[CI] Update dummy-variable regex for pylint

[Lunderberg]

Prior to this commit, the regex used for pylint to identify dummy variables would correctly identify variables that start with an underscore (e.g. _scale), unless they have an underscore elsewhere in the name (e.g. _scale_factor). This leads to false positives from pylint for unused variables, as prefixing a variable with an underscore should mark a variable as intentionally unused.

This commit updates the regex in TVM's pylintrc to match the current default value for dummy-variables-rgx, to allow unused variables to be named with a leading underscore, even if they also contain another underscore.

[Lunderberg]

I ran into this false positive for #17201 for a variable named _in_group (screenshot below). While I could work around it by renaming to _ingroup, it's also good to have a long-term fix for the lint rule.

[quic-sanirudh]

Since we're already using names starting with _ to represent dummy variables, it makes sense to allow _ in other places as part of the name.

One thing though is that PEP 8 style guide suggests using names starting with _ to be used as "weak" internal use variables, and the convention I've used till now is that unused variables are named with just an _ like x, _ = get_xy(). So, my question is why do we want to name unused variables in general? Could we not just use _?

[Lunderberg]

So, my question is why do we want to name unused variables in general? Could we not just use _?

Mainly to indicate to a reader which return values are being ignored. As an example, suppose there's a function that produces a

# A reader may not know what the first two return values are.
_, _, attrs = define_function(relax_args)

# The first two return values are a function and parameters to 
# that function.
_func, _params, attrs = define_function(relax_args)

I've run into a few cases where the same return value was re-generated, because a later developer wasn't aware that it was already being generated. By explicitly naming the unused return value, this becomes less likely. (e.g. Seeing _func instead of _, and checking whether it is already the function that should be used.)

Obviously, this shouldn't be applied in all cases, as _, y = get_xy() is already perfectly readable, but it can help in cases where it isn't as clear.

[quic-sanirudh]

So, my question is why do we want to name unused variables in general? Could we not just use _?

Mainly to indicate to a reader which return values are being ignored. As an example, suppose there's a function that produces a

# A reader may not know what the first two return values are.
_, _, attrs = define_function(relax_args)

# The first two return values are a function and parameters to 
# that function.
_func, _params, attrs = define_function(relax_args)

I've run into a few cases where the same return value was re-generated, because a later developer wasn't aware that it was already being generated. By explicitly naming the unused return value, this becomes less likely. (e.g. Seeing _func instead of _, and checking whether it is already the function that should be used.)

Obviously, this shouldn't be applied in all cases, as _, y = get_xy() is already perfectly readable, but it can help in cases where it isn't as clear.

Makes sense, thanks for your reply.