sedataktas / story

Storyly

• For installation, you have to create conf.yml file. This file stores database credentials. I created story-conf-sample.yml file as a sample. You can use that file. After create the config file, you have to change viper.SetConfigName("story-conf-sample") and viper.AddConfigPath("/Users/sedat/go/src/story") in config.go and:

  go run main.go

• Application listen port 8080

• I sed gin for handling and routing. You can check from here: https://github.com/gin-gonic/gin

• For getting to .yml, viper was used. You can check from here: https://github.com/spf13/viper

• I used Postgresql for a relational database. I used gorm for connection and sql commands. You can check form here : https://github.com/go-gorm/gorm

• I created 3 tables named Events, Tokens and Stories.

• Events table stores : app_id, story_id,type and count.

• Tokens table stores : app_id and token.

• Stories table stores : app_id,story_id and metadata.

• For logging to errors, logrus was used because it provides sentry implementation easily. You can check it from here : https://github.com/sirupsen/logrus

• APIs return StatusInternalServer error, If there is a server related problem.

• APIs return StatusNotFound error, If record is not found.

• Added comments for exportable functions.

• I used go.mod for dependency management.

First Assignment

• In this assignment you can send GET request with app_token as parameter and get stories information about application.

• API :

  Method : GET

  Path : $ curl "http://localhost:8080/stories/token_1"

  Response :

{
    "app_id": 1,
    "ts": 1612638478,
    "metadata": [
        {
            "id": 1,
            "metadata": {
                "img": "image1.png"
            }
        },
        {
            "id": 2,
            "metadata": {
                "img": "image2.png"
            }
        },
        {
            "id": 3,
            "metadata": {
                "img": "image3.png"
            }
        }
    ]
}

Second Assignment

• Firstly, I used gin. Because it provides some convenience. But then I used native net/http package for better performance.

• I used this command for Apache ab tool : ab -k -n 100000 -c 100 http://localhost:8080/stories/token_1

• • k : keep alive, don't close connection
• • n 100000 : I send 100000 requests totally
• • c 100 : concurrency
• • with this command, 100 user sends 100000 request

• Gin Results :

• Again, at the beginning, I reach to db 2 times. First to check app token exists. Second is to take metadatas. I also check is app token exist in second query but with first query I can return to user, app_token not found information. For improving performance, I waive this so now I reach to db 1 time.

• I also created index for token field in tokens table because I always query to token.

• net/http, 1 reach to db and with app_token index results:

• Additionally, if there are not many applications, tokens can be store memory.

Third Assignment

• In this assignment api take POST request if there is a record matching the values in request body, record's count increments 1. If there is no record matching the values in request body, creates new record and set count value 1.

• If there is no record matching to story_id, app_id and event type, returns error.

• For type field in events table, valdiator tag can be use(impression/close ...)

• API :

  Method : POST

  Path : $ curl "http://localhost:8080/event/token_1"

  Request Body :

{
    "story_id": 1,
    "event_type": "impression"
}