README.md

Description

Ths microservice is the simplest, but production-ready version of http://old.makala.com :)

Enjoy it!

Developers

In order to start the microservice, you need to run external services with the following command:

make services_up

Start the microservice itself with the command:

make api_up

API will be accessible at the port :8600 by default

About Tests

The microservice is divided into three layers:

• handler or request parser (./server/rest/handler/*.go)
• service or business logic (./service/postfeed/*)
• datastore & cache (./provider/*)

Note: tests in the microservice are to show how I cover my code with tests. I did not intend to cover all the code with tests, because I think this is not necessary at the moment

Unit Tests

Run all

make unit_test

Test for handlers

go test -v ./server/rest/handler/*.go

Test for service

go test -v ./service/postfeed/*.go

Test for providers

Redis

go test -v ./provider/adstore/*.go

PostgreSQL

There is no need to write for PostgreSQL as sqlc generates a type-safe, tested go code.

Integration Tests

Run all

make integration_test

Test for Redis (cache)

go test -v tests/integration/provider/adstore/ads_test.go	
go test -v tests/integration/provider/feedstore/feed_test.go

Test for PostgreSQL (cache)

go test -v tests/integration/provider/poststore/poststore_test.go

Task

Task is to build a microservice that:

1. accepts posts and stores them
2. generates a feed based on the score of each post
3. generates a feed, which contains promoted posts (ads)

Solution

Posts & Ads:

• All posts are stored in PostgreSQL datastore, including promoted ones (ads)
• Feed and ads are cached in Redis (ids of posts are stored in cache)

Feed:

• There is a background worker, which periodically creates a feed in Redis and updates versions
• Feed is stored in Redis in a sorted set. The set is sorted based on the score of each post.
• The number of posts in a feed is limited. You can set a limit (at this moment it is 1000)
• Redis holds two versions of feed 'new' and 'old'.
• 'New' version for users, who started scrolling feed (started with page = 0 or without started_fetching_at_unix_nano_utc in query)
• 'Old' version for users, who were scrolling, while feed gets updated

Feed Update & Background worker:

• Background worker goes through all posts in PostgreSQL & generates a feed with limited number\
• The newly created sorted set, will replace 'new' version of feed, while it replaces 'old' version
• Microservice remembers when user started to scroll feed. This is done based on the feed page (page = 0 is the start of feed scroll)
• Microservice also remembers feed update time
• Example of how feed version is picked:

Let's say a user started scrolling feed at 10:00:00 (hh:mm:ss) If feed was updated at 09:00:00, then a user accesses the 'new' version of feed. Then let's say feed is updated at 10:05:00 then a user will get the 'old' version of a feed.

As at 10:05:00 newly-generated-feed (by background worker) -- copied --> key_feed_version_new -- copied --> key_feed_version_old

Note: it is still not perfect (there should be several versions of feed for those, who scroll for a long time)

User Feed & Ads

• Ads (promoted posts) itself are stored in PostgreSQL, while their ids will be stored in Redis in a list
• Microservice remembers index of ad, which a user has seen. This means users will not miss any of the ads.
• If the index has reached the end, it starts from the beginning. This means that users will receive all ads in a circular fashion

API Reference

You can find documentation of the microservice here

POST /api/makala/v1/post

• there are restrictions that are described in the task (.pdf file)

GET /api/makala/v1/feed?count=10&page=0&author=t2_authorme&started_fetching_at_unix_nano_utc=1746560444234496670

• count - number of posts (default 27)
• page - page number, if it is 0 then started_fetching_at_unix_nano_utc will be set to current time & feed version will be 'new'
• author - at this moment, there is no restriction on the name of an author. This must be taken from token/header/session, but to make you life easier I put it here
• started_fetching_at_unix_nano_utc - time, when a user started to scroll feed. This is used to switch between versions of feed

Requests

Fetch feed 'new' version of feed:

curl --location --request GET 'http://0.0.0.0:8600/api/makala/v1/feed?page=0&count=27&author=t2_author12' \
--header 'Content-Type: application/json; charset=utf-8'

curl --location --request GET 'http://0.0.0.0:8600/api/makala/v1/feed?page=1&count=27&author=t2_author12&started_fetching_at_unix_nano_utc=<current_time>' \
--header 'Content-Type: application/json; charset=utf-8'

Fetch feed 'old' version of feed:

curl --location --request GET 'http://0.0.0.0:8600/api/makala/v1/feed?page=1&count=27&author=t2_author12&started_fetching_at_unix_nano_utc=1146674939827247437' \
--header 'Content-Type: application/json; charset=utf-8'

Create post:

curl --location --request POST 'http://0.0.0.0:8600/api/makala/v1/post' \
--header 'Content-Type: application/json' \
--data-raw '{
    "author": "t2_author12",
    "content": "some content",
    "nsfw": false,
    "promoted": true,
    "score": 15.05,
    "submakala": "submakala",
    "title": "title 1"
}'

curl --location --request POST 'http://0.0.0.0:8600/api/makala/v1/post' \
--header 'Content-Type: application/json' \
--data-raw '{
    "author": "t2_author12",
    "link": "https://makala.com",
    "nsfw": false,
    "promoted": true,
    "score": 15.05,
    "submakala": "submakala",
    "title": "title 1"
}'

Points of improvement

• The microservice is not fully covered with unit tests, because this is just a task, not a microservice shifting to production :)
• The url validation might be improved. For example, we should check a protocol (https, http).
• The integration test should be commented properly and needs refactoring
• Ids of promoted posts are stored in Redis, while they are fetched from PostgreSQL. The storage in Redis could be eliminated