Docyrus Data View Config Design

A data view is a saved, named configuration over a single data source's records. It does not change the schema or the data — it controls how a grid/gallery surface (TanStack Table + react-querybuilder) presents that data source: which columns show, in what order, how rows are filtered/sorted/grouped/colored, and how paging/inline-editing/forms behave. Views back the "tab strip" / view picker in Docyrus list pages (useDocyrusDataViewSelect).

Author views with docyrus studio *-data-view, then validate (read back) and test (confirm the saved JSON unpacks to the intended grid state). An unvalidated view is not done.

For the schema being viewed see docyrus-data-source-design; for the React grid that renders views see docyrus-data-grid-page-design; for full CLI flag reference see docyrus-cli-app.

Critical model: opaque JSON blobs with packed sub-keys

The backend stores a view as a flat row (tenant_app_config_data_view) with four opaque jsonb columns — columns, filters, sort, color_rules — plus scalar columns (name, description, icon, color, is_default, sort_order, quick_filter_fields, archived, tenant_app_id). The backend does not validate the inner shape of the four jsonb blobs (the DTO only checks they are objects). The grid UI packs/unpacks them with specific sub-keys — if you write the wrong sub-keys the view saves fine but renders as empty defaults.

The single most important fact: the stored sub-key names differ from the TanStack/UI property names. e.g. visible-columns is stored as columns.visibility (not columnVisibility); color rules are color_rules.row / color_rules.cell (not rowColorRules). Get these exact key names from references/view-config-fields.md — do not guess them.

columns      → { visibility, order, pinning, grouping, rowHeight, displayMode,
                 pagingEnabled, pagingMode, pageSize, inlineEditingEnabled,
                 readOnlyColumns, columnOptions, formId }
filters      → { columnFilters, filterQuery }
sort         → { sorting }
color_rules  → { row, cell }

Workflow

Confirm auth, app, and data source. A view belongs to a data source inside an app.
```
docyrus auth who --json
docyrus apps list --json                                   # find appSlug
docyrus studio list-fields --appSlug <app> --dataSourceSlug <ds> --json
```
The list-fields output gives the column ids / field slugs you reference in visibility, order, grouping, sorting, filters, and color rules. For enum/select/status filter values you need enum row UUIDs — docyrus studio list-enums --appSlug <app> --dataSourceSlug <ds> --fieldSlug <f> --json. If there is no session, stop and ask the user to run docyrus auth login.
Decide the view's intent. Per view settle: name (tab label), which columns are visible and in what order, default sort, any filter (simple column filters and/or a query), grouping, row height, display mode (table or gallery), color rules, paging, inline editing, whether it is the default view. Sketch this for the user when ambiguous. If the data source has a status/single-select field, default to status-based views — see Status fields → design views per status.
Build the payload. Prefer --from-file ./view.json for anything beyond a trivial view — the blobs are nested JSON and awkward inline. Use the per-blob convenience flags (--columns, --filters, --sort, --colorRules, --quickFilterFields take JSON strings; --name, --description, --icon, --color, --isDefault, --sortOrder are scalars) only for small views. See references/view-config-fields.md for every key's exact name, type, and accepted values, and references/workflow-examples.md for ready-to-adapt payloads.
Create the view. See Create.
Validate. Re-read with get-data-view and confirm every blob has the intended sub-keys with the right column ids / enum UUIDs. See Validate.
Test. Confirm the saved JSON round-trips to the grid state you intended (and, if possible, that it filters/sorts real records correctly). See Test.

Status fields → design views per status

A status field (status / single-select / "stage"-style enum) is a strong signal to design status-based views. The existence of such a field means the records move through stages, and the most useful tab strip on a list page is one that splits records by stage. When a data source has a status field, default to creating a set of status views (plus a default "All" view) unless the user wants otherwise.

How many views depends on how many enum options the status field has:

3–4 options → one view per option. e.g. status New, Active, Won, Lost → four views New / Active / Won / Lost, each filtering status = <enum uuid>.
More than ~4 options → group related statuses into fewer, meaningful views. One tab per value becomes an unusable strip. Group them by stage of life, e.g. New, In Progress, On Hold → an "Open" view; Done, Cancelled → a "Closed" view. Pick group names the user's domain would recognise.
Usually also keep an "All" view (no status filter) and mark the most-used view (often "All" or "Open") as is_default.

Each status view is an ordinary view whose filters blob targets the status field by enum row UUID, never the label (read UUIDs from docyrus studio list-enums ... --fieldSlug status --json):

Single status: one filters.filterQuery rule with = (or a columnFilters entry).
Grouped (multi-status) view: a filters.filterQuery rule using in with an array of UUIDs.

Grouped "Open" view payload (filter only — merge with the rest of your view config):

{
  "name": "Open",
  "icon": "circle-dot",
  "sort_order": 2,
  "filters": {
    "columnFilters": [],
    "filterQuery": {
      "combinator": "and",
      "rules": [
        { "field": "status", "operator": "in",
          "value": ["<new-uuid>", "<in-progress-uuid>", "<on-hold-uuid>"] }
      ]
    }
  }
}

These status views are exactly what backs the view-type / status tab strip planned during project planning — keep one view per visible tab.

Studio data-view command cheat-sheet

Selectors — pass exactly one of each pair; the CLI resolves the other and routes to /v1/apps/<appSlug>/data-sources/<dataSourceSlug>/views: --appId | --appSlug, --dataSourceId | --dataSourceSlug. Mutations also accept --data '<json>' or --from-file ./view.json (JSON only); explicit per-blob flags are merged over the file/data payload. Append --json for machine output.

docyrus studio list-data-views   --appSlug crm --dataSourceSlug contacts --json
docyrus studio get-data-view     --appSlug crm --dataSourceSlug contacts --viewId <id> --json
docyrus studio create-data-view  --appSlug crm --dataSourceSlug contacts --from-file ./view.json --json
docyrus studio update-data-view  --appSlug crm --dataSourceSlug contacts --viewId <id> --from-file ./view.json --json
docyrus studio delete-data-view  --appSlug crm --dataSourceSlug contacts --viewId <id> --json

list-data-views / create-data-view accept an optional --tenantAppId (maps to tenant_app_id) to scope a view to a different app than the data source's owner. Most views omit it.
update-data-view additionally accepts --archived (soft-delete / hide from the tab strip). archived is not the same as deleting — prefer --archived true over delete-data-view when you just want to retire a view.
create-data-view requires name. Everything else is optional; an omitted blob means "grid defaults" for that aspect.

Create a data view

Minimal (just a named view — inherits all grid defaults):

docyrus studio create-data-view --appSlug crm --dataSourceSlug contacts \
  --name "All Contacts" --isDefault true --sortOrder 1 --json

Configured (recommended via file — see references/workflow-examples.md for full examples):

docyrus studio create-data-view --appSlug crm --dataSourceSlug contacts \
  --from-file ./active-leads-view.json --json

Critical rules

Stored sub-keys ≠ UI property names. Visible columns → columns.visibility, order → columns.order, pinning → columns.pinning, grouping → columns.grouping, color rules → color_rules.row / color_rules.cell, sort → sort.sorting, filters → filters.columnFilters / filters.filterQuery. Wrong key = silently ignored (view renders as defaults). Full list in references/view-config-fields.md.
Blobs are not validated server-side. The DTO only asserts each of columns/filters/sort/color_rules is an object. Malformed inner JSON is accepted and then ignored by the UI — so always validate by reading back and checking the unpacked shape; the create call succeeding proves nothing about correctness.
Column ids are field slugs (plus built-ins like id, name, autonumber_id). Get them from list-fields. A typo'd column id in visibility/order/sorting is dropped silently.
Enum/relation filter values are UUIDs, never labels (same rule as records). Read them from list-enums / the related data source before writing a filters blob that targets a select/status/relation field.
columns.visibility is an allow/deny map, not a column list. It is { "<colId>": true|false }; columns absent from the map use their default visibility. To hide a column set it false; to force-show set true. Use columns.order to control sequence.
Many features live inside the columns blob, not their own column: paging (pagingEnabled/pagingMode/pageSize), inline editing (inlineEditingEnabled/readOnlyColumns), per-column options (columnOptions), and the bound form (formId). This is so the backend needs no schema change — see references/view-config-fields.md.
is_default is per data source. Marking a view default does not auto-unset other views server-side; if exactly-one-default matters, clear the previous default's is_default yourself. The UI fallback chain is: persisted localStorage selection → is_default view → first view.
quick_filter_fields is a top-level uuid[] column (field ids), not part of any blob.
Validate then test, every time. Read back the view and confirm the unpacked shape; delete throwaway test views you create.

Validate

After authoring, confirm the saved view is exactly what you intended:

docyrus studio get-data-view --appSlug <app> --dataSourceSlug <ds> --viewId <id> --json — confirm scalars (name, is_default, sort_order, icon, color, archived) and that each blob holds the expected sub-keys with valid column ids / enum UUIDs.
Cross-check every column id in columns.visibility/order/grouping, every sort.sorting[].id, every filter field, and every color_rules formula reference against list-fields output — drop anything that does not resolve.
docyrus studio list-data-views ... --json — confirm the view appears, ordering (sort_order), and that at most one is_default is set if that was the intent.

Per-field "what correct looks like" checklist is in references/workflow-examples.md.

Test

Prove the view behaves as intended:

Round-trip check — re-read the view and unpack each blob using the key map in references/view-config-fields.md; confirm it yields the visible columns / order / sort / filter you designed.
Filter sanity — run the equivalent query against real records to confirm the filter selects the right rows, e.g. docyrus ds list <app> <ds> --columns "..." --filter "..." (see docyrus-cli-app), or docyrus dsql for complex filters. The view's filters.filterQuery mirrors react-querybuilder; the column filters mirror TanStack columnFilters.
Clean up — docyrus studio delete-data-view ... --viewId <id> for throwaway views (or --archived true to retire without deleting).

Full test playbook is in references/workflow-examples.md.

References

references/view-config-fields.md — The authoritative field reference: every scalar column, the exact packed sub-keys of columns / filters / sort / color_rules, their types, accepted values (display modes, row heights, paging modes), the UI-name↔stored-key mapping, and gotchas. Read this before writing any non-trivial view payload.
references/workflow-examples.md — End-to-end worked examples (default view, filtered+sorted+grouped view, color-coded view, gallery view, paging + inline-editing view, form-bound view) as copy-adaptable JSON payloads, plus the validation checklist and test playbook.

docyrus-data-view-config-design