bug in caching assumptions

[certik]

While implementing the issue 3428 , I discovered a very nasty bug. Try this:

In [1]: re(x)
Out[1]: re(x)

In [2]: x.series(x,1)
Out[2]: O(x)

In [3]: re(x)
Out[3]: re(x)

ok, that's fine. Now go to core/function.py and comment out the caching,
like this (line 62):

```
#@cache_it
def __new__(cls, *args, **kwargs):
```

and do it again:

In [1]: re(x)
Out[1]: re(x)

In [2]: x.series(x,1)
Out[2]: O(x)

In [3]: re(x)
Out[3]: x

This bug then influences many other tests etc. Please don't fix it though,
until I commit my changes from the branch.

Original issue for #3468: http://code.google.com/p/sympy/issues/detail?id=369

Original author: https://code.google.com/u/104039945248245758823/

Referenced issues: #3428

[certik]

Just for the impatient - the problematic line is the hack in sympy/series/order.py:162:

```
    for s in symbols:
        assume_dict = {}
        if not s.is_infinitesimal:
            assume_dict['infinitesimal'] = True
        if s.is_positive is None:
            assume_dict['positive'] = True
        if assume_dict:
            s.assume(**assume_dict)
```

which adds assumptions to the symbols, and they propagate out of the function, of
course. The caching mechanism probably fixes that.

Note again - please don't fix it, until I commit, probably late at night.

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c1

Original author: https://code.google.com/u/104039945248245758823/

[certik]

Rising this to high, because I need to take some action now, since it is breaking my
new limits code.

**Labels:** -Priority-Medium Priority-High  

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c2

Original author: https://code.google.com/u/104039945248245758823/

[certik]

Lowering to medium, the new limits code seems to work quite well now.

**Summary:** bug in caching assumptions  
**Labels:** -Priority-High Priority-Medium  

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c3

Original author: https://code.google.com/u/104039945248245758823/

[navytux]

Smaller steps to reproduce:

In [1]: O(x**2)
Out[1]: O(x**2)

In [2]: re(x)
Out[2]: x

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c4

Original author: https://code.google.com/u/111152560333599832822/

[navytux]

Patch sent for review: http://groups.google.com/group/sympy-patches/msg/f0ac17b32be586d8

**Status:** Started  
**Labels:** -Type-Defect Type-Patch  

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c5

Original author: https://code.google.com/u/111152560333599832822/

[certik]

Please commit.

**Labels:** Milestone-Release0.5.12  

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c6

Original author: https://code.google.com/u/104039945248245758823/

[navytux]

committed: http://hg.sympy.org/sympy/rev/941cad99f705 What's the state of this issue?

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c7

Original author: https://code.google.com/u/111152560333599832822/

[certik]

I think it's fixed by your patch. Test also looks good. 

Now maybe some other problematic things with series could start working, or at least
be much more easier to fix.

Thanks for the patch!

**Status:** Fixed  

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c8

Original author: https://code.google.com/u/104039945248245758823/

[navytux]

Thanks!

I've seen there are some new XFAILs :)
btw: this was so easy to fix, why had we waited so long?

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c9

Original author: https://code.google.com/u/111152560333599832822/

[certik]

> btw: this was so easy to fix, why had we waited so long?

It wasn't easy at all in September. But we polished and fixed sympy a lot since then,
so it was easy now.

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c10

Original author: https://code.google.com/u/104039945248245758823/

[navytux]

yep, goodbye old bug.

Original comment: http://code.google.com/p/sympy/issues/detail?id=369#c11

Original author: https://code.google.com/u/111152560333599832822/