Incorrect caching of variables' latex_name attribute

[sagetrac-wonder]

Saving and loading variables apparently doesn't reload the latex_name attribute:

┌────────────────────────────────────────────────────────────────────┐
│ Sage Version 6.4.1, Release Date: 2014-11-23                       │
│ Type "notebook()" for the browser-based notebook interface.        │
│ Type "help()" for help.                                            │
└────────────────────────────────────────────────────────────────────┘
sage: xhat = SR.symbol( 'xhat', latex_name='\hat{x}' )
sage: latex(xhat)
\hat{x}
sage: save_session( 'xhat' )
sage: l = load( 'xhat.sobj' )
sage: latex( l['xhat'] )
\hat{x}
sage: 
Exiting Sage (CPU time 0m0.20s, Wall time 2m49.25s).

┌────────────────────────────────────────────────────────────────────┐
│ Sage Version 6.4.1, Release Date: 2014-11-23                       │
│ Type "notebook()" for the browser-based notebook interface.        │
│ Type "help()" for help.                                            │
└────────────────────────────────────────────────────────────────────┘
sage: l = load( 'xhat.sobj' )
sage: latex( l['xhat'] )
\mathit{xhat}

It looks as if it tries to keep these objects unique, and load() will leave an existing variable's latex_name in place, but won't set it when creating a variable.

I can see arguments for and against considering SR.symbol('x', latex_name='y') a different variable from SR.symbol('x'). It seems reasonable to consider them the same, in which case the latex name should be discarded if there's already a variable of the same name. But it should use the saved latex name when loading a variable that isn't already defined.

Seems possibly related: #5466

Update: I think this report, posted to sage-support today, is the same issue:

If you do something like this

sage: a = var('a',latex_name='BLAH')

and then
sage: a = var('a') 
sage: latex(a) 
BLAH 
So the first a=var(...) did have an impact on the second one, no?

Component: symbolics

Issue created by migration from https://trac.sagemath.org/ticket/17559

[sagetrac-wonder]

Description changed:

--- 
+++ 
@@ -27,10 +27,23 @@
 sage: latex( l['xhat'] )
 \mathit{xhat}
 ```
-
 It looks as if it tries to keep these objects unique, and `load()` will leave an existing variable's `latex_name` in place, but won't set it when creating a variable.
 
 I can see arguments for and against considering `SR.symbol('x', latex_name='y')` a different variable from `SR.symbol('x')`.  It seems reasonable to consider them the same, in which case the latex name should be discarded if there's already a variable of the same name.  But it should use the saved latex name when loading a variable that isn't already defined.
 
 Seems possibly related: #5466
 
+Update: I think this report, [posted to sage-support today](https://groups.google.com/d/msg/sage-support/pZ4rCvBYtTY/b_KTxC-6BIoJ), is the same issue:
+
+> If you do something like this 
+> 
+> sage: a = var('a',latex_name='BLAH') 
+> 
+> and then 
+
+> sage: a = var('a') 
+> > sage: latex(a) 
+> 
+> BLAH 
+> 
+> So the first a=var(...) did have an impact on the second one, no? 

[sagetrac-wonder]

Description changed:

--- 
+++ 
@@ -35,15 +35,12 @@
 
 Update: I think this report, [posted to sage-support today](https://groups.google.com/d/msg/sage-support/pZ4rCvBYtTY/b_KTxC-6BIoJ), is the same issue:
 
-> If you do something like this 
+> If you do something like this 
 > 
-> sage: a = var('a',latex_name='BLAH') 
+> sage: a = var('a',latex_name='BLAH') 
 > 
-> and then 
-
+> and then
 > sage: a = var('a') 
-> > sage: latex(a) 
-> 
+> sage: latex(a) 
 > BLAH 
-> 
-> So the first a=var(...) did have an impact on the second one, no? 
+> So the first a=var(...) did have an impact on the second one, no?

[kcrisman]

comment:5

According to the sage-support thread in question, perhaps the more recent bug is different

The problem above is caused by sage reusing 
"already existing variables" in the symbol function 
in sage/symbolic/ring.pyx. This "fixes" the above bug: 

diff --git a/src/sage/symbolic/ring.pyx b/src/sage/symbolic/ring.pyx 
index f87f622..6972018 100644 
--- a/src/sage/symbolic/ring.pyx 
+++ b/src/sage/symbolic/ring.pyx 
@@ -555,7 +555,7 @@ cdef class SymbolicRing(CommutativeRing): 
         e = pynac_symbol_registry.get(name) 

         # fast path to get an already existing variable 
-        if e is not None: 
+        if False: # e is not None: 
             if domain is None: 
                 if latex_name is None: 
                     return e