Delegate error printing to the ajax caller.

[Aiosa]

By default, OSD prints an error if ajax call results in an error response code (other than 2xx). This is a bit problematic since one can use custom error codes to route the behaviour and the console gets bloated by unnecessary error messages (I am having a scenario where tiles are expected to be very unpredictable, and the system can automatically detect that and remove the tiled image: having hundreds identical errors in the console does not help much). Other problem is that each error response generates two error messages, following the how not to scenario:

try:
    raise Error("Test error")
except e:
    print(str(e))  # do not print error messages if the routine error handling happens elsewhere
    raise e

so the ajax function should print an error only if no error handler is registered, otherwise, it should be the handler's responsibility to do so. Following the future PR I am planning, the user will have also the ability to re-define how tiles are downloaded and thus control the errors completely.

[iangilman]

Awesome... Great cleanup!

[iangilman]

Actually, one thought...

[iangilman]

Should we not write to the console here if somebody has subscribed to the open-failed event? We may need to add something like hasHandler to EventSource to make that possible, but it shouldn't be hard.

[Aiosa]

Good idea!

[Aiosa]

Actually, maybe not... The problem is open-failed event is registered inside OpenSeadragon to print the message into the error html container, so in 99,9999% scenarios this equals to not logging the message at all...

[Aiosa]

I found out while adding test case for this commit (yeah, should do it in one commit) that OSD actually uses the event internally, so the event handler count is almost always > 0 and therefore the error gets never printed... I would revert the last commit and just go with it as it was... I mean 'open' event happens once per tile source, so it is not that big a problem, unlike the image tile fetching itself. What do you think?

[iangilman]

Good catch! Yeah, please undo that change. Might as well keep numberOfHandlers though; seems like it could be useful in the future (if even just for testing).

[Aiosa]

Done

[iangilman]

Fantastic! Thank you for the tests for numberOfHandlers :)