mmattax / url-shortner Goto Github PK

Allow the url field to be changed. Because (account_id, url) is unique, we'll need to check for an existing url before we update.

TBD: When a url is changed, should we reset visit_count and delete the visit records?

Terraform all AWS components:

• IAM
• VPC
• ECS Fargate
• RDS
• SQS
• Elasticache

• Insert a visits record
• Increment url.visit_count
• 302 to the destination url

We should be careful when incrementing url.visit_count. Write the SQL to perform the update atomically:

update visit_count = visit_count + 1 where id = ?

Move insertion of visits and incrementing visit_count to a background script that consumes an SQS message that will be queued when a url is visited.

Let :id be the id or slug

See: https://echo.labstack.com/cookbook/middleware/

Look for X-API-Key in the header and then lookup the api key in the accounts table. Pass the account struct on the request context.

Build out the following schema. Let's use https://github.com/rubenv/sql-migrate to handle migrations.

Should we use a cookie? For now let's not (avoid GDPR?). Instead, store a fingerprint (hash) of soem request metadata: UserAgent + IP?

Use multi-stage build to place a static Go binary on a scratch image

• Should support date filtering: start, end
• Allow date sorting direction ?sort=asc|desc
• Don't worry about paging for now

Response should look something like:

{
    total_visits: 2344,
    total_unique: 1567,
    visits: [
      {
        date: '2020-04-26'
        total: 345,
        unique: 298
      },
      ...
    ]
}

Queries look something like:

Total:

select date(created_at) as d, count(*) from visits where created_at >= $start and created <= $end group by d

Unique:

select date(created_at) as d, count(unique(vistor_fingerprint)) from visits where created_at >= $start and created <= $end group by d

Build docker image + push to ECR + update ECS service definition

Stub out a working scaffold of the tech stack with Docker Compose. Containers needed:

• golang:alpine running Echo
• postgres
• memcache

The biggest concern here is generating the shortened slug that's going to be used in the URL. After a bit of research, it looks like base64 or base62 encoding the id is a tried and true method:

https://stackoverflow.com/questions/2375968/url-shortener-best-encoding-method

Found a Go library that does the encoding and swaps chars into url encodable values: https://github.com/lytics/base62