Sometimes metrics may generate warnings, i.e. for incomplete results or unexpected but ignorable exceptions, for example one of the 1000 Java files has a parse error. The platform should be able to communicate such warnings to the end-user such that she can decide if there is an influence on the validity of the results.
A proposal would be to have an IPlatformMonitor callback interface which is passed as a parameter to ITransientMetricProvider.measure(...)
.
Discussion
-
jrwilliams
2014-03-11These warnings are likely to be version specific aren't they? I.e. the parse error may be fixed in the latest version. Do you want to store the entire history of warnings or just the latest? Should the user be aware that the historic metrics for certain dates may be invalid due to parse errors?
-
Jurgen vinju
2014-03-11All valid questions. So just to be safe, let's store all warnings, such that later we may decide which to present and how to the user. Also, while developing the platform this should help us with testing and diagnostics.