py3: change the repr of Finite families

[chapoton]

by pprinting their dict (sort by key)

[chapoton]

New commits:

​ab5fe56

py3: change the repr of Finite family (pprint their dict)

[git]

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

​d6f7560

fixing some doctests

[tscrim]

General +1 to this change, but see comment 4 of #24438. In particular, having a _keys means the iteration order for FiniteFamily is fixed, and I think the repr should reflect that.

[git]

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

​5fd1450

trac 26225 better _repr_ for families, using _keys if not None

[git]

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

​8fd4354

trac 26225 fixing doctests

[embray]

src/sage/sets/family.py

@@ -35 +35 @@
 #*****************************************************************************
 import types
 from copy import copy
+from pprint import pformat, saferepr
 
 from six import itervalues
 from six.moves import range
@@ -659 +660 @@
             sage: from sage.sets.family import FiniteFamily
             sage: FiniteFamily({3: 'a'}) # indirect doctest
             Finite family {3: 'a'}
+
+            sage: FiniteFamily({3: 'a', 4: 'b'}) # indirect doctest
+            Finite family {3: 'a', 4: 'b'}
+
+            sage: FiniteFamily({3: 'a', 4: 'b'}, keys=[4,3]) # indirect doctest
+            Finite family {4: 'b', 3: 'a'}
         """
-        return "Finite family %s"%self._dictionary
+        if self._keys is None:
+            d = ' '.join(pformat(self._dictionary).splitlines())
+            return "Finite family %s" % d
+
+        d = ', '.join('{}: {}'.format(saferepr(key),
+                                      saferepr(self._dictionary[key]))
+                      for key in self._keys)
+        return 'Finite family {{{}}}'.format(d)

You could avoid some duplication (and inconsistency in string formatting methods) by writing this like:

if self._keys:
    d = ', '.join('{}: {}'.format(saferepr(key),
                                  saferepr(self._dictionary[key]))
                  for key in self._keys)
else:
    d = ' '.join(pformat(self._dictionary)[1:-1].splitlines())

return 'Finite family {{{}}}'.format(d)

or similar.

Travis, does this address the problems with #24438? I haven't checked carefully. Doesn't this version mean that elements not in the keys argument won't be displayed at all? Is that desirable?

[tscrim]

My understanding is this ticket would be another way to fix #24438, but a nuance may be lost on me. It should be that keys should match the keys of the dict object used in the construction (well, more generally, keys is treated as the input to a function -- equivalently, a sequence indexed by keys -- so things not in keys should not be displayed).

[embray]

Ok. Other than my above code style suggestion this is better then.

[git]

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

​10a2126

coding style

[chapoton]

There is an annoying doctest failure in

src/sage/algebras/lie_algebras/classical_lie_algebra.py

which is not clear to me. The order seems to change from one machine to another..

[tscrim]

So it has to do with the fact that roots and coroots are incomparable, hence sorted is random but the underlying dict is small enough that it doesn't visibly suffer from this. What we should do is just add a fixed set of keys. So for str_keys, it would be

keys = (['e{}'.format(i) for i in index_set] + ['f{}'.format(i) for i in index_set]
        + ['h{}'.format(i) for i in index_set])

and the other case would be

keys = ([alpha[i] for i in index_set] + [-alpha[i] for i in index_set]
        + [alphacheck[i] for i in index_set])

and then

Family(keys, ret.__getitem__)

[git]

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

​239f345

py3: change the repr of Finite family (pprint their dict)

[chapoton]

thanks, done. I am now launching my patchbots.

[chapoton]

ok, both bots are morally green

[tscrim]

Can you check that this change does not affect #26078? If it does, could you make #26078 a dependency?

Otherwise LGTM (I don't know of any other tickets that would otherwise have a conflict with this change).

[chapoton]

It should not interfere, as here we only change the output of "Finite family", not the output of "Family". And there is no "Finite Family" in #26078.

[tscrim]

Ah, right, they are all instances of TrivialFamily. Thanks.