Cool. Got a patch?

Not working on it, yet, but I'll probably get to it eventually if someone doesn't beat me to it. I figured any ideas should be logged in case anyone gets a wild hair and tries to tackle it before me.

Another idea on the same topic: aggregators for code inspectors (PMD, CheckStyle, FindBugs) provide only one value, the total sum of errors or violations for all files.

Although this is useful, the distribution of the errors is lost. For example, "1000" doesn't tell you if you have 2 errors in each of 500 files or 999 errors in one and 1 in another.

I think an aggregator that would calculate some descriptive statistics of the values would be more helpful; for example the median and deviations from it. Those numbers could be presented in a single bar chart, similar to the progress bar used now for percentages. Let's say the aggregator returns a series of numbers in the format N1:N2:N3:N4 (which could represent first deviation, median, second deviation and total). We would draw a chart with 4 segments in different colors, scaled to match the default bar width.

For a nice demo on how to read mean ane median values, check: http://www.ruf.rice.edu/~lane/stat_sim/descriptive/

Sure, go for it and write new "median" aggregators if you're interested