Hacker News Reader

Hacker News

Comment by chillee | Hacker News Reader

Full thread

chillee

I agree that the spam is problematic, but I feel the need to call out his "data collection".

As evidence that CodeWithHarry is the cause, he says

> A search for "Amazing Project" is now showing 21,177 issues.

and

> A search for "improve docs" shows 319,251 issues.

???

Improve docs has been a common beginner level PR since the beginning of time. Do you really think that 319k spam issues have been filed in the last day?

Amazing Project is more relevant, but if you take a look at those issues (sorted by newest), you see that most of the PRs that show aren't spam at all: https://github.com/search?o=desc&q=amazing+project&s=created...

If you want to count how many spam PRs were directly inspired by the video, you can try searching "an amazing project" (with quotes), and filtering by newest. It looks like there have only been about 50 PRs directly inspired by it.

https://github.com/search?o=desc&q=%22an+amazing+project%22&...

View on HN

Github allows you to search by dates (newer than, date ranges, etc), so for example, you can add the query "created:>=2020-09-30" (the date the youtube video was uploaded) to get a more accurate list.

From this search, it looks like it's definitely on the order of hundreds:

https://github.com/search?p=5&q=amazing+project+created%3A%3...

I opened a few at random to spot-check, and they all seemed like garbage to me. Like this one, ugh https://github.com/mongodb/docs/pull/4402/files

Sorting by relevance doesn't get you a representative sample - that bubbles PRs who say "amazing project" in the title to the top(which are indeed mostly spam).

If you restrict it to "in:title", you see that there's only 44: https://github.com/search?q=in%3Atitle+amazing+project+creat...

And if you sort by "newest" or "oldest", which provide a more representative sample, you can see that most of them are not spam: https://github.com/search?o=desc&q=amazing+project+created%3...

Why restrict in title though? From parent reply it is obvious that the phrase isn't in title yet it is undeniably spam.

I'm trying to get a lower bound on spam PRs filed. If you sort by oldest (instead of by "relevance" as the parent commenter did) and do a spot check, you'll see that a lot of the PR's are legitimate: https://github.com/search?o=asc&q=amazing+project+created%3A...

Certainly, restricting to the title misses plenty of PRs. But his search provides 252 PRs, and I'd estimate that easily <100 of them are spam.

Typically the PRs are titled something like "Improve docs" and the change is to add "an amazing project" somewhere in the README.md.

Here's a search that finds a lot: https://github.com/search?q=in%3Atitle+improve+created%3A%3E...

This search makes the same mistake I mentioned above - you think it finds a lot because you're sorting by "relevant" and sampling.

Sort by newest and you'll see that the vast majority of these PR's are legitimate: https://github.com/search?o=desc&q=in%3Atitle+improve+create...

The results I got show 4 spam PRs out of 10 on the first page. Same on the second and third pages. 60% is not a vast majority legitimate and 40% spam is not inconsequential when it's 86k PRs.

That doesn't match my experience. I looked at the first and 5th page from my link. No spam PRs.

https://i.imgur.com/VgOteKy.png

https://i.imgur.com/aAZ8xbo.jpg

Presumably the spam onslaught is slowing down with time from the video being posted so looking at the most recent results will give an ever fewer percentage of spam. If you look at page 100 of the results (for example, currently 18 hours old) there's a lot more spam.

For example: https://imgur.com/a/vxCHvcO

It is so indiscriminate that repos that could be considered dead (7+ years without activity) by the maintainer seems to get massive attention from new accounts just to get that T-Shirt. O_O

Love this one — somebody is trying to rename "package.json" into "package":

https://github.com/dinosaurjs/website/pull/28

No, he renamed it to "package: An Amazing Project.json"

OMG even better :-))

My mind did not parse anything after ":"

Windows doesn't even let you have a colon in a filename!

Had a hell of a time downloading a torrent with a colon in the name.

You mean it was a pain in the colon? :)

Wonder if NTFS 'lets' you. I know the API does not let it. But I wonder if you could get it in a different way. I know you can sneak case sensitivity in here and there.

Think I shall go ruin a VM tonight... :)

I suppose it is because the colon is used for NTFS alternate data streams, see the example in the introduction here: https://docs.microsoft.com/en-us/sysinternals/downloads/stre...

This is my favourite trash PR

Not even that, the file is actually renamed "package: An Amazing Project.json". Amazing indeed!

My favorite one I saw renamed "README.md" to "My Reading".

Link, please :)

What’s this bitboxx? Somebody decided the spam PRs weren’t enough spam so they created a bot to amplify it?

"An amazing project!" is going to part of dev culture for years, at least some circles.

You can search for all PRs by a user: https://github.com/pulls?q=is%3Apr+author%3Asalimjaved18

A lot of those accounts also seem to be created just a few hours before October 1st. Can GitHub time gate new accounts before that can make public PRs?

I feel like that would be counterproductive given that GitHub is a tool, not a social network you need to gather account age or fake internet points on to use. Don't really see anything wrong with somebody signing up and immediately creating a PR.

I realize GitHub is one of the most popular choices out there but the activity stream is already social mediaesque enough for my taste. There's other ways to tackle this, DO could make their marketing stunt opt-in and GH could respond to the people wanting to add friction to their repo's workflows in a way that provides value, without crippling the general user experience.

I think this would be a bad direction. I bet that most people create an account on GitHub to report an issue or create a pull request directly. I think this should be encouraged.

I was wondering about all of that. Appreciate the clarification.

Reply on news.ycombinator.com