Research: SQL Formatting & Pretty Printing Theory

[productdevbook]

Summary

Research on SQL formatting, pretty printing theory, and code generation approaches.

Key Findings

Foundational Papers

• Oppen (1980): "Prettyprinting" — original linear-time algorithm
• Hughes (1995): "The Design of a Pretty-printing Library" — algebraic approach
• Wadler (1998): "A prettier printer" — simplified Hughes with group/nest/line combinators
• Lindig (2000): "Strictly Pretty" — imperative implementation of Wadler's algorithm

SQL Formatters Analyzed

• sql-formatter: Tokenizer → Parser → CST → Formatter pipeline
• prettier-plugin-sql-cst: CST-based parsing with Wadler-style layout
• pgFormatter: Perl-based, PostgreSQL-specific
• SQLFluff: Python linter+formatter, rule-based
• CockroachDB sqlfmt: Extends Wadler with SQL-specific right-alignment

Architecture Recommendation

Use Wadler-style document algebra as intermediate representation:

• text(s) — literal string
• line — newline or space
• nest(n, doc) — increase indent
• group(doc) — try flat, break if too wide
• This separates "what to format" from "how to format"

SQL Style Consensus

• One clause per line (SELECT, FROM, WHERE, etc.)
• 2-4 space indentation
• CTEs over subqueries
• Explicit JOIN syntax
• UPPERCASE keywords

Action Items

• Implement Wadler-style document IR in printer/formatter.ts
• Support 3 output modes: compact, formatted, debug (with params inlined)
• Follow Holywell/Kickstarter style guide conventions
• Add configurable indentation and keyword casing

[productdevbook]

Research completed. Findings implemented in Wadler-style document algebra (src/printer/document.ts) and keyword-aware SQL formatter (src/printer/formatter.ts). See #24 for implementation details.