FileMaker XML Compressor

Converts a FileMaker Save a Copy as XML file (FileMaker 2024 / v21–22) into compact, LLM-friendly text files that Claude can actually work with.

What It Does

Takes your large <FMSaveAsXML> export and produces ~10 small text files organized by category, typically 95-97% smaller than the original. It:

• Strips all styling/positioning data (bounds, CSS, coordinates, colors, fonts)
• Strips theme definitions, UUIDs, and internal tracking data
• Preserves every table, field, calculation, relationship, script step, layout object, value list, custom function, and security setting
• Deduplicates data that FileMaker repeats across sections
• Handles UTF-16 LE/BE with BOM natively (no encoding gymnastics on your end)
• Preserves emoji intact — no restoration shim needed
• Uses a custom shorthand format (not XML or JSON) designed for maximum token efficiency

Requirements

• macOS (the AppleScript droplet) — Python 3.6+ for the standalone script
• Python 3.6+ (no external packages — uses only the standard library)

Two Ways to Use It

A. The Mac App (recommended)

build/FileMaker XML Compressor.app — a drag-and-drop droplet built from build/XMLCompressor.applescript.

1. Drop a *_fmp12.xml file (your Save-a-Copy-as-XML export) onto the app icon
   • or double-click the app to get a file picker
2. Pick where the compressed output folder should go
3. The app runs the compressor and offers to reveal the result in Finder

First-launch note: unsigned, so Gatekeeper blocks the first launch. Right-click → Open once and approve.

B. The Python script

# Basic usage — outputs to a "<filename>_Compressed" subfolder next to your XML
python3 fm_saxml_compress.py /path/to/YourFile_fmp12.xml

# Custom output directory
python3 fm_saxml_compress.py /path/to/YourFile_fmp12.xml --output-dir /path/to/output

How to Generate the Source XML

In FileMaker Pro (2024 / v21–22):

1. Open the file you want to export
2. File → Save a Copy as… (or the "Save a Copy as XML" item on newer versions)
3. Choose XML as the format
4. Save

This produces a single *.xml file. Feed that to the app or script.

Output Files

The output folder contains 10 files. Upload what you need based on your task:

File	Contents	Upload when...
01_SCHEMA.txt	Tables, fields, calcs, auto-enter, validation, storage	Working on data model, field calcs, or schema questions
02_RELATIONSHIPS.txt	Table occurrences, relationship predicates, sort orders	Debugging relationships or building new ones
03_SCRIPTS.txt	All scripts with every step and parameter	Working on script logic, debugging, or writing new scripts
04_LAYOUTS.txt	Layout objects, portals, triggers, conditional visibility	Working on UI, layout triggers, or object scripting
05_VALUELISTS.txt	Value list definitions (custom and field-based)	Working with dropdowns, pop-up menus
06_CUSTOM_FUNCS.txt	Custom function names, params, and full bodies	Debugging or writing calculations using custom functions
07_ACCOUNTS.txt	Accounts, privilege sets, extended privileges	Security questions
08_MENUS.txt	Custom menus and menu sets	Menu customization
09_FILE_REFS.txt	External data source paths	Multi-file solutions
10_SUMMARY.txt	Stats overview and file guide	Quick reference

Common combos:

• Schema work: 01 + 02 (+ 06 if you use custom functions in calcs)
• Script development: 03 (+ 01 if scripts reference field structures)
• Layout/UI work: 04 + 03 (scripts are often triggered from layouts)
• Full picture: 01 + 02 + 03 + 06 (usually fits in context)

Rebuilding the Mac App

The compiled .app is gitignored (it's a build artifact). To rebuild from source:

cd build
osacompile -o "FileMaker XML Compressor.app" XMLCompressor.applescript
cp ../fm_saxml_compress.py "FileMaker XML Compressor.app/Contents/Resources/"

(Recompiling wipes the Resources folder, so the cp always has to follow.)

If Files Are Still Too Large

The 03_SCRIPTS.txt and 04_LAYOUTS.txt files tend to be the largest. If they're still too big:

1. Open in a text editor
2. Search for SCRIPT: (or LAYOUT:) to find item boundaries
3. Copy just the items you're working on into a separate file
4. Upload that instead

Output Format Explained

The format uses a compact shorthand with a key at the top of each file. Example from SCHEMA:

T: Contacts (1523 records) [id:129]
  F: __pk_ContactID | N | T [id:1]
    AE: Serial(OnCreation,next:1524,inc:1)
    VAL: when:Always, notEmpty, unique
    STOR: idx:All
  F: NameFirst | N | T [id:2]
    STOR: idx:Minimal
  F: NameFull | C | T [id:5]
    CALC: NameFirst & " " & NameLast
    STOR: UNSTORED
  F: _cAge | C | N [id:12]
    CALC: Year(Get(CurrentDate)) - Year(DOB)
    STOR: UNSTORED
    CMT: Calculated age in years

This is roughly 10-20x more token-efficient than the equivalent XML while preserving all the information you'd need for development work.

Troubleshooting

MemoryError on very large files → The script loads the entire XML into memory. For files over 500MB, you may need a machine with 8GB+ RAM available. Close other applications first.

Missing sections in output → Some sections only exist if your file uses them (e.g., custom menus, external data sources).

App says "Compression failed" → The dialog shows the underlying Python traceback. Most often this means the XML isn't a Save-a-Copy-as-XML file (e.g., it's the old DDR format, or a record export). The root element should be <FMSaveAsXML>.

Legacy DDR Support

Earlier versions of this project compressed the old Database Design Report XML (<FMPReport> root). That code path has been retired in favor of the newer Save-a-Copy-as-XML format, which is faster, handles emoji natively, and contains everything the DDR did. The retired DDR scripts are preserved in git history if ever needed.