Installing root_numpy and rootpy via conda

We also provide conda distributions of root-numpy (the interface between ROOT and NumPy) and rootpy. When installing root-numpy, ROOT's latest version will be picked up as a dependency:

$ conda install root-numpy

as can be seen from the conda installation plan, the currently-latest (6.04) ROOT version will be picked up:

The following NEW packages will be INSTALLED:
    ...
    numexpr:       2.4.6-np110py27_1    defaults                                 
    numpy:         1.10.4-py27_1        defaults                                
    readline:      6.2.5-15             https://conda.binstar.org/NLeSC/linux-64/
    root:          6.04-py2.7_gcc4.8.2  https://conda.binstar.org/NLeSC/linux-64/
    root-numpy:    4.4.0-root6.04_py2.7 https://conda.binstar.org/NLeSC/linux-64/    
    ...           

Proceed ([y]/n)?

If you rather want to have another ROOT version, specify that version explicitly:

$ conda install root-numpy root=5

Now conda proposes:

The following NEW packages will be INSTALLED:
    ...
    root:          5.34.32-py2.7_gcc4.8.2  https://conda.binstar.org/NLeSC/linux-64/
    root-numpy:    4.4.0-root5.34.32_py2.7 https://conda.binstar.org/NLeSC/linux-64/
    ...

When installing rootpy, both root-numpy and ROOT will be picked up as dependencies automatically. When installing rootpy, the same above holds for choosing your ROOT or Python version.

rootpy.stl: STL Dictionary Generation

From the rootpy documentation:

This module allows C++ template types to be generated on demand with ease, automatically building dictionaries with ROOT’s ACLiC as necessary. Unlike vanilla ACLiC, rootpy’s stl module generates and compiles dictionaries without creating a mess of temporary files in your current working directory. Dictionaries are also cached in ~/.cache/rootpy/ and used by any future request for the same dictionary instead of compiling from scratch again. Templates can be arbitrarily nested, limited only by what ACLiC and CINT can handle.

Note: The compiled libraries are by default stored in ~/.cache/rootpy/{architecture}-{root-version}. If that directory is inaccessible, they will be placed in /tmp.