libs/textwrap/no_line_continuations

[thatch]

Finds cases like

textwrap.dedent("""
x \
y
""")

which can't be safely reindented without changing their final value.

[nedbat]

I'm not sure what's going on, but when I try to apply this to coverage, it says NEEDS_WORK, but --patch doesn't show a patch and --apply makes no changes:

% git checkout 6bb59f4efa80a4b30ddbe8d958bf1748421dc0c7
HEAD is now at 6bb59f4e chore: bump actions/download-artifact in the action-dependencies group (#2147)

% ick --rules-repo=https://github.com/codrsquad/rulez@thatch/textwrap-line-continuations run libs/textwrap/no_line_continuations
Running...  [------------------------------------]    0%

-> libs/textwrap/no_line_continuations: NEEDS_WORK
     tests/test_phystokens.py:114: line-continuation backslash in textwrap.dedent string


% ick --rules-repo=https://github.com/codrsquad/rulez@thatch/textwrap-line-continuations run libs/textwrap/no_line_continuations --patch
Running...  [------------------------------------]    0%

-> libs/textwrap/no_line_continuations: NEEDS_WORK
     tests/test_phystokens.py:114: line-continuation backslash in textwrap.dedent string


% ick --rules-repo=https://github.com/codrsquad/rulez@thatch/textwrap-line-continuations run libs/textwrap/no_line_continuations --apply
Running...  [------------------------------------]    0%

-> libs/textwrap/no_line_continuations: NEEDS_WORK
     tests/test_phystokens.py:114: line-continuation backslash in textwrap.dedent string


% git status
HEAD detached at 6bb59f4e
nothing to commit, working tree clean

% bat -r 110:120 tests/test_phystokens.py
  ─────┬──────────────────────
   110 │     def test_1828(self) -> None:
   111 │         # https://github.com/coveragepy/coveragepy/pull/1828
   112 │         tokens = list(
   113 │             source_token_lines(
   114 │                 textwrap.dedent("""
   115 │             x = \
   116 │                 1
   117 │             a = ["aaa",\\
   118 │                  "bbb \\
   119 │                  ccc"]
   120 │             """)
  ─────┴───────────────────────

[thatch]

That's expected, it prints a message and exits 99 on affected files. I couldn't figure out an autofix I was happy with -- getting rid of the continuation but keeping the same (weird) value didn't seem good, and doubling the backslash to do what the user probably intended isn't really better because it will probably break behavior you were relying on.

[nedbat]

Maybe we should have ick print a message then: "No fix available"

[thatch]

NEEDS_WORK is supposed to include both "no autofix" and "incomplete autofix" -- as an example, we might have one batch exit 0 and one batch exit 99 -- "human, take a look" is what you need to do because it's not fully automatic. Fingers crossed this is just a wordsmithing problem?

We could potentially add another status that's just "unsafe fix" if that's more user-friendly to retry with another flag.

[nedbat]

Maybe instead of status codes we should piggyback on structured metadata. It would definitely help to clearly indicate that the rules wants a change but cannot make the change.