
export const Title = "Hello again, again";
export { BlogPage as default } from "@/Components";
import { Meta } from "@/Components";

<Meta title={Title} />

# {Title}

I don't think of myself as someone who loves to write. But it's hard to deny that my most basic instinct whenever I need to express something is to scribble and draw and write. 13 years ago, I convinced my parents to get me the [aryanrocks.com](https://web.archive.org/web/20150919152819/http://aryanrocks.com/) domain and help me point my WordPress blog to it. 

For several years, I used it as my personal mood board, collecting articles, infographics, video games, and jokes I found all across the internet, along with my own posts and games. I couldn't say it was really my greatest work, but I think it speaks to a very human desire to share and truly capture and represent yourself. And that's what I wanted to bring back here.

I am a terminally online person. I'm always listening to podcasts, reading Substack, and following the daily ebbs and flows of the internet town square.  That sometimes feels like an unstoppable faucet of information that I'm not actively retaining. I think that's the core idea behind this blog. I want to deep-dive on specific things I come across on the internet. I want to call out the beautiful things I find online and connect them to what I care about.

As a disclaimer: this (and most of my articles) aren't meant to be well-written or comprehensive prose. This is somewhere between being a public mirror of my obsidian vault and a journal to point friends to.

There are a lot of projects that I start and end up not finishing. Hopefully, this won't be one of them :)


Hello again, again


export const Title = "Running a interpreter in WASM";
export { BlogPage as default } from "@/Components";
import { Meta } from "@/Components";
import Imp from "@/ImpCompiler";

<Meta title={Title} />

# {Title}

Do not try this at home.

<Imp />

This uses [HeavyImp](https://github.com/aryansw/heavyimp), which is an interpreter in Rust for a simple imperative language with pointers, written for a [Programming Languages](https://www.cs.purdue.edu/homes/bendy/cs456/fall22/) class. This interpreter can report two types of errors: parsing and type-checking errors. Errors would not be possible while actually running the program, as the type-checker would have already caught them.

To use this Rust project in the browser, I used [wasm-pack](https://rustwasm.github.io/wasm-pack/), to create an npm package that can be imported into the project. Some (a lot) changes were necessary in the `wasm-refactor` branch to get this working, as well as a lot of wrangling with Next.js and Webpack.

Here's also probably a good place to shout out [MDX](https://mdxjs.com/), for making it so much nicer to write posts, with React components right in the middle of a markdown document. It was the only thing that didn't let me down when trying to get this working.


Using WASM to run a simple imperative language interpreter, originally written in Rust, on the browser

Running HeavyImp on WASM


export const Title = "gql-db, Explained.";
export { BlogPage as default } from "@/Components";
import { Meta, DesignDiagrams } from "@/Components";

<Meta title={Title} />

# {Title}

gql-db is an SQL database server with version control integrated into the database itself. It's written in Rust and uses Protocol Buffers/gRPC for communication. This was our Capstone Software Engineering project for Purdue University.

The motivation behind this, would be to allow developers to work on separate "branches" of the database, and then merge their changes into a main branch. Reverting changes, now termed as commits, in a database also becomes much easier.

Implementing this as a _component_ of the database system, instead of as a "layer" on top of the database system allows us to take advantage of the same underlying infrastructure used by the database system to efficiently store and access version control information.

Learn more at our repos: [GQL-Project/gql_db](https://github.com/GQL-Project/gql_db) and [GQL-Project/gql_client](https://github.com/GQL-Project/gql_client)


## Features

- SQL:
  - Basic SQL operations: `SELECT`, `INSERT`, `UPDATE`, `DELETE`, `CREATE TABLE`, `DROP TABLE`
  - Support for `JOIN` operations, as well as `GROUP BY` and `ORDER BY` operations
  - BTrees for indexing
  - User Permisions
- Version Control:
  - Creating, Reverting and Squashing Commits
  - Creating, Switching and Deleting Branches
  - Pulling Changes from source branch
  - Merging Branches, with conflict resolution policies
  - Logs, Viewing Specific Commits and Viewing Database Schema at a specific commit

## Language Choices

- When choosing a language for the database server, we looked for one that would be fast, safe, and with a great package management system. This is what brought us to Rust.

  - Rust is fast, safe, provides zero-cost abstractions, and has a great type-system.
  - The _structure_ of the language forces us to write code in a manner that eliminates huge sets of concurrency and memory bugs.
  - Iterators in Rust are iterators done right: performing transformations and selections on arrays becomes really straightforward and concise.
  - The toolchain that Rust provides with its compiler makes a very seamless experience. The `rustc` compiler additionally provides very helpful error messages and warnings. `cargo` allows us to painlessly introduce helpful packages and resources into our project.
  - There were some _drawbacks_, particularly its steep learning curve and slow compile times, but overall, using Rust allowed us to iterate and add features very quickly.

- To efficiently send data between the client and server, we chose Protocol Buffers and gRPC. Protocol Buffers are a language-neutral, platform-neutral, extensible way of serializing structured data. It's extremely efficient, and allows us to define a "structure" for our queries that needed to be enforced.

- We chose React for the client because it's a popular framework for building web applications, and it's also easy to learn.

  - Adding Next.js made using multiple pages much cleaner, and allowed us to also clearly separate the client and server.
  - Using Material UI additionally allowed us to make a consistent theme through the entire site, as well as draw from a large library of components in our application.

## Design Overview

- The database is represented as a "folder", with each "file" inside it representing a table. This table "file" is a binary format file, which stores the schema for the table, the rows for the table, and any indexes the table uses.

  - These table files store information in "pages", with each page holding `4096` bytes of information.
  - Any accesses to the table are made as sequences of page accesses, rather than individual row accesses.
  - Using this page format helps reduce the number of IO operations required when working with tables.
  - Using a binary representation for this file, rather than a simple `.csv` representation, allows us to more efficiently store the rows of the table.
  - Additionally, it makes updating rows and removing rows, two crucial operations for any database, much faster.
  - For more information, see: [Page File Structure](https://github.com/GQL-Project/gql_db/blob/main/src/fileio/README.md)

- All branches, commits and files are stored in the database server itself. This differs from git, where commits are first stored in a local repository and then pushed to a remote repository.

  - The branches and commits for a database are stored across four files.
  - `deltas.gql`: Stores the actual changes made in the commits, as well as any important metadata. This uses a modified version of the page representation format, to better handle variable-sized information.
  - `commitheaders.gql`: A table to quickly lookup the location of a commit in the actual `delta.gql` files using the commit hash.
  - `branches.gql`: A table that stores the various _nodes_ that are linked together to form branches.
  - `branchheads.gql`: A table to store the current _HEAD_ for a particular branch. In other words, the latest commit made on a particular branch.
  - For more information, see: [Version Control System](https://github.com/GQL-Project/gql_db/blob/main/src/version_control/README.md)

- Here's some additional helpful diagrams we made while designing the server:

<DesignDiagrams />

## Developers

#### [Aryan Wadhwani](https://github.com/aryansw)

#### [June Seo](https://github.com/JuneSeo27)

#### [Kade Boltjes](https://github.com/Kboltjes)

#### [Ryan Hawks](https://github.com/rhawks00)

#### [Shayne Marques](https://github.com/shayne243)

#### [Umang Sharma](https://github.com/UmangSharma24)

## Screenshots

![gql terminal example](/projects/gqldb.png)

![gql client example](/projects/gqlclient.png)

![gql client example](/projects/gqlclient2.png)

![gql client example](/projects/gqlclient3.png)


What's so special about gql-db, why did we make it, and what can it do currently?

Aryan Wadhwani

Systems Software Engineer at Apple

Blog

Hello again, again

Running HeavyImp on WASM

What is gql-db?