Squiggle Programming Language

The @quri/content package is an internal utility used for maintaining Content Collections. It serves as a centralized hub for managing structured content used across the Squiggle ecosystem, including documentation for the website and raw data for AI system prompts.

This package re-exports all generated content collections and provides helper functions for interacting with them.

You can find Squiggle models and related analytical writeups in several locations across the ecosystem:

• Squiggle Hub: The primary application for hosting and interacting with models (squigglehub.org).
• Effective Altruism Forum: Search for the Squiggle tag to find models and discussions.
• Squiggle Models Repository: A dedicated list of QURI Squiggle models can be found in the squiggle-models repository.

The apps/hub/src/lib directory contains shared logic for the Squiggle Hub application. It is organized by the intended runtime environment and usage pattern:

• lib/server: Contains code that must only run on the server (e.g., database access, sensitive logic).
• lib/hooks: Contains common React hooks intended for use in client-side components.
• Other files: General shared logic that can be used in both server and client contexts.

The Plot module provides functions for visualizing distributions and mathematical functions.

There are two ways to generate plots:

1. Direct Plotting: Using raw functions and distributions, which are plotted using default parameters.
2. Plot Objects: Using specific functions from this module that return plot objects. These objects provide granular control over chart parameters and allow for the creation of more complex chart types.

The Danger library provides experimental modules that are less stable than the core Squiggle library. These modules are intended for testing new ideas rather than production environments. The available modules include:

• Javascript: Provides near 1-to-1 matches of Javascript functions.
• Math: Mathematical functions.
• Combinatorics: Combinatorial functions.
• Distributions: Statistical distribution functions.
• Integration: Integration functions.
• Optimization: Optimization functions.
• JSON: JSON-like object serialization (Danger.json and Danger.jsonString).

The Metaforecast website uses a specific stack for GraphQL implementation:

• Backend: Pothos for the GraphQL server.
• Frontend: urql for client-side queries.
• Code Generation: GraphQL Code Generator for schema and query generation.

Key Files:

• schema.graphql: The generated GraphQL schema.
• codegen.yml: Configuration for graphql-code-generator.
• graphql.config.yaml: GraphQL Config for VS Code integration (enables type-checking in .graphql files).
• src/pages/api/graphql.ts: The GraphQL HTTP endpoint and playground.
• src/web/urql.ts: Helper functions for urql.
• src/graphql/: Contains the GraphQL server implementation.
• src/web/**/*.graphql: Contains fragments, queries, and mutations used on the frontend.

Squiggle is a minimalist programming language designed for intuitively-driven quantitative estimation. It allows users to perform math with probability distributions and encode functions as forecasts.

Key Characteristics:

• Purpose: Intuitively-driven estimation rather than data analysis or data-driven statistical techniques.
• Embeddability: It is an embeddable language that can be integrated into JavaScript applications.
• Not a Probabilistic Programming Language: Unlike standard probabilistic programming languages, Squiggle does not support Bayesian inference (backwards inference).
• Not for Data Analysis: It is not intended to replace tools like Python or Julia for substantial data analysis, nor is it a replacement for enterprise Risk Analysis tools (e.g., Crystal Ball, @Risk).

Starting with @quri/squiggle-lang v0.8.0, the expression a to b always evaluates to a lognormal distribution. Note that the arguments must be positive-only. Previous versions allowed these to fall back to normal distributions for values zero or less, but this behavior has changed.

// Evaluates to a lognormal distribution
5 to 10