| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
| |
standalone is equivalent to distributing scipy_core.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
manipulate VTK objects.
|
| |
|
|
| |
weave. The source for the SWIG wrapped extension modules are also included.
|
| | |
|
| |
|
|
| |
PyCObject and not a string to encode the pointer to the C++ object.
|
| |
|
|
| |
arch-dependent Fortran compiler flags.
|
| |
|
|
| |
weave.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
it isn't available.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Fernando Perez fix for random uncessary recompilation with weave.
|
| |
|
|
| |
Python 2.3)
|
| | |
|
| |
|
|
|
|
| |
- set self.compiler_cxx to g++ in __init__. This should really be done through the set_exectuables method,
but it would break pre 2.3 support because compiler_cxx is not an accepted option.
- added target_lang argument to link() method.
|
| |
|
|
| |
avoid problems with Windows installer tools.
|
| | |
|
| |
|
|
| |
module is not installed but getmodule tries to access its __dict__
|
| |
|
|
| |
namespaces in C++. This is used in the freetype and agg wrappers.
|
| | |
|
| |
|
|
| |
that C++ files are linked correctly. If fails over to distutils if scipy_distutils isn't present in hopes that it will do the right thing.
|
| |
|
|
|
|
|
|
|
|
|
| |
distutils.core.Extension object for a weave generated extension. This
is helpful in using weave extensions within setup.py files.
added logic inside ext_module.generate_module to check whether the
new extension module code is different from the code that may already
exist in a previously generated extension module file. If the code is
the same, the file is not overwritten. This prevents setup.py from thinking
the module is always out of date and must be recompiled.
|
| |
|
|
| |
Py2.2)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
New scipy import timings on PII 400MHz, Debian
with postponed import hooks enabled:
pearu@localhost:~$ time python2.2 -c 'import scipy'
real 0m1.392s
user 0m1.320s
sys 0m0.060s
pearu@localhost:~$ time python2.3 -c 'import scipy'
real 0m0.691s
user 0m0.650s
sys 0m0.030s
For reference, importing scipy to Python 2.2 used
to take 2-16 seconds depending on whether scipy was
loaded before or not.
2) New feature in ppimport:
When pre_<modulename>.py* exists for ppimported
module <modulename> then this pre_*.py file will
be executed to initialize additional attributes
to postponed module loader. For a python package,
<packagename>/pre___init__.py is used to add
attributes to package loader.
This feature is useful for importing documentation
strings of a package/module without importing them.
Currently this works with scipy.info but I haven't
tested with pydoc.help yet.
3) Made initial steps for getting scipy documentation
strings parsable by docutils.
|
| |
|
|
| |
to escape \n characters so that they would not be interpreted by python and would actually get written into the generated C++ file. This patch fixes that.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
makes it where the code doesn't have to be read out of a file in the directory path and makes it easier on py2exe like programs that bundle code into an exe file. [see bug 37]
I think weave will still not work in bundled applications because it referes to a lot of other external files in the scxx directory and blitz directories. Still, these are not read during import, so they at least don't cause the scipy import to fail.
We should look for a way to make weave work from py2exe binaries in the future.
|
| |
|
|
| |
-shared instead of dllwrap for building dlls
|
| | |
|
| |
|
|
| |
undoing part of the previous fix added to the main chaco setup.py file.
|
| |
|
|
| |
Split blitz tests out so that the individual numeric types are tested separately.
|
| |
|
|
|
|
|
|
| |
modules. It is missing -Wno-long-double to supress the output of this
warning. CFLAGS has this flag, so I just set OPT to CFLAGS in
distutils.sysconfig.config_vars. THis also adds a flag that turns off
pre-compiled headers in c++ builds. I don't know if this will prove an issue
in the future or not, but it doesn't affect any of the weave tests.
|
| |
|
|
| |
more generic purposes beyond just getting the compiler information.
|
| |
|
|
|
|
|
| |
This prevents the possibility of compilers from different platforms
from looking in the python_intermediate directory and finding an object
file built by a different compiler and trying to link it in -- which will
fail.
|
| |
|
|
|
|
|
|
| |
This is needed so that we can create compiler specific directories for
intermediate files (.o) created by weave. An md5 check sum of the
compiler binary provides a reasonably reliable way to differentiate
between compilers. This approach works across multiple platforms that
share the same file system.
|
| |
|
|
| |
repr() and str() now return std::string instead of py::object values
|
| |
|
|
| |
freetype library.
|
| |
|
|
|
|
| |
Numeric arrays. If we could use std::wstring reliably on all platforms,
the need for this would go away. My notes say that it doesn't compile
correctly on gcc2.95.3 though.
|
| | |
|