Frappe — Bake Customizations

Move live-database Custom Fields and Property Setters into the codebase so they survive bench migrate on fresh installs, are version-controlled, and no longer depend on ad-hoc DB state.

Quick start

1. Fetch customizations:

   BENCH=$(pwd)
   PYTHON="$BENCH/env/bin/python"
   SCRIPT=$(find "$BENCH" -path "*/frappe-customizations-baker/scripts/fetch_customizations.py" | head -1)
   "$PYTHON" "$SCRIPT" --app <app_name>
   

2. Choose baking target based on app ownership:

   • Same-app → Use frappe-doctype-schema (native field)
   • Different-app → Use frappe-customizations-writer (sync_on_migrate custom/ file)

3. Generate cleanup patch to delete DB records after baking

Workflows

Primary workflow: Bake customizations

Context: User requests customizations from DB → code (e.g. "persist field customizations", "bake in Employee fields").

Steps:

1. Fetch customizations via bundled script
2. Filter/validate candidates (skip custom doctypes, handle duplicates)
3. Initiate frappe-doctype-schema or frappe-customizations-writer for each DocType group
4. Draft cleanup patches (one per app) to delete now-redundant DB records
5. Run bench migrate to apply DB changes

Feature workflow: Ensure all necessary fields are baked in

Context: User is developing a feature and asks to ensure all required fields are baked in.

Steps:

1. Build a feature field map by reading code and listing (DocType, fieldname, usage location).
2. Check whether each field is already defined in code (doctype/*.json or custom/*.json).
3. For missing fields only, run a targeted search in dev DB customizations and continue with the bake plan.
4. If missing fields exist in DB customizations, continue from the detailed workflow (analysis -> bake -> cleanup patch).
5. If no matching DB customizations exist, proceed only with Step 3 of the detailed workflow (define fields in code), then continue the feature plan.

Advanced features

See REFERENCE.md for:

• Mental model: doctype_app vs customization_app distinction
• Detailed step-by-step workflow with edge cases
• Feature field-map workflow for feature development
• Decision tree for filtering candidates
• Common pitfalls (blank modules, Property Setter quirks, default_print_format)