See #44. Basically, with lists and tuples, there is no difference at the bytecode level (originally I assumed that tuples would be faster, which wasn't correct). sets on the other hand will be faster for larger datasets, since lookups are O(1). Because of this, sets emit different bytecode.

To answer your question though, I do think I users should be specify whether they want to use []/()/ {} formatting, but there currently is no way to pass arbitrary data to a check.

To fix this, we could implement something like this for the command line:

refurb file.py --data FURB109.style=list

And something like this for the config file:

[tool.refurb]
# your settings here

[tool.refurb.data.FURB109]
style = "list"

Then, when running the FURB109 check, we would pass in the key values that the user supplied (if any). The check would be free to do what it wants with the data (though we would want to document this for the user).

What are your thoughts?