t3rn0
commented
1 year ago Hi. Thanks for creating this PR. I understand that in some cases proposed behavior may be convenient. But I think the current one is ok. Consider following example:
source = """
a = 1; b = 2
"""
tree = ast.parse(source)
for node in tree.body: # len(tree.body) == 2
print(ast.get_source_segment(node))
## prints ##
a = 1
b = 2There are two separate nodes and there are two separate source segments. Now, let's change the b-node to a comment:
source = """
a = 1 # some comment
"""
tree = ast_comments.parse(source)
for node in tree.body: # len(tree.body) == 2
print(ast.get_source_segment(node))
## prints ##
a = 1
# some commentAgain, we have two separate nodes (Expr and Comment) and two separate source segments. I don't see any difference between the first and second examples, except for the ast node type. If a Comment node is part of the body of some other node (function, forloop, etc. ), its source segment must also be part of the corresponding node's source segment:
source = """
def foo(): # some comment
a = 1
"""
tree = ast_comments.parse(source)
for node in tree.body: # len(tree.body) == 1 (!!!)
print(ast.get_source_segment(node))
## prints ##
def foo(): # some comment
a = 1About PR:
From "testing" perspective I find that ast_comments.get_source_segment doesn't actually do anything. If you remove it pytest results won't change.
I recommend creating this modification on your side (in radon).
It's really great you raised this issue. Review helped me to find one minor bug 🙂
devdanzin
I've made some changes to ast-comments while trying to get a version of ast.get_source_segment that roundtrips comments to use in radon.
They allow getting trailing comments from within function bodies. Unfortunately, it'll also incorporate comments that appear right after a function's body. I'm not sure this is of interest to include in ast-comments, but wanted to keep a record of the exploration and allow assessing whether (an improved version of?) this would be desirable.