SimpleInMemoryDatabase

Implement an in-memory database similar to Redis. Be able to read commands via standard input (stdin), and print appropriate responses to standard output (stdout).

Test coverage

File	% Stmts	% Branch	% Funcs	% Lines	Uncovered Line #s
All files	97.14	94.59	100	97.14
DatabaseManager.ts	100	100	100	100
Processor.ts	93.75	94.44	100	93.75	53
TransactionManager.ts	95.83	85.71	100	95.83	39
constants.ts	100	100	100	100

• Note that this test coverage is generated once the file is commited. This is a placeholder for later CI/CD

---

Prerequisites

Node	Yarn (Optional)	Docker (Optional)
18.x.x	Install	Install

• Note: Can skip install yarn if running the project in production mode using Docker

Setup

1. Install dependencies

   $ yarn set version stable
   $ yarn install

2. Setup repository

   $ yarn setup

   • Running this command to make sure the pre-commit hook would be installed correctly
     • Pre-commit hook will be triggered for all *.ts files and all other file types supported by Prettier

---

Development

Start application

We can either start the application using Docker or Yarn Note that the docker version is not yet supported for file changes watch

• Using yarn

  $ yarn dev

• Using Docker

  $ npm run docker:dev

Test

$ yarn test

---

Production

Build and start application

$ yarn build
$ yarn start

---

Q&A

How can I check the linter issues?

$ yarn lint

• This command will run linter for:
  • Typescript checking
  • Eslint for *.ts files
  • Prettier for the other file types (E.g. *.md, *.json)

How can I quickly fix the linter issues?

$ yarn format

• This command will fix some auto-fixable issues. The rest must be yours

How can I find out dependencies I accidentially added without using?

$ yarn depcheck .

• Note that the command will work only if you have not imported the package anywhere

---

Explaination and TODOs

Checkbox

• Execute successfully with test cases provided in the requirement
• Compatible with Typescript
• Has linter, dependencies checker, and pre-commit hook
• Able to run with docker (to skip Yarn installation)
• Test coverage > 90%
• Ready for production deployment

Optimize memory consumption

Current behavior

• The current implementation is not the best solution, due to the fact that TransactionManager will initialize another layer of DatabaseManager whenever a transaction is initialized
• Because the time is over (3 hours), I would like to stop the development at this stage, as long as the program can be executed properly (tested with test cases in src/index.spec.ts)

Proposed solution

• In each transaction, the number of changed variables are not varied. Thus, instead of taking a layer as DatabaseManager, we can create another class like TransactionLayer, which contains:
  • List of changed variables within the current transaction
  • Values of changed variables within the current transaction
• TransactionManager in this case will have an additional hashmap ({[key: string]: TransactionLayer[]}) to store the key and its corresponding TransactionLayer
  • Thus, when execute GET, we will retrieve the TransactionLayer based on the key and get the value by the TransactionLayer
• Further explanation:
  • COMMIT will initialize an empty layer uppon called
  • ROLLBACK will pop the layer, and scan through changed keys in order to update key - TransactionLayer hashmap of TransactionManager
  • COMMIT will merge the changed values of current layer with previous layer
  • SET, UNSET will be executed on the current layer, and add the key to changed-keys list
• Note:
  • Using this approach, the data will be duplicated at the level of DatabaseManager and TransactionManager, since we will need 2 hashmaps to store the key-value and key-TransactionLayer, also 2 hashmaps to handle inverted index of value list