Please keep the comment position in csgrep output

[praiskup]

See the "Added warnings" comment, it is not assigned to any report:

$ cat /tmp/test.err
# ================
#  Added warnings
# ================

Error: MYPY_ERROR:
copr_backend/constants.py:9: mypy[error]: Need type annotation for "DEF_REPOS" (hint: "DEF_REPOS: List[<type>] = ...")

Error: MYPY_ERROR:
copr_backend/constants.py:13: mypy[error]: Need type annotation for "DEF_MACROS" (hint: "DEF_MACROS: Dict[<type>, <type>] = ...")

Error: MYPY_ERROR:
copr_backend/helpers.py:26: mypy[error]: Library stubs not installed for "pytz" (or incompatible with Python 3.11)

Error: MYPY_NOTE:
copr_backend/helpers.py:26: mypy[note]: Hint: "python3 -m pip install types-pytz"

Error: MYPY_NOTE:
copr_backend/helpers.py:26: mypy[note]: (or run "mypy --install-types" to install all missing stub packages)

Error: MYPY_NOTE:
copr_backend/helpers.py:26: mypy[note]: See https://mypy.readthedocs.io/en/stable/running_mypy.html#missing-imports

Error: MYPY_ERROR:
copr_backend/helpers.py:31: mypy[error]: Library stubs not installed for "redis" (or incompatible with Python 3.11)

Error: MYPY_NOTE:
copr_backend/helpers.py:31: mypy[note]: Hint: "python3 -m pip install types-redis"
i

But csgrep assigns the rport to one of the issues:

$ csgrep -v --path copr_backend/helpers.py < /tmp/test.err
Error: MYPY_ERROR:
# ================
#  Added warnings 
# ================
copr_backend/constants.py:9: mypy[error]: Need type annotation for "DEF_REPOS" (hint: "DEF_REPOS: List[<type>] = ...")

Error: MYPY_ERROR:
copr_backend/constants.py:13: mypy[error]: Need type annotation for "DEF_MACROS" (hint: "DEF_MACROS: Dict[<type>, <type>] = ...")

[kdudka]

This is currently not possible. Each comment belongs to a specific defect in the internal representation, which is shared by all the supported data formats. At the moment, csdiff supports 8 input and 3 output formats.

[praiskup]

I'm not in hurry at all. I just wanted to report this.

[kdudka]

If we allowed comments not "owned" by any defect, how would the plain-text parser distinguish them from the per-defect comments that we use today? Some developers write comments before the defects while others write comments after the defects. Empty lines are intentionally ignored in this format.

[praiskup]

Aaah, you are right, now I see (I have a bad feeling I already reported this before, sorry).

I.e. indeed, there would have to be an additional rule in the format. Like \n\n implying an independent comment (it would be OK if csgrep filtered this out entirely). To me, this still doesn't sound that bad compared to the current behavior - a seemingly error-related comment (might be actually related to a set of errors) is moved down so it looks like a event-related one.

Perhaps we could define three types of comments, generic (separated by multiple newlines), error-related (on top of the ERROR header, single newline-separated) and event-related (on top of the event, single-newline-separated, this way csgrep prints the comment now).

I can imagine this is not a trivial thing to implement though.