Hi all,
Marcus here, and ready to share with you what i've accomplished in the while working in the lab this Spring. This project was quite an undertaking from the very beginning; I had just conquered the mighty task of “Hello World” in java when I found out that I’d be paired with Joe to prevent more Excel made graphs from entering publication. We were given the option to either work in R, a completely unfamiliar language to both Joe and I, or to see if there were any ways to get Python to co-operate. Enter matplotlib (MPL).
This library was so combative that we had to make use of a separate application called Jupyter, which already had MPL integrated in it, to start working. Determining a starting location was a task in itself. Our first graph made extensive use of an oh-too-kind stackoverflow users code which gave us a 7x7 grid of information. That was a lot to take in, so we started looking into how get more specific with what we were presenting. The next graphs we created were specifically targeting lead, in-channel vs overbank and resample vs original, including variants that had data that accounted for negative values and those that didn’t. The nature of working with lead values is that they all had crazy error-bars, which got to be distracting visually. Suffice to say, there were a lot of moving pieces that didn’t want to work together at first.
This is when we started changing up our approach, Joe had delegated me to be more in charge of creating the graphs themselves while she continued working on her already built code that was able to grab and pair related information sets. My job required me to understand what data I was being given and how to use that in the graphs I would be creating, so naturally I needed to understand, at least some parts of, the code Joe was already working on. Now my to-do list included: learning how Jupyter worked, learning how to use MatPlotLib, understanding how Joes code was structured. Maybe it would’ve been easier to work in R. Jokes aside, it was a somewhat daunting task, so I figured it’d be best to start by seeing how Joes code worked, as she was much more accessible than the authors of the other two applications.
It was a really interesting experience getting to see how Joe went about setting up her code to retrieve data and return meaningful results. The structure she had set-up worked in a way that was fairly straightforward, so figuring out how to add to and build off of what was there already wasn’t too difficult. It was a nice introductory period before we got our hands dirty with MatPlotLib. As I mentioned earlier, we really started off making plots that looked nothing like what our final graphs became. It was definitely a learning experiencing as we figured out what worked best and how to get information dense graphs that were still intelligible. Stackover-flow and the many documentation web pages quickly became purple links in my google searches as to why this wasn’t working or how to change this seemingly obvious part of the graph. Changing from regular plots to scatterplots, playing with subplots and legends and even changing the font size, color and shapes all seemed to have specific intricacies that want to provide un-intuitive results for seemingly small modifications. Ultimately though, Joe and I were able to work through these bugs as they popped up and create some good figures in the process.
No comments:
Post a Comment