Skip to content
DevMeme
7178 of 7590
CodeReviews Post #7872 · source on Telegram

Code Review Diff Stats: +534,441 Additions Spark Reviewer Concern

Description

A dark-mode screenshot of a GitHub pull request comment by user 'kinlay0' posted on Jan 4. The comment shows a diff statistics widget displaying '+534,441' in green and '-46' in red, accompanied by five diff-ratio squares (four green, one dark). Below a divider, the reviewer writes 'I have a lot of questions for you', followed by emoji reactions: a grinning face with 29 reactions, a confused face with 2, and a slightly smiling face with 27. The humor lies in the absurdly massive pull request - over half a million lines added with only 46 removed - and the reviewer's deadpan understatement, a scenario painfully familiar to anyone who has been asked to review an unreviewable mega-PR, likely containing committed dependencies, generated code, or an entire vendored project

Comments

19
Anonymous ★ Top Pick Per the well-known law of code review: a 10-line PR gets 10 comments, a 534,441-line PR gets 'LGTM' - this reviewer is bravely violating physics
  1. Anonymous ★ Top Pick

    Per the well-known law of code review: a 10-line PR gets 10 comments, a 534,441-line PR gets 'LGTM' - this reviewer is bravely violating physics

  2. Anonymous

    At this scale, `LGTM` is no longer approval; it is a timeout.

  3. @drbogar 4mo

    lgtm

  4. @deadgnom32 4mo

    echo venv >>.gitignore

  5. @evorobyov 4mo

    "Small fixes"

  6. @peajack 4mo

    average chinese dev

  7. @sukhrob_ikromov 4mo

    ruff check --fix

  8. @ArtemVoikov 4mo

    experience doesn't matter anymore it seems. no one would ever be able to follow the code changes at this pace

    1. @death_by_oom 4mo

      Imo any pr with more then 200 lines of actual logic code should be rejected and split up

      1. @tema3210 4mo

        Only if you're really comitted to squashing on merge and representing a single logical step as commit. Isn't it easier that all features get their own pulls?

        1. @death_by_oom 4mo

          Reviewing a commit that is longer then 200 lines of logic too hard, because you need to keep all of this in your brain while you are reviewing. Unlike the person who actually wrote the commit you need to do this quite fast, making commits bigger the 200 lines of logic is just bad for the review process more bugs will go through

        2. @death_by_oom 4mo

          Just squash on merge to main, so your main commit history is clean

      2. @purplesyringa 4mo

        that's just not realistic though maybe you can split it into multiple commits in a PR, but often not into different PRs

        1. @death_by_oom 4mo

          I misspoke here, I mean commits. But commit is a complete unit of work, so you can review it as a standalone piece of work

          1. @tema3210 4mo

            Ideally, yes, but practically? Not so much in my experience

            1. @death_by_oom 4mo

              It isn't very often that it's not possible to break up a 300+ line commit into more manageable logical chunks

      3. @sysoevyarik 4mo

        This 500kloc commit surely contains no more than "200 lines of actual logic". It's just enterprise quality what makes it look bigger

        1. @TheFloofyFloof 4mo

          "Enterprise Quality" So its all bloat with 5 lines of actual code

          1. @sysoevyarik 4mo

            Doubt that "enterprise quality" is that bad. In place where I work ratio is near 300 lines / 1 line of payload. Maybe we are not a true enterprise...

Use J and K for navigation