To install you need latest node.js
and npm
.
Node is only used to organize the code locally and push it to your couch. It's not needed on your Couch instance.
First clone this repo.
Then install the couchapp
dependency with npm
:
npm install couchapp
And then simply push this source code to your couchdb instance:
couchapp push app.js http://yourcouch.com/dbname
Remember to look into the config.js
file so you can change certain words to fit your app needs.
Such as the title of your app, and you can even translate other things into your own language.
This is basically a Hacker News clone for CouchDB. As a Couchapp. I figured it would be awesome to have a Couchapp act as an entire news application such as HN. This would enable anyone to easily host their own HN site, with their own topic.
Here's what an item
looks like:
{
"_id": "784289f1ac926d7d9ab85b3a22005546",
"_rev": "5-7d72e6583ddd567d700a242618c3ce6f",
"created_at": "2012-05-20T21:09:52.045Z",
"author": "test",
"title": "Nottingam",
"url": "http://dfsf.com",
"type": "item",
"voted": [
"test"
],
"comments": [
{
"comment_id": "2012-05-21T07:43:01.410Z",
"parent_id": 0,
"text": "Ciao Bella",
"voted": [
],
"author": "test"
},
{
"comment_id": "2012-05-21T07:48:50.963Z",
"parent_id": "2012-05-21T07:43:01.410Z",
"text": "My born super",
"voted": [
],
"author": "test"
}
]
}
But what about document conflicts? Easy, I deal with document conflicts on the Browser. Resend the request if it failed. Simple and relaxed approach.
Each item
has a voted
property which is an array of all the users
that have "upvoted" that item. In my all
view I calculate the score
based on the number of items in the voted array (which I call points
) and the created_at
property.
Then I simply emit()
this score as the key of my view, which is then naturally ordered. This is the heart of the ranking algorithm and the order of the front-page, as well as comments.
Couch does only the very necessary. The rest is left to the browser to do. The hierarchy of the comments for example is done by the browser. However, everything is returned nicely by couch so that search engines can crawl it easily.
The HTML returned is also the same as the one on Hacker News. This enables the re-use of already built crawlers and add-ons.
CouchDB is not very good in serving static content because it doesn't allow you to modify the Cache-Control property to tell the browser to not fetch the content over and over. So an easy fix is to simply host the assets (things like images, .css, .js files etc.) on a CDN. A good one is CloudFare.
The only asset that you need to keep an eye on is site.js
. It has all the logic for the frontend of the app. So be sure to
update it from this repo, whenever an update is available.
If you want to install this app yourself, I strongly advise using the great IrisCouch hosting service. It's free to start with and comes with a great pricing plan. This is where hckr.it is hosted.
Simply click on the big "Sign Up Now" button on their homepage and create your own Couch instance. You should get a http://foo.iriscouch.com domain. So to put hckr.it on it you can do:
couchapp push app.js http://username:[email protected]/dbname
With these type of systems where there's voting involved, spammers are always an issue. This Couchapp enforces very stringent rules and allows users to only upvote once. However, someone could take the time and create new accounts and use these new accounts to upvote several items, therefore resulting in multiple votes coming from a single person. This should be avoided, since voting is at the heart of this application.
There might be several ways of solving this issue:
This also is not very accurate as changing IPs is a very easy thing to do, by using proxies and such.
If we make it harder for users to create new accounts - need to validate the email address, for example - this will annoy spammers greatly. Spammers will need to provide a valid email for each of the accounts they create. But, a committed spammer will surely find many email addresses and create many different accounts and therefore will continue spamming the system.
This is indeed what will work at the end of the day. Sites such as StackOverflow and even HN itself have a little "flag" link on each of their post. The community decides whether a post is coming from a spammer or not.