Buckaroo Banzai vs. Search & Replace
Back when I was as old as my students are now, I daydreamed about being a better-looking Buckaroo Banzai: rock star, brain surgeon, and all-around hero. These days, I daydream about search and replace. I’ve spent almost twenty hours that I really don’t have trying to get the new slides for the Software Carpentry course to format properly. Even with help from a couple of very smart people, there are still weird formatting glitches on Firefox, and they still look like crap on Internet Explorer. I’d really, really like to do a global search and replace on CSS—and when I say “global”, I mean it literally. I’d like to replace every shred of CSS on the planet with something that works and makes sense and doesn’t suck this much life force out of the people trying to use it.
I wouldn’t stop there, though. No, no, no, I wouldn’t stop there. [TODO: link to an MP3 of a mad scientist laughing maniacally.] My students have now spent more than a week figuring out how to work around Django‘s handling of unauthenticated users—search and replace. Ssetting up to webcast Software Carpentry lectures between Toronto and Edmonton is proving much harder than expected (our options seem to be riding a dinosaur, spending money that we don’t have, or using a low-res system with no guarantees of service quality). If I could have one superpower, just one, I’d replace ‘em all with something—anything—that just plain worked.
Two and a half weeks until we start trying to teach scientists how to use computers more effectively. I feel distinctly unqualified for the task right now; here’s hoping the next 18 days leave me feeling qualified…
But on the bright side, I have lunches like this to look forward to almost every day.
I am a graduate student in robotics with a good amount of software experience (see website). While gaining that that experience I have dealt with a lot of people, including my self, who can really benefit from the program you are creating here. So I am really glad you are creating it.
After looking over your slides, I feel that in #5 automated builds you are doing your students a disservice by teaching them just make, or even make at all. Its great to know the concepts behind automating build systems: file dependency, compiler/linker command generation, and system/environment introspection. Showing how Make does this is a decent way to teach that, but if you are trying to teach them best practices for building software, you should teach them a tool like CMake or SCons. Both save the user tons of time compared to make, and provide lots of useful features. CMake even has a GUI for configuring its build parameters, and generates build files that can be opened by every major IDE (VS, Xcode, Eclipse, Code Blocks) or run with make.
Thanks for your comment on my blog. I’d be happy to teach something else, as long as (a) it’s easily available on Windows and Mac as well as in widely-used Linux distros, (b) there’s good integration with IDEs, (c) it’s well documented, (d) it likely has a long life, (e) it’s easy for newbies to use, and and (f) someone is willing to write up the lecture. CONS, SCons, and Rake all fail at least two of these tests; CMake comes closer. Would you be willing to tackle (f)?
p.s. a bit of history: SCons was actually produced as part of the original Software Carpentry project back in 2000-01.
Which ones does SCons fail? (b) and (c) maybe?
Do you still need help with the CSS? That I can do. The first time around you were asking for design, which is not my forte.