crdis — Crystal Reports Disassembler

A clean-room, cross-platform parser for Crystal Reports .rpt files.

Goal

Given a .rpt file, extract everything in it without depending on the proprietary Crystal Reports runtime, COM/OLE Automation, or any Windows-only component:

• All text content (field names, labels, formulas, embedded SQL, parameter definitions)
• Embedded images (extracted as standalone PNG/JPEG/BMP files)
• Element geometry — for every visible element, its bounding box (x, y, width, height) and section
• Eventually: full structural dump matching what the Designer view shows

Why

Crystal Reports .rpt is a closed, undocumented binary format. Existing OSS tooling (e.g. rpt-to-xml) wraps the proprietary engine and runs only on Windows with the runtime installed — which defeats the point of breaking free of the legacy stack. This project builds the parser from scratch.

Target

• Format generation: Crystal Reports for Visual Studio (CRforVS), version 13.x — files produced by Visual Studio 2010 through 2019. One format family, mostly stable across that range.
• Other versions: out of scope for now; revisit once 13.x is solid.

Methodology

We do not have a spec, so we make one. The loop:

1. Controlled-pair sampling. In the CRforVS Designer, build a minimal .rpt file (e.g. empty report). Then build a second one differing by exactly one known change (e.g. one extra label at known coordinates). Save both with ground-truth notes (screenshots, property-panel values, DB bindings).
2. Binary diff. The bytes that changed between the pair encode that one change. Read them.
3. Hypothesise. Form a structural hypothesis: "bytes at offset X are the element's x-coordinate in twips," etc.
4. Encode the hypothesis as a Kaitai Struct spec (.ksy) under spec/. Kaitai compiles to parsers in Python, JS, C, C#, Java, Go, Rust — the spec is the deliverable, the parsers are downstream.
5. Verify against more samples. Refine. Commit. Repeat.

Layout

crdis/
  spec/          Kaitai Struct (.ksy) specification — the canonical output of this project
  crdis/         Python reference parser + CLI (uses spec/ via kaitaistruct runtime)
  samples/       Paired .rpt files + ground-truth screenshots/notes (see samples/README.md)
  research/      Running RE notebook — diff logs, hypotheses, dead ends
  tests/         Verifies parser against samples

Status

Phase 0 — project setup. Awaiting first controlled sample pair (samples/001_empty_vs_one_label/).

Phases

• Phase 1: Carve text + images, locate elements with bounding boxes (x/y/w/h).
• Phase 2: Section tree, element type discrimination (Text vs Field vs Line vs Box vs Image), font/style attributes.
• Phase 3: Formulas, parameters, sub-reports, group/sort definitions, DB binding metadata.
• Phase 4: Round-trip — emit a structural JSON that fully reproduces what Designer shows.

Non-goals

• Rendering reports.
• Modifying or writing .rpt files (read-only parser).
• Supporting non-CRforVS versions until 13.x is solid.