fix(pages): read a pluggable widget's microflow datasource from MicroflowSettings

[ako]

Fixes mendixlabs/mxcli#795 — DESCRIBE PAGE drops a datagrid's microflow DataSource, so round-tripped MDL recreates the grid with no data source.

(The issue number refers to mendixlabs/mxcli, not this fork.)

The bug

A DataGrid2 bound to a microflow described as:

datagrid g1 {
  column "Key" (Attribute: Key, Caption: 'Key')
}

— no DataSource. A database from grid on the same page described correctly, which is what made the loss look selective.

The reporter was right that the model stores it correctly. The page BSON holds:

"DataSource": {
  "$Type": "Forms$MicroflowSource",
  "MicroflowSettings": {
    "$Type": "Forms$MicroflowSettings",
    "Microflow": "MyModule.DBG_ListBucketObjects"
  }
}

The name lives in the nested Forms$MicroflowSettings; there is no top-level Microflow key. extractDataGrid2DataSource looked up ds["Microflow"], got "", and returned nil. The describe formatter's case "microflow" branch was correct all along — so neither the write path nor the stored model was ever wrong. Purely a read bug.

The fix

Four readers each had their own copy of this lookup and disagreed with each other — two read the nested settings, one read the top-level key, one did both. They now share microflowSourceRef / nanoflowSourceRef, which read the nested settings and fall back to the legacy top-level key so older files still round-trip.

Two further gaps the regression tests turned up while covering this:

• A Gallery whose source used the legacy flat shape was unreadable too. Gallery delegates to parseCustomWidgetDataSource, which handled only the nested form, and the flat fallback in its own switch is not reached for a CustomWidget gallery.
• Forms$NanoflowSource was missing entirely from the DataGrid2 and Gallery switches, so a Studio-Pro-authored nanoflow datasource described as nothing.

Left alone deliberately: CustomWidgets$CustomWidgetNanoflowSource (a different metamodel type whose Nanoflow key really is top-level) and the Forms$MicroflowAction reads (actions, not datasources).

Also corrects the datasource table in the create-page skill, which documented datasource: microflow Module.GetData() — the trailing () does not parse, as the reporter noted.

Verification

Reproduced on main before the change, against a real project:

	before	after
datagrid + microflow	datagrid g1 {	datagrid g1 (DataSource: microflow …)
datagrid + database	correct	correct (unchanged)
gallery + microflow	correct	correct (unchanged)

The round-trip the issue is actually about — describe → exec → describe — now produces byte-identical output.

make build, make test, make lint and make check-mdl all pass.

New coverage in mdl/executor/cmd_pages_describe_flowsource_test.go: a table over both extractors covering the nested shape, the legacy flat shape, and nanoflow, plus direct tests of the shared helpers including non-map MicroflowSettings (no panic). Verified these fail against the pre-fix code — including the two extra gaps, which I had not predicted from reading alone.

Repro fixture: mdl-examples/bug-tests/795-datagrid-microflow-datasource-describe.mdl, and a row added to the .claude/skills/fix-issue.md symptom table.

Out of scope, found while testing

Two things I did not touch, worth their own issues:

1. A nanoflow datasource on a pluggable datagrid is dropped at write time. mxcli exec logs modelsdk: pluggable widget data source *pages.NanoflowSource not yet supported — rerun with MXCLI_ENGINE=legacy and then still prints Created page. The read fix here covers Studio-Pro-authored nanoflow sources; mxcli still cannot author one, and it fails quietly.
2. The -- Context: comment names the microflow as the context entity — $currentObject (MyModule.DBG_ListBucketObjects) instead of the entity the microflow returns. Affects datagrid and gallery alike; cosmetic, describe-output only.

🤖 Generated with Claude Code

https://claude.ai/code/session_012XR649rKk68z6gBpngu6MA

---

Generated by Claude Code

[ako]

Superseded — split into four topic PRs, one per area, so each can be reviewed and taken upstream on its own:

• fix(pages): read a pluggable widget's flow datasource, and stop dropping children silently #59 — pages / describe: mendixlabs/mxcli#795 + two follow-ups (3 commits)
• fix(odata): read an external entity's OData source, and keep its local-changes flag #60 — OData external entities: mendixlabs/mxcli#782 (2 commits)
• fix(test): leave the project as mxcli test found it #61 — test runner: mendixlabs/mxcli#802, #803, #804 (1 commit)
• fix(microflows): serialize break/continue events, and size loop boxes to their contents #62 — microflows: mendixlabs/mxcli#791, #790 (2 commits)

All 8 commits and every changed file are accounted for across the four branches — verified by diffing the file sets, with nothing added or dropped. Each branch passes make build, make test, make lint and make check-mdl independently off current main.

Closing this one in favour of those.

---

Generated by Claude Code