fix: render docstrings for class attributes and constants

[souradeep-das]

What was wrong?

#9

How was it fixed?

• As the cst nodes are traversed, assign the statements index position as the child offset
  Since, the child_offset for the cst node is incremented both during enter_class_def and after visiting the node at exit(), incrementing on every iteration sometimes incorrectly assigns different child offset indexes (skipped by 2)

• while traversing cst, do not revisit docstring lines as statements (since they are attached to their parent attribute)

• For the class body, the cst children starts with a Index 0: TrailingWhitespace(), while for modules (outside classes) index starts with constants - selectively handle that when determining the offset of the docstring

Testing

Verified against different class structures and conditions

Sample Class structures and results
bytes.py (in src_docc).pdf

[SamWilsn]

If I'm reading this correctly, this detects any string alone on a line? I suspect you tested it, but this works for multiline strings, and for strings that aren't the first child? Like:

def foo():
    a = 1
    "String"
    return a

[souradeep-das]

so for functions the functionality is unchanged. For class definitions it would still take the very next docstring line after the attribute (also multilines). Dangling strings not attached to class attributes or class descriptions wont show up.

I have attached some test outputs here (also with functions inside classes), which covers these cases: #18 (comment)

do lmk if something is off, or needs to be handled.

[SamWilsn]

Is this a stylistic change, or is this important for fixing the issue? If it's the latter, can you briefly explain what went wrong here? When would += 1 be different from idx?

Even if it's just stylistic, it's definitely an improvement.

[souradeep-das]

I noticed that the child_offset was being incremented also at exit() L843, which resulted in statements being assigned offsets in double increments (multiples of 2). so used idx to assign it to its true position

[SamWilsn]

Hm, interesting. Again, there's a good chance I'm wrong here, but it looks like enter_class_def returns SkipChildren, so do the children of the class definition actually get their exit called?

[souradeep-das]

I think the exit() for simpleStatementLine is still called since we do a statement.visit(self) on L573

I can see offsets being incremented after visiting every child

..
child_offset=8: SimpleStatementLine(
    body=[
        AnnAssign(
            target=Name(
                value='parent3_hash',
...
child_offset incremented to 9 after exiting CstNode(SimpleStatementLine(...), start=12:4, end=12:24)
  Source line: parent3_hash: Hash32
  
child_offset=10: SimpleStatementLine(
    body=[
        Expr(
            value=SimpleString(
                value='"""Number of bytes3 in each instance of this class."""',
..