I find myself increasingly preferring the exists()-then-open() sequence to open()-and-catch-NotFound (not just in Python, but generally).
Yes, it is classically frowned upon, but:
1. It is typical that there are multiple checks that need to be done in addition to exists(). It is a given that conditionals are more flexible than exceptions (even in Python). As long as some checks are in conditionals, it makes the code more legible and easier to reason about when all preliminary checks are in the same fashion and in the same place.
2. You still need to handle the other possible sub-varieties of OSError and other exceptions that can arise due to, say, I/O being famously unreliable. If you implement that handling gracefully, you will have the FileNotFoundError covered for free. From end user experience perspective, since a file being deleted between exists() and immediate open() on a single system within a single thread is an out of ordinary one-in-a-thousand-years case, it is OK if it is handled slightly less gracefully than if it occurs during the preliminary checks.
