Installing Archangel, MacPorts, numpy, cfitsio, matplotlib, Xcode3, and pyfits on a Mac
This isn’t a “normal blog post” but I’m putting it here for my future reference (and the reference of those going through the Archangel installation process). Archangel is the last step in this setup, so this guide should be beneficial to many others when attempting to install packages like numpy or pyfits on a mac.
For background, Archangel is a utility for analyzing and extracting the photometric properties of extended objects in astronomical data. For more information, visit the Archangel Homepage.
A note before we get started: This guide is for installing these programs on a REMOTE Mac box, however all of these commands still work for if you are installing it locally.
Strategy
Our goal is to install archangel, but there’s a ton of stuff to do first. Archangel directly depends on 5 packages: python, numpy, pyfits, matplotlib, and cfitsio. These are NOT trivial to install. Things like numpy are an absolute mess because they rely on packages like BLAS and LAPACK, linear algebra routines that are about as easy to install as catching a greased pig. So, we will use MacPorts to assist us. MacPorts is essentially an “easy install” application for many unix utilities. It automatically downloads, configures, and installs a ton of great software and any prerequisates for the software you want to install. We will use it to install python and numpy (the two hardest things to install). Note that MacPorts is VERY comprehensive, almost any large utility can be found in its database, ready to install…visit macports.org for more information. I should note that the default python that ships with Mac OSX is terrible, it is a chopped-down version of the real thing, so we need to make sure we install the real thing using MacPorts.
However, MacPorts is dependent on a few things as well. First, X11, which mac ships with so we don’t have to worry about it. Second, it depends on XCode, the apple software development suite.
Installation
XCode3
To install XCode3, visit the Apple Developer’s site here and register for a free developer’s account. After registration, download Xcode 3 using the links towards the bottom of the page. Make sure you download Xcode 3! not Xcode 4 (unless you want to pay for it). Note: This is a .dmg disk image file that is about 4 GB in size. Be ready for a long download. If you are installing it locally, just double click the .dmg file and off you go. However, I am currently having to use the command line to install it because I’m installing it remotely, so here are the steps for that: (This is for OSX 10.6.6, you need to make sure your xcode download is compatible with your system. Previous Xcode’s can be found here under “Developer Tools”. You will want Xcode 3.1.4 if you are running Leopard, 3.2.6 for Snow Leopard)
- scp the downloaded file to the target system
- ssh into the target system
- Mount the .dmg disk image using the following (assuming 3.2.6 is being installed, just change all of the variables if installing 3.1.4).
hdiutil attach ./xcode_3.2.6_and_ios_sdk_4.3.dmg
- Next, run the installer on the .mpkg file inside of the mounted volume. This installer command tells it to install the .mpkg on the device “/” (our main boot partition).
cd /Volumes/Xcode\ and\ iOS\ SDK sudo installer -verbose -pkg ./Xcode\ and\ iOS\ SDK.mpkg -target /
- Now that it is installed, unmount the disk image with
hdiutil detach /Volumes/Xcode\ and\ iOS\ SDK
- You can delete the .dmg now as well.
MacPorts
Next up is MacPorts. It is fairly straightforward to install:
- Go to a new temporary directory on the target machine
- Get the tarball from macports.org (About 1 meg)
wget http://distfiles.macports.org/MacPorts/MacPorts-1.9.2.tar.gz
- CRAP! wget isn’t in Mac! We have to use MacPorts to install wget…so we’re stuck with downloading the file to our current computer, and scp’ing it to the temporary directory on the target machine. Download the URL above and SCP it to your target machine (or just keep it locally if you are installing all of this on your local machine).
- Extract it and install it:
tar -zxvf MacPorts-1.9.2.tar.gz cd MacPorts-1.9.2 ./configure make sudo make install cd ../ rm -rf MacPorts*
- Ta-da! All done installing, now we can get to business. We need to tell MacPorts to syncronize with the MacPorts server so issue the following command:
sudo port -v selfupdate
- This will update MacPorts with all of the most recent repository changes. As a little tutorial, to install something with MacPorts, you simply enter “sudo port install <package name>”. Note: you should run the selfupdate command on a regular basis to keep everything up to date. You can “man port” for more information after you install MacPorts.
- Note: If your system doesn’t recognize the “port” command, you need to add /opt/local/bin to your PATH (add the following line to your .cshrc file):
setenv PATH "/opt/local/bin:$PATH"
Python, Numpy, Scipy, matplotlib (2.6), & cfitsio (THANK GOD FOR MACPORTS)
Start by installing Python 2.6. We do this separately from the rest because we want to make sure we set it as our default python version before proceeding. In short, the python that ships with OSX is chopped up so much because Apple uses it a lot when booting up the computer, so they take out a whole bunch of stuff that slows the boot a little bit to make it as fast of a boot as possible. Unfortunately, this means half of python is missing!
sudo port -v install python26 python_select
Give it time, quit being impatient! (that’s a note to myself) Now, set our python install as the default for our system:
sudo port select --set python python26
Note that doing “python_select -l” will list all of the python installs to choose from on your machine.
Now, we’ll install numpy, scipy and all of their dependencies all at once! Scipy is not necessary, but it’s extremely useful for other things, so I’m installing it as well. Scipy depends on numpy, which depends on python…so if we tell MacPorts to install Scipy for python 2.6, everything will be installed! (also there are a ton of dependencies here I’m not going into. When you run the command it will tell you all the stuff it is going to install in addition to scipy, numpy and python…you can see it is a HUGE list of dependencies. Thank goodness we don’t have to install all of these by hand!). We also throw matplotlib and cfitsio into this command to get them taken care of as well (and wget, I can’t live without wget).
sudo port -v install py26-scipy py26-matplotlib cfitsio wget
Warning: This will take FOREVER…hours on a decent machine (took one of the ones I did this on over 5 hours). There are so many libraries to install and BLAS, LAPACK, numpy, and scipy alone take forever to configure and install. Go get a delicious beverage, eat a donut, or better yet, sleep…it might be done by morning if you’re lucky.
After all that work, we need to test it. Fire up a new terminal and try importing each of the libraries, if there’s no error, everything went smoothly 
% python Python 2.6.6 >>> import numpy >>> import scipy >>> import matplotlib >>> exit()
Pyfits
Pyfits is maintained by STSci and is a simple ‘normal python install’:
wget http://www.stsci.edu/resources/software_hardware/pyfits/pyfits-2.4.0.tar.gz tar -zxvf pyfits-2.4.0.tar.gz cd pyfits-2.4.0 sudo python setup.py install
It’s always good to test the install so, in a DIFFERENT DIRECTORY (never ever ever ever run a python package that you just installed from the install directory) do:
% python Python 2.6.6 >>> import pyfits >>> exit()
If there’s no error, you’re good!
Archangel
FINALLY, we’re ready to install archangel. It has been a long road, but if all went smoothly, this should be SIMPLE. Go into a NON-temporary directory (the directory where you want to store the archangel utility subdirectory, I did it in a /archangel root directory so everyone would have access to it) and issue:
mkdir /archangel cd /archangel sudo wget http://abyss.uoregon.edu/~js/archangel/archangel_v2.5.tar.gz tar -zxvf archangel_v2.5.tar.gz mv archangel/* .
This next part is important…you choose. If you are just setting archangel up for one account, you can run the following line of code NOT as sudo and choose the place you want to place the binaries for yourself. Then you will need to modify your .cshrc file to reflect where these binaries are. Here’s how:
python setup.py build (choose binary directory)
Next, view the readme file in that directory to see how to modify your .cshrc file.
For my install, I want everyone to be able to use it so I run:
python setup.py build Binaries going to -> *location* Ok? /usr/local/bin
And it should put everything into my system path at /usr/local/bin. /usr/local/bin should be in everyone’s default path so there isn’t a need to do that .cshrc stuff BUT you do need to do one .cshrc thing:
setenv ARCHANGEL_HOME /archangel
Note what I did here. I made the archangel_home directory a new directory at the / level. This is so that all users can access it. EVERYONE who is going to use archangel then needs to put that in their .cshrc file.
Test it! Source the .cshrc file (I would actually recommend totally logging out and back in), go to your home directory (or wherever you want to store your archangel outputs) and run:
cp /archangel/examples/ic5271_j.fits . profile ic5271_j.fits
If all goes well, you have successfully installed Archangel . Now go read the documentation!