If I missed an unquoted string (most likely within a list or as a sub field) please let me know as well, and I will quote it. But the goal of this PR isn't necessarily 100% coverage, it's 99%+ to help others form good habits / avoid bad habits. This PR is probably too huge to aim for completely perfection in a realistic time frame.

Nice! Since this diff is enormous and you're willing/prepared to handle merge conflicts, I'm assuming it's done by a script? It might be useful to publish the script on a gist as well, so that it can be reviewed alongside the actual diff.

Actually, it was done by hand with the aid of grep and vim macros. I didn't trust a script enough to do the proper thing without reviewing every point it changed anyway. And since I wanted at least one pair of human eyes on every change I figured I might as well just process it by hand. It took about 8 hours. (And I didn't think that there was a handy nix lexer I could just hook into easily, and I didn't fancy writing one.)

For the record, my grep call was: vim $(egrep '^\s*[A-Za-z0-9.]+\s*=\s*\w+://' * -Rl)

And my macro was roughly: /^\s*[\w.]\+\s*=\s*\w\+:\/\/<Enter>f=wi"<Esc>/\s*;\s*\(\#.*\)*<Enter>i"<Esc>

However that macro underwent a few refinements as I found failure modes. The following failure modes for the various versions of the macro are/were:

1. fails when the semicolon is followed by an inline comment, it puts the ending quote on the next line instead. This was double checked for at the end, no are present to my knowledge. (Old only)
2. It quotes a trailing space if there is a space between the URL and the semicolon. This is more difficult to search for, but an attempt was made to find any examples that I missed. (Old only)
3. Does not find all unquoted URL strings. And I'm ok with that, though I have tried to track down other types as well. (All versions)

EDIT: @bhipple thanks for the question. It gave me an opportunity to think about failure mode 2 a little bit harder, and I was able to come up with a better way of searching for missed cases and found 6 strings that were improperly quoted. They have been fixed with the latest force push.

EDIT2: Additionally, I am a programmer. @a<leader>n My job is very important and serious @@<leader>n I have very high standards for myself and my own professionalism @@@@@@@@<leader>n And I contribute to open source projects during the quarantine 20@a<leader>n using only the best and most sophisticated techniques. <right arrow><right arrow><right arrow><right arrow><right arrow>i"<esc><mouse over and click>i"<esc>

One conflict for today, renamed commit per @bhipple.

Fixed 23 packages this morning. That's more like it!

3 due to a removed patches
4 due to updating adjacent metadata
8 generated files updated (generated with bareURLs)
3 files removed entirely
3 due to marking package as broken or partially broken
1 due to the URL directly changing
1 due to whitespace differences (indenting fixed)

Additionally, 5 pkgs that were either just added or previously correct had bare URLs introduced, they were also fixed.

Three conflicts fixed today. Fixed one new bare URL in the new commits.

Question. Would you like me to restructure this commit into smaller batch commits (like 1000 apiece) and/or keep the daily merge fixes as separate commits, then rebase at the end? Will that help with the review process? @bhipple @grahamc @FRidh ?

@GrahamcOfBorg eval

I'm not sure, to be honest. Such a large change is ... well, a very large change. Even broken up in to 15 PRs, it is quite significant. I'm not sure who will have the guts to press the merge button :) Something like this and #84122 I can only really imagine happening nicely over time, gradually converging to the "correct" format. I am just one person, though.

Fixed 6 conflicts:

5 changed homepage URLs (3 committed as bare URLs)
1 changed description

Fixed 1 newly pushed bare URLs.

However, my + and - numbers have diverged. So I'm attempting to figure out why.

EDIT: that's sorted.

Re: who could review this — @infinisil by comparing with the automated version, I guess.

Thanks a lot for detailed and frequent reports on merge conflicts. With just the automated tool, we were never sure how much merge churn there is

Fixed 7 conflicts:

1 whitespace change in adjacent line
3 files entirely deleted
1 indentation adjustment
2 URLs changed

No new bare URLs were committed today.

Fixed 8 conflicts: (Total commit size: 15410 -> 15387)

2 due to updating adjacent metadata
4 changed URLs
2 due to simplified nix expressions on adjacent lines (cleanup)

Fixed 3 conflicts (Total commit size 15387 -> 15373)

2 files removed entirely
1 changed URL

If you send a PR with just the lib/ and nixos/ subdirs, I can review and merge that; then we're down to just the huge pkgs/ subdir (which the bot is slowly reducing in scope, at least). For pkgs, we have OfBorg telling us that it's 0 rebuilds, at least.

I splitted the diff of aaf2b1e into smaller 10000 lines chunks:

• splitai.txt
• splitah.txt
• splitag.txt
• splitaf.txt
• splitae.txt
• splitad.txt
• splitac.txt
• splitab.txt
• splitaa.txt

Pick a random one and post if you have reviewed it. Best viewed with: nix-shell -p bat and bat -l diff ./splitaa.txt

Cherry-picked into 84cf00f
I had to fix the merge conflict with doctl.

Thanks!

🥳