m3libea / random-coding-challenge

#Random Coding Challenge

Part 1

1. Create an users table with the following fields and insert 10 million random rows into the table

• id
• full_name
• email
• city

2. Find all the users who have john in their name so John smith, Smith john, Johnny etc should all be returned

3. Optimize query so it doesn’t take more than few milli seconds.

##My solution

I'm using three nodejs scripts to run several tasks (db creation, populating db and performing queries). I'm using several libraries: node-mysql, async and measure. This example is meant to be run in a MySQL server as is taking advantage of a FULLTEXT index to allow fast querying among the records (MyISAM engine).

Every script will provide timing information in milliseconds when they are done.

Instructions

1. cd partOne

2. npm install

   will install all the abovementioned dependencies.

3. npm start

   It will run scripts inside bootstrap folder create_db and populate_db. Populate DB will queue 10000 INSERT tasks in series. Each task will insert 1000 random user records. It will take a while to populate the DB depending on your machine. (As a stretch I would explore parallel db query and see how they could improve exec speed).

4. node index.js [term]

   It will run a pair of queries against the DB. The first query is a slow query using classic LIKE syntax and therefore not taking advantage of the FULLTEXT index. (It's provided as a comparison). The second query it's a fast query using MATCH ... AGAINST in binary mode returning the same result set but with almost a ten-fold improvement in execution time by leveraging a FULLTEXT index.

Better results could be achieved by using Apache Lucerne or Solr as they will allow suffix searches, matching for example 'John' => John Johnson DeJohn. This is a limitation as FULLTEXT indexes use a B-Tree.

Example

$ node partOne/index.js John
[+] Slow query using LIKE. NOT INDEXED
 ->  SELECT * FROM Users WHERE full_name LIKE '%John%'
[+] Slow query stats: { ct: 1, total: 3951, max: 3951, min: 3951 } num rows: 39995
[+] Fast query using MATCH AGAINST. INDEXED
 ->  SELECT * FROM Users WHERE MATCH(full_name) AGAINST('John*' IN BOOLEAN MODE);
39995
[+] Fast query stats: { ct: 1, total: 510, max: 510, min: 510 } num rows: 39995

Part 2

1. Make a chat application using PHP, nodejs/socket.io. It should be simple chat application with

one public channel.

2. Populate chat with 1000 random messages by random users every 5 seconds.

3. Make a list of 100 words. These words should be blocked on public channel.

4. Deploy chat on a server.

##My solution

Here I'm using socket-io with express. When you enter in the chat, you get a prompt to introduce your username. I'm using bower in order to install the frontend dependencies. Then, I'm deploying it on Heroku.

#Instructions

1. cd partTwo

2. npm install

   will install all the dependencies

3. node index.js

   It will set up and start the chat. A script will populate 1000 random messages by random users every 5 seconds. Frontend part delete old messages in order to improve the chat, if you don't do it, you can get tons of messages that will block your client.

#Example

You can see a functional version on Heroku: https://mysterious-citadel-93288.herokuapp.com/