quick scan through lobsters's source code

let’s dive deep into the source to see what’s going on there.

a few links one could find helpful:

• github.com/lobsters/lobsters
• lobste.rs
• lobste.rs/about
• db/schema.rb
• Gemfile
• config/routes.rb

here is some interesting bits that I found(to the best of my understanding)

• replies on stories with negative comments are not shown ( source ) (in your /replies)
• usernames are limited to 25 characters, alphanumeric start + (alphanumeric _ -) (ascii) ( source )
• there is a list of banned usernames like admin/moderator/root/etc. :) ( source )
• 7 days for “new” users, 10 karma to suggest title/tags, 50 karma to downvote, -4 karma to submit stories/invitations, 51% of posts from you = “heavy self promoter” (at least 2 stories), ( source )
• markdown is a custom parser based on nokogiri on top of commonmarker that follows the CommonMark spec + github flavored markdown ( source ):
  • h1, h2, etc. are converted to bold (so good)
  • images to links for images (except moderators :) ( source ) (alt of the image is used for the text of the link if provided)
  • nofollow for links
  • @username are auto-linked to the user’s profile on lobsters
• IP addresses (linked to user id/username/url) are logged on login, sending invites, and when fetching urls (sent in the User-Agent header) ( source ) (also on upvotes it seems ( source ))
  • This is the Internet, you have no privacy.
• there is a bunch of mod-only urls (e.g. /mod*) (on top of all the mod actions obviously) ( source )
• if you post “this!” as a comment, you get a “nope!” ( source )
  • tldr -> “wow! a blue car!”
  • me too -> plz upvote parent
• plaintext comments/stories may be coming ( source )
• it got something to find people with consistently downvoted comments ( source )
• users become inactive and can disown past comments source
• story.rb :
  • does not accept all those annoying tracking domains (t.co, bit.ly, etc.)
  • 14 days to downvote, -5 score is the minimum, 6 hours to edit, 30 days before re-submitting, 2 users to make suggestion live, hot stories for 22 hours
  • drops a bunch of words like a/an/and/but/in/of/etc. when making url slugs
  • no emojis in titles :sad: (:joy:)
• there was a “Low Quality” reason for story downvotes that got removed ( source )

thanks to pushcx, jcs and all the current and past contributors for the amazing work BTW, and to everyone in the lobsters community, quite an amazing prowess.

what I was actually looking for

• the comments about “enabling postgresql” are a lie :) (schema + search depends on mysql at least)
• all assets are through the asset pipeline
• Gemfile is pretty minimalistic, nothing fancy except the rotp (2FA code) and rqrcode (QR code) gems used for security
  • ruumba , rubocop linting for erb, quite cool
• activity has been quite stable and it has been open to contributors for a quite long time
• issues/PRs are managed (e.g. not abandoned/ignored like most projects)
  • “Issues and PRs are typically reviewed Wednesday and some Thursday mornings”
  • “Feature requests are no longer being accepted, unless in the form of a pull request with code.”
• uses jquery/select2 and a bit of JS, nothing crazy, I would say “pragmatic” (no turbolinks)
• long files instead of multiple files are preferred in most cases
• no spring/bootsnap it seems (quite a good idea as it’s the source of 90% of my rails dev experience frustration)
• follows the “small controllers, fat models” mantra
• erb for templates
• everything is very minimalistic, pragmatic, simple in general
• looks like some BSD/MIT license but doesn’t have a name (I think it’s a 3 licenses thing)
• processing done with cron ( source )
• has ansible and docker setups
  • mariadb actually :)
  • on xen on prgmr.com
  • postfix for mails, nginx, tarsnap, unicorn, a funny vimrc
• ruby 2.3 (just got EOL)
  • most gems are close to latest
  • bundle-audit, nothing
  • brakeman has a lot of warnings (mostly false positive, might be worth checking them out though)
• the validate do ... end and ... && errors.add styles are new to me and quite cool I think link
• some html templating in a model, kind of weird .
• raise "too many hash collisions" best error yet :)
• seems like sqlite was also supported at some point source
• formatting is pretty consistent with community guidelines (e.g. it’s using rubocop + pragmatic spacing)
• consistent hash rocket syntax, a little pet peeve of mine :)
• search is based on some MATCH (mysql-specific)
• no redis cache / in-memory cache (except memoizing) (e.g. no Rails.cache)
• has_secure_password (bcryt)

having another web application that is still active and which source code I can almost read in ~2 hours is quite an amazing feat in itself.

(where is my “deploy to heroku (and pay $Xk/month)” button :D)

(never thought of having a comma in my directory names)

(one could think I would contribute, I may actually, but I kind of don’t want to deal with mariadb/mysql (or docker/remote server with ansible))

source of this post on github