Deezer project: Tag app

Exercise for an internship. It uses Neo4j, Vue.js and Express

Running the app

Run the following commands in the main folder.

For the server:

> cd tag-server
> docker build -t tag-sever .
> docker run -p 8081:8081 tag-server

For the front:

> cd tag-front
> docker build -t tag-front .
> docker run -p 8080:8080 tag-front

The app is then available at http://localhost:8080/#/track or http://0.0.0.0:8080/#/track

While the API listens on port 8081.

The app can also be used without docker by running npm install and npm start in both ./tag-server and ./tag-front.

Accessing the db

Informations

As this is only a demo, I chose to conveniently host the data base (for free) on graphendb which means it is:

• Asleep when inactive so the very first connexion is very slow.
• Hosted in Northern Virginia (US) which leads somewhat slow response time.
• Limited to 1k nodes
• Limited to 10k relations

Visualization

The db can be visualized here using these credentials

• login: [email protected]
• password: dzrtagdemopwd

Log in, then go at the bottom of the page, Tools: Neo4j browser, click Launch.

Server side use

DB app credentials

• username: server-user
• password: b.TIFIvn8dL3qz.ewj9o7s3m2Xs7iHk
• connexion string (BOLT): bolt://hobby-ecmbfbmekcnhgbkeephhacal.dbs.graphenedb.com:24786

Overview

As there is a lot of N:M relations and high potential for redundancy, the natural choice is graph databases.

This could also facilitate the implementation of more complex request later on.

Relations between artists, albums and tracks are not implemented in this demo.

I chose Neo4j as it the most well known graph DB and seemed to be the easiest to use. However, Orientdb seems to be more suitable for alrge scale projects.

Possibilities for future

• Add a weight on the TAGS relations to quantify how rock a song is.
• Implement the missing relations between artists, albums and tracks.

Stack

Neo4j + Express +Node.js + Vue.js

• Server API
• Front that can talk to the server and to deezer's API.

Front functionalities

• Add and delete content by ids
• Search for content given a set of tags
• Get more infos on clicked content and edit its tags
• Edited tags will only be sent to the db when Submit is clicked

Server functionalities actually used in the front

• GET tagged content
• POST create content
• POST replace content
• DELETE content

Server functionalities

Add tags to content

POST /artist/123 or POST /track/456 or POST /album/789

Request body (tag list) ["tag_a", "tag_b", ...]

Replace content

Replaces the tag list by a new one

/!\ suboptimal implementation

POST /artist/123/replace or POST /track/456/replace

Request body (tag list) ["tag_a", "tag_b", ...]

Get tags associated with content

GET /artist/123

Response body (list of tags) ["tag_a", "tag_b", ...]

Get content list from a tag set

Returns the content that has all the requested tags

GET /album?tags[]=tag_a&tags[]=tag_b or GET /track?tags[]=... or GET /artist?tags[]=...

Response body (list of ids and tags)

[
    { id: 123 tags: [tag_a, tag_b, tag_c...] },
    { id: 456 tags: [tag_a, tag_b,...] },
    ...
]

Export all the tagged content

Returns one json object per line.

GET /export

Response body

{"type":"artist", "id": 123, "tags":["tag_a", "tag_b", ...]}
{"type":"track",  "id": 456, "tags":["tag_c", "tag_d", ...]}
{"type":"album",  "id": 789, "tags":["tag_e", "tag_f", ...]}
...

Delete tags

DELETE /artist/123/tag

Request body (tag list) ["tag_a", "tag_b", ...]

Delete content

DELETE /artist/123

Why no names/titles in the table ?

Limited to 50 requests per 30s to deezer's server + the current architecture would be very inefficient.