py3: Stop to use sort on KeyConvertingDict object

[vklein]

With python3 we currently have 40 failures of type

TypeError: '<' not supported between instances of 'KeyConvertingDict' and 'KeyConvertingDict'

These errors comes from the call of V.sort() in multi_polynomial_ideal.py", line 2381, variety function.

The root reason for calling sort is for doctest displaying :

commit 17f42824d8941b23825fea1a6d82278bd5ef6c74 Author: Carl Witty <cwitty@…> Date: Wed Oct 24 18:36:37 2007 -0700

Sort results of variety(), for better cross-architecture doctest compatibility. This compensates for the fact that the results of Singular's triangular decomposition algorithm triangLfak appear in a different order on different architectures. (It would be better to fix this in triangular_decomposition(), but the obvious approach of sorting its result failed due to TRAC #991.)

As it will not make a lot of sense to implement comparison operators for KeyConvertingDict? this ticket aim at solving the sorting problems at doctests level rather than at python source code level.

[git]

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

​4b8a3e5

Trac #26781: replace dict.iteritems calls by dict.items

[vklein]

Add #23572 as dependency for mpoly renaming.

[git]

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

​b32170f	reviewer suggestion about independent set test
​324b483	Merge branch 'public/french_book_python3' in 8.6.b0
​a00c64c	livre sage doctests: fix floating errors
​3c6139a	Merge branch 'public/french_book_python3' in 8.6.b1
​6d6fb70	partial fix for numerical noise on osX
​17d97d8	Merge branch 'public/french_book_python3' in 8.6
​6da3629	book doctests : adding some tolerance again
​5fd7867	Trac #26792: py3: Stop to use sort on KeyConvertingDict object
​140623f	Trac #26792: py3: fix test.french_book_mpoly module
​b78d1c4	Trac #26792: replace dict.iteritems calls by dict.items

[git]

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

​822e4de	Trac #26792: py3: Stop to use sort on KeyConvertingDict object
​ceee97f	Trac #26792: replace dict.iteritems calls by dict.items

[vklein]

Ticket rebased on 8.7.beta1.

[git]

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

​3c0660c

Trac #26792: fix mpoly.py doctests

[vklein]

Fix mpoly.py doctests and remove #23572 dependency.

[git]

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

​9f3984b

Trac #26792: fix mpoly.py doctests

[git]

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

​f5ce2fa

Trac #26792: fix mpoly.py doctests

[vklein]

ticket on hold in order to let priority for #23572

[chapoton]

time to start again with this ticket ; this should be a great step forward when done

[chapoton]

New commits:

​2f908f0

Merge branch 'u/vklein/26792' in 8.7.b4

[chapoton]

remains one failing doctest

[git]

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

​f4a2bfa

trac 26792 fixing one doctest

[jhpalmieri]

I get a merge conflict with 8.7.beta4: src/sage/tests/books/computational-mathematics-with-sagemath/mpoly_doctest.py

[jhpalmieri]

There is a related doctest failure in misc/converting_dict.py. Four Python 3 failures in the develop branch are replaced by 1 with this branch.

[git]

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

​945857f

trac 26792 trying to fix a doctest

[jhpalmieri]

The new test passes with py3 but not py2.

[chapoton]

and before my last commit it was the contrary. Probably one should make sure that the repr of these "dicts" is conveniently sorted in the same way for both py2 and py3.

[vklein]

I will fix that.

[git]

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

​59c980e

Trac #26792: Fix doctest in converting_dict.py

[vklein]

Replying to chapoton:

and before my last commit it was the contrary. Probably one should make sure that the repr of these "dicts" is conveniently sorted in the same way for both py2 and py3.

The __repr__ called by KeyConvertingDict is the one inherited from python's dict i don't think we should override it just for doctests compatibility. I prefer to modify the doctests.

[chapoton]

I was talking about the ipython representation. This is different (already for dict) from using str or repr. Using that in the repr would be better.

[git]

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

​4dc152a

Trac #26792: Fix a doctest in multi_polynomial_ideal.py

[vklein]

Replying to chapoton:

I was talking about the ipython representation. This is different (already for dict) from using str or repr. Using that in the repr would be better.

I think you mean using Ipython pretty print. I don't know why it like it's not called by default for KeyConvertingDict as it is for dict:

Adding some traces :

diff --git a/src/sage/misc/converting_dict.py b/src/sage/misc/converting_dict.py
index f9bdb17..f08e13f 100644
--- a/src/sage/misc/converting_dict.py
+++ b/src/sage/misc/converting_dict.py
@@ -140,6 +140,13 @@ class KeyConvertingDict(dict):
         key = self.key_conversion_function(key)
         return super(KeyConvertingDict, self).__setitem__(key, value)
 
+    def __repr__(self):
+        print("call __repr__")
+        return super(KeyConvertingDict, self).__repr__()
+
+    def _repr_pretty_(self, p, cycle):
+        print("pretty")
+
     def __delitem__(self, key):
         r"""
         Remove a mapping from the dictionary.

You get this:

sage: K.<x,y> = QQ[]
sage: I = ideal([x^2+2*y-5,x+y+3])
sage: v = I.variety(AA)[0]; v
call __repr__
{x: 4.464101615137755?, y: -7.464101615137755?}
sage: from IPython.lib.pretty import pprint
sage: pprint(v)
pretty

[git]

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

​c9c1047

Trac #26792: py3 fix doctests failing because ...

[vklein]

I fixed the remaining cases using #py2 #py3 flags as it seems to be the most readable solution right now.

There is still one failling test sort related to KeyConvertingDict:

File "src/sage/tests/books/computational-mathematics-with-sagemath/mpoly_doctest.py", line 409, in sage.tests.books.computational-mathematics-with-sagemath.mpoly_doctest
Failed example:
    J.variety(RealField(51))
Expected:
    [{y: 396340.890166545, x: -14.1660266425312}]
Got:
    [{y: 396340.890166545, x: -35.0231212211684}]

The difference being the value of x i think this is out the scope of this ticket.

[chapoton]

ok, let us move on here.