We've been using Skywalker to generate plots for our different codes, and then eyeballing those plots. It would be tremendously useful to have a tool that accepted two Skywalker output modules and emitted a report giving some measures of the differences in the output quantities in these modules. Here's a proposal for such a tool:

1. The tool accepts the names of two skywalker output modules as command line arguments
2. It imports the two skywalker outputs modules and mines them for output variables. Maybe it can also do some simple validation to ensure that the output modules contain data for the same variables.
3. For each output variable, the tool computes some set of error norms (Inf, 1, 2 norms?), perhaps including both absolute and relative measures. We could establish a convention where the first module passed as an argument is the "reference" dataset, for purposes of computing relative error norms.
4. When it has computed all the error norms for the output modules, the tool emits a text file containing a report of the error norms. We can start with a simple plain text file and play around with it. Down the line, if we can think of a workflow that makes use of this information in some appropriate format, we can switch to that format.

@overfelt , how does this sound? Any thoughts?

Hi,
I am getting this error:

./build-haero.sh: line 76: ${BUILD_TYPE^}: bad substitution

After running the build-haero.sh script:

./build-haero.sh build

I am using a mac.