Fetching data from REST APIs or a plain HTTP server is usually a part of my work. Many problems may arise during fetching data; for example, a server goes offline, my storage is out of space, an API gives the wrong URL, a server sends data in an invalid format, etc. To illustrate the situation, I can oversimplify my task to the function below:
def fetch(uri):
requests.get(uri)
A fetcher usually faces one of these problems. I don't want to rerun a fetcher after running it for an hour or a day. Maybe I can write my fetcher to resume working after it stops at any state, but it will take time, and my fetcher will be more complex. If I code in Python, the fetch function can raise many exceptions, and my fetcher will stop because my code didn't handle any exceptions.
I see two types of environments that will help me fix the problem.
Type checker may help me handle as many types of exceptions as possible upfront. Java definitely can do this because, in Java, I have to declare the list of exceptions as a part of the fetch function, and Javac checks if the caller handles these exceptions. Rust doesn't use a Java-like exception, but the result type encodes the types of errors, so the Rust compiler can check if a caller handles these errors.
Common Lisp runtime will run a debugger if unhandled errors occur. So I can choose what to do next. For example, I can empty my trash folder if my storage is out of space. I still don't know if I can handle the storage space problem later with Erlang runtime.
Python doesn't seem to belong to (1) nor (2). I tried mypy, a static type checker for Python, but it didn't warn me anything. And I don't know how to retry or do something else with Python's exception system rather than let it stop.
Top comments (0)