PyFlink: Shall we disallow relative URL for filesystem path?

[Yik San Chan <ev...@gmail.com>]

Hi community,

When using Filesystem SQL Connector, users need to provide a path. When
running a PyFlink job using the mini-cluster mode by simply `python
WordCount.py`, the path can be a relative path, such as, `words.txt`.
However, trying to submit the job to `flink run` will fail without
question, because `flink run` expects an absolute file path prefix-ed with
a `file://`, for example, `file:///tmp/input/words.txt`.

Since we will need to submit our job implementation to run on prod anyway,
I recommend disallowing relative paths such as `words.txt`. This will save
users from spending hours to figure out "Why I can `python` run this job,
while `flink run` fails?"

Look forward to your feedback.

Best,
Yik San