Who the hell are you?

What the hell is this?

In the beginning.

There was SQL.

And it was okay.

And then...

There was noSQL!

And it scaled and it was all javascripty and all the hipsters loved it and everyone was happy.

But then you want to actually use it

var request = http.request({
  host: 'couchdb.internet.com',
  path: '/awesomedb/_all_docs',
  auth: 'zer0c00l:God'
});
request.end();

var newrows = []
request.on('response', function (response) {
  response.on('data', function (data) {
    data.rows.forEach(function(row){
      row.value.walkThe = "dinosaur";
      newrows.push(row.value);
    });
  });
});
request.end();

var request = http.request({
  host:   'couchdb.internet.com',
  method: 'post'
  path:   '/awesomedb/_bulk_docs',
  auth:   'zer0c00l:God'
});

request.write(JSON.stringify({docs:newrows}));

request.end()

unql-node

https://github.com/sgentle/unql-node

It's like SQL for NoSQL

Deletin' stuff

> delete from checkins where username.match(/\d{4}$/) && timestamp > (new Date()).getTime()-1000*60*60*24*7
						

Why don't we take all the data...

> create collection nag_old_users
> insert into nag_old_users select {username: username, email: email} from user where last_login < (new Date()).getTime()-1000*60*60*24*7
						

And put it over here!

Want to see my awesome LALR recursive descent parser?

Nope!

handle /delete from (\S+)(?: where (.*))?/, (db, expr, cb) ->
  query = "update #{db} set _deleted = true"
  query += " where #{expr}" if expr
  processExpr query, cb
						

There I fixed it.

noDB

So you want to add some persistence to your simple web app.

Okay so just require that each user have a couch installation.

Or make redis a dependency. Probably include some build scripts or something to make it easier since it's the only thing in the app that needs to be compiled.

Maybe it would be easier to build SQLite?

Nope!

noDB

SQL < noSQL < noDB

Super simple

var users = JSON.parse(fs.readFileSync('data/users.json'))

users.push({name: "newguy"});
fs.writeFileSync('data/users.json', JSON.stringify(users));
res.send('success');
						

Session storage too!

var store = new express.session.MemoryStore;
store.sessions = JSON.parse(fs.readFileSync('data/sessions.json'));

setInterval(function(){
  fs.writeFileSync('data/sessions.json', JSON.stringify(store.sessions));
}, 2000);
						

It's basically Redis

But it is:

• Deployable
• Dependency free
• Easily backed up
• Debuggable

Databases: Fixed

I would like to make some information available on the internet.

So find a shared hosting provider or something.

But they're kind of crappy

So I'll just run my own

On nginx or express or something!

But load balancing and updates and bugs and oh no.

Oh I know!

Augh. I wish I could just shove all this right on Akamai

...

...

...

Why not?

Step 0: Generate static site

Use whatever you want. We really like Jade.

!!! 5
include includes/head.jade
body
  include includes/header.jade
  // BEGIN Content 
  .inner-bg
  #content-wrapper
    Welcome to zombocom!
						

make

JADE = $(shell find pages/*.jade)
HTML = $(JADE:.jade=.html)

all: $(HTML)

%.html: %.jade
      ./jade/bin/jade < $< --path $< > $@

clean:
      rm -f $(HTML)

.PHONY: clean
						

Rackspace == Akamai

It's cheap. 18c per GB

You don't have to deal directly with Akamai. Keep your soul!

http://www.rackspace.com/cloud/public/files/

Getting it there

There's a node library for that.

https://github.com/nodejitsu/node-cloudfiles

There's also some tooling for that.

https://github.com/PinionTech/cloud-loader

coffee cloud-loader.coffee -x localDir cloudContainer
						

Deep magic

curl -i -X GET -H "X-Auth-User: pinion" -H "X-Auth-Key: XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX" "https://auth.api.rackspacecloud.com/v1.0"

curl -X POST -H "X-Container-Meta-Web-Index: index.html" -H "X-Auth-Token: XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX" "https://storage101.dfw1.clouddrive.com/v1/MossoCloudFS_ad499859-aaed-43c6-813b-5e1c545f85cf/pinion-www
						

Relax

Never ever care about whether your website is up.

Go do something useful instead.

Yeah, it's harder to do pretty URLs.

So fixed

It's not always the right thing to just fix it

But sometimes it is