it depends, I'm a data scientist so my code gets sent a lot of "bad" data. Sometimes it is a sign of improperly configured read mechanism (character encoding, CSV dialect are favourites). Sometimes your source does not fulfil your expectations (supplies a None when you expected an int) - this might happen millions of rows into a data source. In these cases I'd handle the issue.

#datascientist

i voted “exceptions”, because i favor the let it crash philosophy, but ill do values if it makes sense, but only if it makes sense

i get eternally annoyed by code that “handles” exceptions because “that’s what good code does”, without actually adding any value, and in a lot of cases hiding problems making it impossible to troubleshoot

I mostly opt to use return values and keep exceptions for "exceptional" situations and internally.

However like so many other things in this industry it depends on the specific situation, such as data source, data format, what it's used for and is it internal or externally provided etc.

If the bad data is coming from outside my control and it’s not something I necessarily anticipated then I would prefer an exception thrown and aggregated somewhere for better visibility.

If that specific case keeps happening often enough as to no longer be considered surprising, I may eventually codify it and make a special return value for it.

I'm in the 'it depends' crowd. If it's suspect, but I can live with it, it gets logged and I move on. If it's completely garbage, it gets logged and ignored. The only time I have exceptions is when someone needs to look at the data and decide what to do in this specific case.

I had one like this recently -- "Really? This script's been in service for six years, running every half hour, and that's the first time we've seen this error?" I implemented a fix, and things started up again.

that’s such a “there be dragons” question 😄

I tend to always throw exceptions when receiving bad data from an initialization-like call, but usually do false returning internally for little stuff.

But if I have to hand data back away from my stuff I’ll throw an exception too if something goes wrong handing it back before exiting.

If it’s something complex where I can be confident exceptions are handled I’ll use them everywhere. There’s a bit of a twisted freedom in that.