@opengovsg/plumberscript
v0.0.2
Published
Programming language for Plumber Compute, implemented in TypeScript
Downloads
8
Readme
PlumberScript
A safe programming sandbox for Plumber Compute, implemented in TypeScript.
Getting Started
Development
Get your repository set up with:
git clone https://github.com/opengovsg/PlumberScript.git
cd PlumberScript
npm install
Library usage
Example in Typescript:
import { PlumberScript } from '@opengovsg/plumberscript'
const plumber = new PlumberScript()
plumber.evaluate(`let x = 1;`)
plumber.evaluate(`let y = 2;`)
const value = plumber.evaluate(`x + y`)
console.log(value) // 3
REPL usage
Build and run the compiled version with
npm run build
npm run plumber # Interpreter mode
npm run plumber examples/fibonacci.plumber # File mode
For development:
npm run dev
which will fire up the interpreter.
Syntax
Arithmetic
- Negation Operator:
-
converts a number to it's negative representation - Multiplication Operator:
*
multiplies two numbers together - Division Operator:
/
divides two numbers together - Addition/Concatenation Operator:
+
adds two numbers together, or concatenates two strings into a single string - Subtraction Operator:
-
subtracts one number from another
Logical Comparisons
- Equality Condition:
==
evaluates totrue
if two values are equal, andfalse
otherwise - Not Equal Condition:
!=
evaluates totrue
if two values are not equal to each other - Greater Than Condition:
>
istrue
when one number is greater than another, andfalse
otherwise - Less Than Condition:
<
istrue
when one number is less than another, andfalse
otherwise - Greater Than or Equal Condition:
>=
istrue
when a value is greater than or equal to another, andfalse
otherwise - Less Than or Equal Condition:
<=
evaluates totrue
when a value is less than or equal to another, andfalse
otherwise - And Condition:
and
evaluates totrue
when two logical expressions both evaluate totrue
- Or Condition:
or
evaluates totrue
when either of the logical expressions evaluate totrue
. The right expression is not evaluated of the left returns a truthy value.
Plumber Functions
The following Plumber Functions are currently available:
ABS(number)
: Returns the absolute value of a numberPOWER(base, exponent)
: Returns the result of a number raised to a power
These functions are implemented in src/lib, but more can be readily implemented when necessary.
Advanced usage
PlumberScript can also be used to define and execute scripts, not just simple expressions.
For example, the following code prints the first ten Fibonacci numbers:
function fibonacci(n) {
if (n == 0) return 0;
if (n == 1) return 1;
return fibonacci(n - 1) + fibonacci(n - 2);
}
for (let i = 0; i < 10; i=i+1) {
print fibonacci(i);
}
Motivation
We were inspired by compute features available in other products, such as Zapier Formatter and Tableau's Calculated Fields, which allowed flexible compute for users to transform their data.
At the same time, safely evaluating untrusted user-submitted code was an extremely difficult problem. We found that it was not trivial to implement this feature using open source options that could strike the right balance between user-friendliness with strict safety requirements. Libraries such as math.js were highly domain-specific, while others such as isolated-vm would expose the full functionality of JavaScript to end users at the same time. Even NodeJS' own VM module put up big red warning signs stating that it is not a security mechanism.
With PlumberScript, we can sandbox untrusted code while retaining control of the syntax at the same time. Dangerous operations such as reading the file system, accessing process
module, or performing network calls cannot be done because they have not been built into the language interpreter.
Performance
PlumberScript is expected to be much slower (at least 100x) compared to an equivalent C or Rust implementation. As a tree-walk interpreter built on top of a high-level NodeJS runtime, it is unable to take advantage of many common compiler optimisations (e.g. memory layout, processor cache). However this is not a priority at the moment as users are expected to submit simple expressions most of the time.
Caveat
This repository has not yet undergone VAPT or tested for code breakout.
Acknowledgements
We would like to thank Robert Nystrom and his invaluable insights in his book Crafting Interpreters.