attempt_to_blog++

It’s unfortunate that this blog is turning into rants about scientific computing. Perhaps I am simply ethnically predisposed to doing science and kvetching.

Today’s rant is brought about by my collaborators’ request to make my paper figures “presentable”. The reason is simple — current figure drafts are straight up Mathematica output, relying mostly on defaults, and they suck. Let’s face it — figures produced by Mathematica can be good, but rarely great, and trying to fine-tune the appearance of frames and axes can be a daunting task.

The problem isn’t that Mathematica sucks at plots — so does most other software. The problem is that to my knowledge there are no reasonable open-source alternatives to something like Origin. So right now, if you are unsatisfied with Mathematica’s plotting capabilities, your options come down to this:

- get as far as you easily can with Mathematica, then switch to Illustrator and Photoshop. Pros: guaranteed to work. Cons: time-consuming, labor-intensive, and requires Illustrator, Photoshop, and Windows.

- import data into one of {R, Matlab, Python/Matplotlib} and hope it can do what you are after. Pros: you might get the output you want. Cons: you need to know R, Matlab, or Python.

What shocks me is that in this day and age, it seems that every few years, somebody sits down to write a plotting package, and ends up reinventing the wheel. So you have a bunch of wheels, none perfectly circular; there are all sorts of bumps and protrusions, all in different places. Within R, for instance, there are at least 3 different ways to create plots (base graphics, ggplot2, lattice), and more are being created (e.g. jjplot). However, despite all the man-hours spent, we still don’t have such basic things as TeX integration. In a software package that claims to be the premier graphics solution for a statistician/applied mathematician! The syntax to add formulas to plots is revolting. Even Excel can do better.

All right, rant over. I need to go generate some figures.