  Polynomial
    Weighted feedback vertex set   :  Unknown
    Weighted independent set       :  Linear
Got:
    Class of graphs : Chordal
    -------------------------
    id                             : gc_32
    name                           : chordal
    parameter                      : [{'name': 'acyclic chromatic number', 'boundedness': 'Unbounded'}, {'name': 'bandwidth', 'boundedness': 'Unbounded'}, {'name': 'book thickness', 'boundedness': 'Unbounded'}, {'name': 'booleanwidth', 'boundedness': 'Unbounded'}, {'name': 'branchwidth', 'boundedness': 'Unbounded'}, {'name': 'carvingwidth', 'boundedness': 'Unbounded'}, {'name': 'chromatic number', 'boundedness': 'Unbounded'}, {'name': 'cliquewidth', 'boundedness': 'Unbounded'}, {'name': 'cochromatic number', 'boundedness': 'Unknown'}, {'name': 'cutwidth', 'boundedness': 'Unbounded'}, {'name': 'degeneracy', 'boundedness': 'Unbounded'}, {'name': 'diameter', 'boundedness': 'Unbounded'}, {'name': 'distance to block', 'boundedness': 'Unbounded'}, {'name': 'distance to clique', 'boundedness': 'Unbounded'}, {'name': 'distance to cluster', 'boundedness': 'Unbounded'}, {'name': 'distance to co-cluster', 'boundedness': 'Unbounded'}, {'name': 'distance to cograph', 'boundedness': 'Unbounded'}, {'name': 'distance to linear forest', 'boundedness': 'Unbounded'}, {'name': 'distance to outerplanar', 'boundedness': 'Unbounded'}, {'name': 'genus', 'boundedness': 'Unbounded'}, {'name': 'max-leaf number', 'boundedness': 'Unbounded'}, {'name': 'maximum clique', 'boundedness': 'Unbounded'}, {'name': 'maximum degree', 'boundedness': 'Unbounded'}, {'name': 'maximum independent set', 'boundedness': 'Unbounded'}, {'name': 'maximum induced matching', 'boundedness': 'Unbounded'}, {'name': 'maximum matching', 'boundedness': 'Unbounded'}, {'name': 'minimum clique cover', 'boundedness': 'Unbounded'}, {'name': 'minimum dominating set', 'boundedness': 'Unbounded'}, {'name': 'pathwidth', 'boundedness': 'Unbounded'}, {'name': 'rankwidth', 'boundedness': 'Unbounded'}, {'name': 'tree depth', 'boundedness': 'Unbounded'}, {'name': 'treewidth', 'boundedness': 'Unbounded'}, {'name': 'vertex cover', 'boundedness': 'Unbounded'}]
    speed                          : {'speed': 'Superfactorial-Theta'}
    type                           : base
    <BLANKLINE>
    Problems :
    -----------
    3-Colourability                : Linear
    Clique                         : Polynomial
    Clique cover                   : Polynomial
    Cochromatic colouring          : Unknown
    Colourability                  : Linear
    Domination                     : NP-complete
    Feedback vertex set            : Polynomial
    Graph isomorphism              : GI-complete
    Hamiltonian cycle              : NP-complete
    Hamiltonian path               : NP-complete
    Independent dominating set     : Linear
    Independent set                : Linear
    Maximum bisection              : Unknown
    Maximum cut                    : NP-complete
    Minimum bisection              : Unknown
    Monopolarity                   : Linear
    Polarity                       : Polynomial
    Recognition                    : Linear
    Weighted clique                : Polynomial
    Weighted feedback vertex set   : Unknown
    Weighted independent dominating set : Unknown
    Weighted independent set       : Linear
    Weighted maximum cut           : NP-complete
    book thickness decomposition   : Unknown
    booleanwidth decomposition     : Unknown
    cliquewidth decomposition      : Unknown
    cutwidth decomposition         : NP-complete
    treewidth decomposition        : Polynomial

[dcoudert]

comment:3

Do you know how to clear the cache / database ?

[dcoudert]

comment:4

It seems that we have to update file local/share/graphs/isgci_sage.xml (don't know how), update doctests accordingly, and decide what to do with parameter and speed in the description of the class (parameters added in 2015 and 2018, details can be found at https://www.graphclasses.org/classes/gc_61.html).

[dimpase]

comment:6

isn't it updateable with graph_classes.update_db() ?

[dcoudert]

comment:7

Yes it is, but I'm not sure if I should push an updated version of the database + corrected doctests in this ticket, or if there is another way to store a version of the database. I don't know where the default file that we have in local/share/graphs/isgci_sage.xml comes from and what's the normal procedure to update it.

[dimpase]

comment:8

if you read the code you'll see it's downloaded, it's not stored anywhere in Sage git tree.

[dimpase]

comment:9

many of these doctests are too verbose, replacing most of them with ... seems OK.

[sagetrac-git]

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

9ec96bc	trac #34211: Merged with 9.7.beta7
0a33cf3	trac #34211: reduce verbosity of some doctests

[sagetrac-git]

Changed commit from 32f05b0 to 0a33cf3

[dcoudert]

comment:11

I have reduced the verbosity of some doctests.

[dcoudert]

comment:12

and we have green bot.

[dimpase]

Reviewer: Dima Pasechnik

[dimpase]

comment:13

lgtm

[dimpase]

comment:14

should we actually turn these untested tests into optional (internet) tests? This would improve chances to timely catch upstream changes.

[dcoudert]

comment:16

Should I just changed to # optional - internet or is it better to add # long time # optional - internet ?

[dimpase]

comment:17

IMHO # optional - internet is good enough.

[sagetrac-git]

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

310cc04	trac #34211: Merged with 9.7.beta8
2302af6	trac #34211: indicate optional doctests

[sagetrac-git]

Changed commit from 0a33cf3 to 2302af6

[dcoudert]

comment:19

done

[dcoudert]

Changed dependencies from #34079 to none

[dimpase]

comment:20

ok

[vbraun]

Changed branch from public/graphs/34211_bug_in_isgci to 2302af6

[seblabbe]

comment:22

see #34612 for a follow-up ticket fixing a doctest failure

[seblabbe]

Changed commit from 2302af6 to none