I finally got the calendar working! It turned out that the form name was different in different places (the calendar elements' attributes were using something else). Since this popup calendar puts the data into a form field, I was able to get rid of some extra processing code.

I showed my work at the meeting today. The automatically-calculated maximum value for the y-axis was sometimes nearly twice as high as appropriate, leaving a lot of white space on the graph. The graphs were also too tall to fit all together on the page, making it difficult to compare data. People seemed to like being able to view 90 days, as this makes it easier to catch intermittent problems. and reaction to the calendar date picker (including the non-tiny font sizes) was positive. We talked about whether to use a weighted sum of the restlessness readings to determine time spent in bed (probably not). We discussed how and whether to use a thermometer in addition to the graphs. It was mentioned that the colors were hard to read, especially since each bar and color is outlined and sometimes the outlines are not much slimmer than the bars themselves. Andy had the idea of a "moving average" - each day could have an indication of what the average was over the past X days from that day, giving a smoother graph.

I addressed a lot of the issues mentioned after I got back to work. I set up the y-maximum to automatically calculate based on the largest value, and that looks a lot better. I set the graph heights to about half of what they were, and now all 3 graphs fit on the screen if the resolution isn't too low. I set up a little more error checking on the dates. The default date range when the page loads is from "yesterday" to 2 weeks prior (I don't like to include "today" by default because the data is incomplete). It can now compensate for lack of an ending date, an ending date before the start date, or an automatically generated ending date (due to one of those errors) that goes past "yesterday." If the popup calendar is used, most of that can't happen. But if javascript is turned off and the user can enter any date they please, any of those (and more!) can happen.

I learned that the syntax to raise x to the y power in Perl is x**y, not x^y.

We decided not to make the other type of bar graph that we had discussed, a graph that has one set of data pointing downward and the other upward for comparison. It seems that what I've got now is going to be more useful.

I need to look over the feedback I got and implement whatever still needs to be done. I need to start on the new visualization, which entails me doing a little research on the best implementation in Catalyst. I need to at least sketch out a rough idea of my paper, especially since that will probably be harder to pick back up after vacation than coding-based projects. And I'm waiting to hear back on a possible bug in the graph that is making small "phantom data" appear.