When I tried to do something with Anaconda (https://store.continuum.io/cshop/anaconda/), I got angry with munmap_chunk (): invalid pointer when reading a large table with pandas, so I gave up. I built the Python 3.4 system from the source and built the environment for scientific calculation, so the summary.

I heard that canopy is very heavy, so I haven't tried it in the first place.
I didn't notice it at this time, but it seems that the error itself was an error of the read_json () method of pandas. .. ..

Prepared environment

Scientific Linux release 6.5 (Carbon)

Python build

For details, it is faster to see the following site http://toomuchdata.com/2014/02/16/how-to-install-python-on-centos/

Before building the Python body, install the necessary libraries. Even if these are not included, the Python installation itself will pass safely, but later I will suffer from a not found error in the library and I will have to reinstall Python, so I will put it in first anyway. (Self-discipline)

$ sudo yum groupinstall --skip-broken "Development tools"
$ sudo yum install zlib-devel bzip2-devel openssl-devel ncurses-devel sqlite-devel readline-devel tk-devel

Build Python itself. (This time we will introduce the current latest version 3.4.2) This time, install it under / usr / local. Check the version from the following site. https://www.python.org/downloads/source/

$ wget https://www.python.org/ftp/python/3.4.2/Python-3.4.2.tgz
$ tar xzvf Python-3.4.2.tgz
$ cd Python-3.4.2
$ ./configure --prefix=/usr/local --enable-unicode=ucs4 --enable-shared LDFLAGS="-Wl,-rpath /usr/local/lib"
$ make
$ sudo make install

Installation is complete if the following command can be executed.

$ which python3
/usr/local/bin/python3

If the Python build is successful, pip should be installed automatically, so check it with the following command.

$ which pip3
/usr/local/bin/pip3

Installation of various packages

Installation of packages for scientific calculations

Install the packages required to install Scipy in advance.

sudo yum install blas-devel lapack-devel

Except for specific packages, it can be introduced smoothly with pip3.

Since env_keep was not done in sudoers, pip3 is called with an absolute path.

$ sudo /usr/local/bin/pip3 install nose pytest numpy scipy scikit-learn pandas

Install matplotlib

Install the packages required to install matplotlib in advance.

$ sudo yum install libpng-devel

When I try to install matplotlib with pip3, the build fails with the following message.

============================================================================

                        * The following required packages can not be built:

                        * freetype

----------------------------------------
Cleaning up...
Command python setup.py egg_info failed with error code 1 in /tmp/pip_build_root/matplotlib
Storing debug log for failure in /root/.pip/pip.log

If you look closely at the error message, you will find the following message.

REQUIRED DEPENDENCIES AND EXTENSIONS
...
              freetype: no  [Requires freetype2 2.4 or later.  Found

                        2.3.11.]

In order to install matplotlib, freetype2 2.4 or higher is required, but it seems that 2.3.11 is included by default. Here, I decided to build freetype from source. The source can be downloaded from the following site. http://download.savannah.gnu.org/releases/freetype/

$ wget http://download.savannah.gnu.org/releases/freetype/freetype-2.5.3.tar.gz
$ tar xzvf freetype-2.5.3.tar.gz
$ cd freetype-2.5.3
$ ./configure
$ make
$ sudo make install

When I tried pip3 again after finishing up to this point, the following error occurred this time.

Downloading/unpacking matplotlib
  Downloading matplotlib-1.4.0.tar.gz (51.2MB): 51.2MB downloaded
  Running setup.py (path:/tmp/pip_build_root/matplotlib/setup.py) egg_info for package matplotlib
    Traceback (most recent call last):
      File "<string>", line 17, in <module>
      File "/tmp/pip_build_root/matplotlib/setup.py", line 154, in <module>
        result = package.check()
      File "/tmp/pip_build_root/matplotlib/setupext.py", line 945, in check
        min_version='2.4', version=version)
      File "/tmp/pip_build_root/matplotlib/setupext.py", line 433, in _check_for_pkg_config
        if (not is_min_version(version, min_version)):
      File "/tmp/pip_build_root/matplotlib/setupext.py", line 166, in is_min_version
        return found_version >= expected_version
      File "/usr/local/lib/python3.4/distutils/version.py", line 76, in __ge__
        c = self._cmp(other)
      File "/usr/local/lib/python3.4/distutils/version.py", line 343, in _cmp
        if self.version < other.version:
    TypeError: unorderable types: str() < int()

After investigating this error, I found that it seems to be a known bug in Python 3.3 / 3.4 series. See the issue below for details. http://bugs.python.org/issue14894

Download the 14894.patch listed here and apply it to Python. (When will this change be applied to Python itself and released ...)

$ wget http://bugs.python.org/file25808/14894.patch
$ cd /usr/local/lib/python3.4/distutils/
$ sudo patch -p3 < /path/to/patch/14894.patch

Matplotlib can be installed by executing pip3 at the end of this process.

$ sudo /usr/local/bin/pip3 install matplotlib

Install the package to run IPython Notebook

Introduced the following packages.

$ sudo /usr/local/bin/pip3 install ipython jinja2 tornado pyzmq

end

After that, I ran IPython Notebook Server and put in pytest / Theano / wabbit_wappa, but it became difficult to put together, so it ends here. For IPython Notebook Server, the following sites are well organized for your reference. http://akiniwa.hatenablog.jp/entry/2013/11/25/001805

Enjoy PyData!

Summary from building Python 3.4. * From source to building a scientific computing environment