I just had a conversation on #python about this. After a bit of talk about "from __future__ import absolute_import", which doesn't work in Python 2.4 and would require us changing all of our non-absolute imports everywhere (the person I talked with said we would pretty much have to use it everywhere in SymPy), explicit relative imports (also not supported in Python 2.4, but supported without __future__ in Python 2.5+), and renaming symbol.py (which would also be a lot of work), I figured out that the solution is much simpler than that. We just need to move ast_parser_python24.py, which is the only file that uses this stdlib_symbol, to a separate directory from sympy/core.
So I propose that we move that file, as well as ast_parser.py and sympify.py, to sympy/parsing. If there are no objections, I will create a patch for this. Aaron Meurer On Dec 28, 2010, at 2:30 PM, Aaron S. Meurer wrote: > This *seems* to fix it. The only failure when that import is removed is > test_sympify in Python 2.4 and Python 2.5, and it still passes after this > diff is applied. The idea is that once you import a module, it is put in > sys.modules, even if you immediately delete the module from the namespace > after you import it. I don't know how supported this behavior is, though. > It seems rather hackish to me. > > diff --git a/sympy/__init__.py b/sympy/__init__.py > index 60da6e1..bbc9654 100644 > --- a/sympy/__init__.py > +++ b/sympy/__init__.py > @@ -19,6 +19,7 @@ def __sympy_debug(): > SYMPY_DEBUG = __sympy_debug() > > import symbol as stdlib_symbol > +del stdlib_symbol > from sympy.core import * > from assumptions import * > from polys import * > diff --git a/sympy/core/ast_parser_python24.py > b/sympy/core/ast_parser_python24.py > index ed7e41b..1e21d62 100644 > --- a/sympy/core/ast_parser_python24.py > +++ b/sympy/core/ast_parser_python24.py > @@ -4,6 +4,7 @@ > from compiler.ast import CallFunc, Name, Const > from compiler.pycodegen import ExpressionCodeGenerator > import re > +import sys > > from basic import Basic > from symbol import Symbol > @@ -52,7 +53,7 @@ def atom_name(self, nodelist): > > def lambdef(self, nodelist): > #this is python stdlib symbol, not SymPy symbol: > - from sympy import stdlib_symbol > + stdlib_symbol = sys.modules['symbol'] > if nodelist[2][0] == stdlib_symbol.varargslist: > names, defaults, flags = self.com_arglist(nodelist[2][1:]) > else: > > Aaron Meurer > > On Dec 27, 2010, at 10:46 PM, Aaron S. Meurer wrote: > >> Well, it's a problem, because we have stdlib_symbol imported into the >> namespace with from sympy import *. I wonder if this can be fixed by >> manipulating sys.modules? >> >> Aaron Meurer >> >> On Dec 21, 2010, at 9:31 PM, Ondrej Certik wrote: >> >>> On Tue, Dec 21, 2010 at 9:54 PM, Aaron S. Meurer <asmeu...@gmail.com> wrote: >>>> Why is this line >>>> >>>> import symbol as stdlib_symbol >>>> >>>> in the main sympy/__init__.py? If I comment it out, there are some test >>>> failures, but it seems that wherever this is used it should be imported >>>> there (there is no reason that it should be imported into the main >>>> namespace). By the way, the symbol module is "Non-terminal symbols of >>>> Python grammar (from "graminit.h")." >>> >>> I think that the problem is that it needs to be imported in >>> sympy/core, and there is already a file called symbol.py, thus there >>> is no way to import the stdlib symbol. Thus we need to import it one >>> level up. >>> >>> Ondrej >>> >>> — > -- You received this message because you are subscribed to the Google Groups "sympy" group. To post to this group, send email to sy...@googlegroups.com. To unsubscribe from this group, send email to sympy+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sympy?hl=en.
