MAD

A report is a file.

MAD is an embedded reporting platform: the MAD Designer and MAD Viewer apps for iOS, the headless engine that runs them, and the open .mad document format. A .mad file holds a report’s whole definition – the query, its tables and joins, and the layout of its elements – in under 100 KB. Anyone who can read the file can run the report against the database and export it.

The code is at github.com/mthomason/MAD, MIT-licensed.

Why MAD

• Written, not assembled. The SQL generation, the data-table engine, the store, the layout model, and the renderer are first-party Objective-C across thirteen frameworks. Third-party code is limited to the wire – MariaDB Connector/C, FreeTDS, SQLite – and the barcode library. No reporting SDK is being resold.
• Mobile-first, memory-budgeted. A refresh reads the result set once into a SQLite row cache and computes from there, so aggregates and paginated rendering stay fast within a phone’s memory.
• Embeddable. Ten of the thirteen frameworks are UIKit-free and build headlessly. The mad CLI exposes the engine as verbs – inspect, schema, refresh, export – and drivers register themselves, so a new data source is a subclass, not a fork.
• Portable. The .mad container is a bzip2 tar archive holding a portable SQLite store; no Apple framework is required to open it. The headless build passes 1,270 tests on Linux, in Docker.

The .mad format

The open specification, one chapter per file:

What a document is · The container · The store · Reading and writing · Companion files · Hazards · Normative v1 · Open questions · Verification

Using the engine

• Build and test – the exact commands for the iOS suites and the Linux build.
• The mad CLI – inspect, refresh, and export a .mad from Docker in one line.
• Driver-authoring tutorial – the intended first contribution: subclass MADDatabase, register the driver, return rows.

The direction

Vision – a report’s result table becomes a data source for other reports, so one document can span several databases.

---

The MAD Designer app – the report designer built on this engine – has its own site at mad.everydayapps.com.

This site is open source. Improve this page.