Error handling
In an ideal world, all software would work perfectly all the time. However, in the real world software encounters errors due to the outside world, bad input, bugs, or programmer error.
Types of errors
Sentinel values
It is the position of this project that error conditions that happen often, typically due to bad data, should be handled with sentinel values returned to indicate the failure condition. These are used for functions such as add_edge!(g, u, v). If you try to add an edge with negative vertex numbers, or vertices that exceed the number of vertices in the graph, then you will get a return value of false.
Errors / exceptions
For more severe failures such as bad arguments or failure to converge, we use exceptions. The primary distinction between Sentinel Values and Argument Errors has to do with the run time of the function being called. In a function that is expected to be a called in a tight loop such as add_edge!, we will use a sentinel value rather than an exception. This is because it is faster to do a simple if statement to handle the error than a full try/catch block. For functions that take longer to run, we use Exceptions. If you find an exception with an error message that isn't helpful for debugging, please file a bug report so that we can improve these messages.
ArgumentError: the inputs to this function are not validInexactError: some types are used that cannot express a computed quantity with the necessary precision