py3: adding .items iterator to vectors

[fchapoton]

to help build doc in python3

CC: @embray @jdemeyer

Component: python3

Author: Frédéric Chapoton

Branch/Commit: 2fd6154

Reviewer: Erik Bray

Issue created by migration from https://trac.sagemath.org/ticket/25381

[fchapoton]

New commits:

cbd1888

adding .items method to vectors for py3 sake

[fchapoton]

Commit: cbd1888

[fchapoton]

Branch: u/chapoton/25381

[embray]

comment:2

+1 in principle.

I have a different approach to this in my python3 branch that I haven't made a ticket for yet. What do you think?

diff --git a/src/sage/modules/free_module_element.pyx b/src/sage/modules/free_module_element.pyx
index 29b2de1..814afa7 100644
--- a/src/sage/modules/free_module_element.pyx
+++ b/src/sage/modules/free_module_element.pyx
@@ -821,6 +821,7 @@ def random_vector(ring, degree=None, *args, **kwds):
     entries = [ring.random_element(*args, **kwds) for _ in range(degree)]
     return vector(ring, degree, entries, sparse)

+
 cdef class FreeModuleElement(Vector):   # abstract base class
     """
     An element of a generic free module.
@@ -1545,6 +1546,28 @@ cdef class FreeModuleElement(Vector):   # abstract base class
         from sage.arith.all import lcm
         return lcm(v)

+    def items(self):
+        """
+        Return an iterable over self.
+
+        On Python 2 this returns the same as ``list(self.iteritems())``, while
+        on Python 3 this is an alias for ``self.items()``.  This is for
+        congruence with ``dict.items()``.
+
+        EXAMPLES::
+
+            sage: v = vector([1, 2/3, pi])
+            sage: v.items()  # py2
+            [(0, 1), (1, 2/3), (2, pi)]
+            sage: v.items()  # py3
+            <generator object at ...>
+        """
+
+        IF PY_MAJOR_VERSION < 3:
+            return list(self.iteritems())
+        ELSE:
+            return self.iteritems()
+
     def iteritems(self):
         """
         Return iterator over self.
@@ -4543,7 +4566,7 @@ cdef class FreeModuleElement_generic_sparse(FreeModuleElement):
                 e = entries
                 entries = {}
                 try:
-                    for k, x in e.iteritems():
+                    for k, x in e.items():
                         x = coefficient_ring(x)
                         if x:
                             entries[k] = x
@@ -4734,11 +4757,11 @@ cdef class FreeModuleElement_generic_sparse(FreeModuleElement):

             sage: v = vector([1,2/3,pi], sparse=True)
             sage: v.iteritems()
-            <dictionary-itemiterator object at ...>
+            <dict...itemiterator object at ...>
             sage: list(v.iteritems())
             [(0, 1), (1, 2/3), (2, pi)]
         """
-        return self._entries.iteritems()
+        return iter(self._entries.iteritems())

     def __reduce__(self):
         """

[fchapoton]

comment:3

oh, well. Given that "items" currently does not exist, one can just provide it directly with the python3 behaviour. Otherwise, this opens the possibility that someone will use that "items" in a python3-incompatible way, and that we will have to fix this bad use later..

[embray]

comment:4

That is true--the question is to be compatible with Python 2 semantics on Python 2, or just be immediately forward-compatible with Python 3. Since FreeModuleElement is not actually a dict subclass I think we have some freedom to pick the latter.

In that case though what you might do instead is actually rename the iteritems() method to just items() and make iteritems() and deprecated (perhaps just informally?) alias for items().

[sagetrac-git]

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

2fd6154

trac 25381 iteritems as an alias of items for vectors

[sagetrac-git]

Changed commit from cbd1888 to 2fd6154

[fchapoton]

comment:6

Done.

I do not think that one should deprecate formally anything.

[embray]

comment:7

I'm inclined to agree in this case, though I can't quite articulate why.

[embray]

Reviewer: Erik Bray

[vbraun]

Changed branch from u/chapoton/25381 to 2fd6154