I have been using the new version of RIVET for some exploratory data analysis, and am general happy with the experience.
However, I've found that the design of the file input dialogue makes it too easy to incorrectly choose the default "points" instead of "metric" when working with a distance matrix without the --datatype metric flag included. I made this mistake a couple of times. It turned out that the results with "points" were actually kind of similar; it is not so obvious from the output that you have done the computation incorrectly, and that is a recipe for problems.
So I suggest that if a .csv file is given without the datatype file, the datatype selector defaults to "no selection" or something similar, and the user has to explicitly choose "points" or "metric" before they can press "compute."
I suppose this is also a potential problem on the command line, but there I'm less sure what the right solution is. Should we eliminate the default datatype altogether?
I have been using the new version of RIVET for some exploratory data analysis, and am general happy with the experience.
However, I've found that the design of the file input dialogue makes it too easy to incorrectly choose the default "points" instead of "metric" when working with a distance matrix without the --datatype metric flag included. I made this mistake a couple of times. It turned out that the results with "points" were actually kind of similar; it is not so obvious from the output that you have done the computation incorrectly, and that is a recipe for problems.
So I suggest that if a .csv file is given without the datatype file, the datatype selector defaults to "no selection" or something similar, and the user has to explicitly choose "points" or "metric" before they can press "compute."
I suppose this is also a potential problem on the command line, but there I'm less sure what the right solution is. Should we eliminate the default datatype altogether?