Since requirements pointed out (should not require a pre-installed container/server) I didn't make a dockerfile, to run the server you need to do the following:

1. Install the required dependencies using the command go mod download
2. Start the server using the command go run cmd/api/main.go

Default HTTP Port is 8080 if you need to change it change the env var PORT by exporting it. ex: export PORT=9001

I added a unit testing to validate concurrent safety, as not to have negative balances

1. Run the command go test ./....
2. The test results will be displayed in the command line.

I added Postman collection you can run commands using it. this project automatically loads accounts from URL was sent in the challenge, and loads it in memory.

Postman Collection: Open Here

if you don't want to install postman you can run it via curl, these are the Endpoints:

you can get all accounts through [GET] localhost:8080/accounts/

response:

{
    "accounts": [
        {
            "id": "2ad31c8b-4ee2-4198-85a1-dfb14248fb51",
            "name": "Babbleblab",
            "balance": "4488.1"
        },
        {
            "id": "3a7389f3-d492-4521-ae73-865cb22f7f8a",
            "name": "Skaboo",
            "balance": "365.09"
        },
	....
]}

By default this is a highly-available endpoint, it won't gurantee balance consistency in case of an account locked (because of some transaction).

To enable safety you can send safe query param as ?safe=true to have a consistent balance sheets. I added this query param as not to make this endpoint very slow due to waiting to get an account during locking phase; because it fetches all accounts. (we should add pagination to fix this bottlenick)

curl example to run endpoint:

curl --location 'localhost:8080/accounts/?safe=true'

you can get a specific account through [GET] localhost:8080/accounts/:id

response:

{
    "account": {
        "id": "0a637cbd-5aec-4c3b-8bf0-d8a5eb95024c",
        "name": "Yambee",
        "balance": "3012.9"
    }
}

This endpoint is consistant. It'll always return the correct balance no matter what happens. and it does not support unsafe operations. curl:

curl --location 'localhost:8080/accounts/0a637cbd-5aec-4c3b-8bf0-d8a5eb95024c'

you can transfer money through [POST] localhost:8080/accounts/:from/transfer/:to

it returns the sender balance after transfering the money:

{
    "Balance": 3002.9
}

this endpoint automatically aquires a lock on sender, and reciever accounts before operating, I designed it to return error in case of a failure in transfering, could've added retrying to it with an exponential backoff on every trial to ehnance user experience.

curl:

curl --location 'localhost:8080/accounts/0a637cbd-5aec-4c3b-8bf0-d8a5eb95024c/transfer/662178e0-e898-4fa0-a5ac-70951a564f7c' \
--header 'Content-Type: application/json' \
--data '{
    "amount": 10
}'

Currently this service stores data on it's memory, it won't scale this way because it's stateful. I added on DefaultContext an interface named Database to allow extendable architecture.

type Database[T memorydb.IdentifiedRecord] interface {
	Set(key string, record T, opts memorydb.Opts)
	Setnx(key string, record T) error
	GetM(terms []string, opts memorydb.Opts) []T
	Get(key string, opts memorydb.Opts) (T, error)
	Keys() []memorydb.Key
	Length() int
	Lock(key string) error
	Unlock(key string) error
}

In case of we needed to scale out another pod or a replicated node of this service, we can add a package that implements thses methods and talks to any other database over network ex: Redis, Memcached, Mongodb, etc.

• Add idempotency key to transfer endpoint, and in memory responses