TODO, by_s_m.38 - error description?

[DougCollinge]

The TODO suggests replacing an error signal return of -99 with a crash but I'm not sure that this is the right thing to do. I don't like the way it works now because -99 has no intrinsic meaning, it just shows up on the report and you the human reader are just supposed to intuit that "something went wrong" and ignore the rest of the report. That is seriously bogus, IMHO. What I think should actually happen here is that the report should just be replaced with a detailed explanation of why the report could not be produced. In order to accomplish this by_s_m needs to be modified to allow an exit that conveys the error message, whatever it is. Perhaps it should always return a list containing an error code, error explanation, and data, if no errors.

[DougCollinge]

In fact, by_s_m has another error situation not handled at line 44. This problem should be rolled into the same fix, IMHO.

[DougCollinge]

by_s_m3 has the same issues, of course.

[boshek]

Absolutely. -99 isn't helpful at all. At least something that gives the user an idea.

[DougCollinge]

It seems that there is a package that implements try/throw/catch exceptions, which would do this job in a standard way. I'm reluctant to add a heavy dependency, however, just for this little problem. Is there something like that in that "master" roll-up package we already depend on? Sorry, can't remember what it is called.

... Nevermind. I see that tryCatch() is in the base so I'll look into using that. I think it is really better practice to use the exceptions features of the host language to handle exceptions, as opposed to cooking up yer very own scheme, as I proposed above.

[DougCollinge]

Ok, tentatively done for by_s_m and by_s_m3. You just use stop() to throw an error and tryCatch can catch the error. I don't know how to document this possibility so I haven't done it. We need a test for both these functions but Sam commented out the test file because testing against FORTRAN output no longer works. So we should still test them with some fake data. I can also test the error if I have some data.

[DougCollinge]

This new code passes the tests as-is but the various things that call by_s_m should maybe catch the stop() and do something about it. It isn't a danger because the error only arises if someone bungles the input parameters, so a simple hard stop() so they can fix their mistake is probably appropriate.

[boshek]

A stop and then maybe an informative error message?