Basic auth
Username + password against your Fusion instance. Passwords are stored only in the Windows Credential Vault — never in SQLite, never in plaintext on disk.
Features
Everything Data Collage does, on one page.
Data Collage is a desktop SQL workbench purpose-built for Oracle Fusion SaaS. The features below cover what ships in v1 — and what's next.
Connect
Multiple connections, two auth methods, zero plaintext credentials.
Save one connection per Fusion environment. Tag them PROD / TEST / DEV so you always know which pod you're hitting before pressing Run.
Fusion SSO with JWT
For environments that require SSO, Data Collage opens the Fusion login window, captures the JWT, and reuses it for BIP calls. Tokens are kept in memory and refreshed automatically when BIP rejects them.
Environment tags
Colour-coded chips (PROD / TEST / DEV / SANDBOX, custom labels supported) appear next to every connection and in the title bar — no more accidentally running an UPDATE-shaped SELECT against the wrong pod.
Test before save
A six-step Test Connection probes BIP, deploys the read-only gateway report under your user folder, and runs a SELECT 1 smoke test. Save is gated on a passing test, so a saved connection is always a working one.
Edit
A real SQL editor, not a textbox.
Monaco — the same editor that powers VS Code — with Oracle-aware tooling layered on top.
Syntax highlighting & autocomplete
Oracle SQL keywords, functions, and operators are highlighted. Identifiers complete from the metadata catalog as you type.
Curated function catalog
Bundled reference for the Oracle SQL functions you actually use, with signatures, descriptions, and one-click insert. Add your own — your additions persist locally.
Custom snippets
Define short codes (e.g. seld → SELECT DISTINCT) and full templates with tabstops. Snippets are stored locally and survive upgrades.
Bind parameters
Use :name placeholders in your SQL. Data Collage detects them, opens a dialog for values, and substitutes them safely — the validator and the BIP gateway never see the bind syntax.
Format & find
One-shot SQL formatting (Shift+Alt+F), full Monaco find/replace with regex, multi-cursor, and toggleable line/block comments. All the keystrokes you already know.
Multiple tabs
Open as many SQL tabs as you need. Each tab keeps its own results, errors, and execution time independently — switch between tabs without losing context.
Run
Read-only, enforced at three boundaries.
Fusion SaaS doesn't expose direct database connectivity. Data Collage routes queries through a deployed BI Publisher report — and validates every statement before it leaves your machine.
Local AST validation
Every query is parsed locally and rejected unless it's a SELECT or a WITH CTE that resolves to a SELECT. DML, DDL, and PL/SQL are blocked at the editor.
BIP gateway you control
On first Test Connection, Data Collage uploads a read-only BIP data model into your Fusion catalog under /~username/NCDataCanvas/. Your DBAs see normal BI Publisher traffic.
No DB password, ever
Auth uses your Fusion identity. The app has no database credentials — because it never needs any.
Errors that point to the right line
BIP fault strings are parsed into structured error panels. ORA codes are extracted, line numbers are mapped back to your editor, and the offending line is highlighted.
Results
A grid that scales — and an export that just works.
Powered by AG Grid, so a quarter-million-row result set behaves the same as a four-row one.
Sort, filter, hide
Column-level filtering and sorting on every result. Hide columns you don't need and the layout sticks across runs of the same query.
Formula columns with chained references
Add computed columns inline using a safe expression language. Reference real columns and other formula columns — chain NET_DUE = INVOICE_AMOUNT - NVL(AMOUNT_PAID, 0) then NET_DUE_PCT = NET_DUE / INVOICE_AMOUNT * 100. Cycle detection, NULL-safe arithmetic, and NVL fallbacks ship out of the box. Rename or delete a formula and dependents are pruned with a toast — no orphan columns.
Per-column number & date formats
Right-click any column header to apply a number or date format — thousands separators, decimal places, percent, scientific, or a currency from a curated catalog (USD, EUR, GBP, CAD, AUD, JPY, INR). Date formats include ISO, US, EU, Oracle, datetime, and relative ("3 days ago"). Formats survive across re-runs and travel with Saved Analyses.
Quick column actions
Right-click a column header for Keep only this column (one click to focus on one field), Copy unique values (deduped, newline-separated, honors filters), and Copy as IN clause. The IN-clause dialog handles text vs numeric, escapes single quotes correctly, wraps long lists onto multiple lines so Monaco can still highlight them, and auto-chunks past Oracle's 1,000-value limit into OR-joined blocks.
Column navigation
Toolbar buttons jump straight to the first or last visible column — handy on wide result sets. The target column briefly flashes so you don't lose your place.
Export to CSV or XLSX
One-click export of the current view (filtered or not) to CSV or Excel. Numeric columns export as numbers, dates as dates — no string-cleanup required at the other end. Currency formats survive the round-trip into Excel via native format strings.
Active-state pills + one-click clear
A pill row above the grid tracks every active overlay — duplicates filter, top-N filter, formatted columns, linked columns. Each pill has a × (or Clear) to remove just that overlay. Right-click any header or cell → Clear all filters wipes the column filter model in one shot, so you don't have to clear three column filters one at a time.
Row count & execution time
Every result tab shows total rows and wall-clock execution time at a glance, so you can spot a regression the second it appears.
Deep links
Open any record in Oracle Fusion — without leaving the grid.
Curated drill-back into Oracle Fusion 26A for the entities Financials and Procurement teams query every day. Right-click a cell, pick the entity, and the right record opens in your browser using your existing Fusion session. No credentials in the URL, no setup beyond the first mapping.
Twelve verified entities, ready out of the box
AP invoice / payment / supplier, AR transaction / receipt, GL journal, fixed asset, expense report, purchase order, purchase requisition, and project overview — all tested end-to-end on a 26A tenant. The catalog ships with the app; no per-customer setup or admin work.
Two-pane picker, grouped by module
Right-click any cell → Open in Oracle Fusion…. Modules listed on the left with entity counts; entities for the selected module on the right. Mouseover switches the right pane (Windows-style menu feel); a single click on an entity opens the mapping dialog — or, if you've already mapped that entity on this tab, opens the record directly.
You choose: column or fixed value per parameter
When the catalog's default column isn't in your result set (e.g. Asset needs BOOK_TYPE_CODE but your query against FA_ADDITIONS_B only selected ASSET_ID), flip that parameter row from Column to Value and type the fixed value — like CORP for the corporate book. Composite-key entities (Journal Entry, AR Receipt, Asset) work transparently — the URL composes from each row's mapped values, including spaces (encoded as %20, never +).
Linkify a column for every row
Tick Make {COLUMN} clickable for every row in this result and every cell in that column becomes an accent-underlined link — click any row to drill straight into it. A "Linked" pill appears above the grid; one click on the × unlinks. The link host doesn't have to be a key column — right-click AMOUNT, link it to AP Invoice, and AMOUNT becomes the clickable label while the URL composes from each row's INVOICE_ID.
Save vs Open
The mapping dialog has three actions: Cancel (no change), Save (persist the mapping + linkify, no browser tab opens), Open (Save plus open the row you right-clicked). Save is the right pick when you only want to enable the linkified column; Open is the right pick when you also want to follow the link you just right-clicked.
Per-tab cache, predictable invalidation
Once you've mapped an entity on a tab, future right-clicks on that entity skip the dialog and open directly. Re-run with a different column shape and the cache invalidates per entity — only mappings whose source columns went missing get dropped. Remove a linked column and that entity's cache also clears, so the next right-click opens the dialog fresh with the Save button ready.
No credentials in the URL
Deep links carry only the entity type, action, and key parameters. Authentication is your existing Fusion browser session — Data Collage never passes a password through the URL.
Directives
Pin the presentation to the SQL itself.
A small set of <code>--@</code> comment directives at the top of any query control how the result is presented. Oracle ignores them as plain comments; Data Collage interprets them on the way to the grid. They're text, so they version-control, paste, and share like any other SQL.
--@title
Set the tab title from the query. Updates live as you type — the tab renames within half a second of finishing the directive line. Survives Save Analysis and shows up as the default name in the Save dialog.
--@hide and --@show
Hide specific columns (--@hide: CREATION_DATE, LAST_UPDATE_DATE) or flip it around and show only the columns you want (--@show: INVOICE_NUM, AMOUNT_PAID). Unknown names are silently ignored — a typo never blanks the grid.
--@column_order
Order columns explicitly with a * wildcard for everything else: --@column_order: INVOICE_NUM, INVOICE_DATE, *, AMOUNT_PAID. Listed columns land at fixed positions; * expands to all unmentioned columns in their original CSV order.
--@format
Per-column number and date formats baked into the query: --@format: INVOICE_AMOUNT = currency(USD), INVOICE_DATE = date(MM/DD/YYYY). The currency function accepts ISO codes (USD/EUR/GBP/CAD/AUD/JPY/INR), and JPY-style zero-decimal currencies just work — Intl.NumberFormat handles the per-currency rules.
Insert from the editor
Right-click in the SQL editor → Insert directive ▸ drops a template at the top of the document with the cursor parked in the right spot. Type once, learn the syntax, type it manually next time.
Stripped from the wire, kept in history
Directives never reach the BIP gateway — we replace them with bare -- in the payload (line numbers stay aligned for error highlighting). They're preserved verbatim in query history and Saved Analyses, so reopening any past query re-applies the same presentation.
Analyses
Save the whole analysis, not just the SQL.
A Saved Analysis is a single portable <code>.dcanalysis</code> file that round-trips your SQL, formula columns, column formats, hidden columns, column order, and bind-parameter defaults. Hand one to a teammate; they open it, click Run, and see exactly what you saw.
One file, full state
JSON-on-disk with a versioned schema. SQL, formulas, hidden columns, column order, per-column number/date formats, and bind defaults all travel together. Open an analysis in a new tab and your view rebuilds itself.
Save / Save as / dirty tracking
File → Save as Analysis… creates it; Save Analysis overwrites and is gated on a real dirty diff. Tab titles show a • when there are unsaved changes, and closing a dirty tab pops a Save / Don't save / Cancel guard.
Analyses panel
Sidebar tab with search, module filter, and per-row Open / Open in new tab / Rename / Delete. Grouped by module, sorted by most-recently updated, so the analysis you touched last is always on top.
Configurable folder, shareable files
Defaults to Documents\Data Collage\Analyses\; point it at a shared OneDrive or network folder in Settings and your whole team sees the same library. Or open a one-off .dcanalysis from anywhere via File → Open Analysis….
Bind defaults travel with the analysis
The last bind values you ran with are saved alongside the SQL. Reopen the analysis and the bind dialog pre-fills — no re-typing the as-of date or business-unit ID every morning.
Save your work
History, library, snippets — all local, all yours.
No cloud sync, no telemetry. Your queries stay on your machine.
Query history
Every successful run is logged with timestamp, connection, row count, and execution time. Reopen any past query in a new tab with one click.
Query library
A curated catalog of useful starter queries you can drop into a new tab. For your own reusable work, save it as an Analysis — full state travels with it, not just the SQL.
Custom snippets & functions
Short codes (ddl → DESCRIBE …) and full multi-line templates with tabstops. Your function catalog and snippet list persist locally and follow the app installation.
AI Assist Coming soon
Generate SQL from natural language — grounded in your Fusion metadata.
A query-generation panel is wired into the UI and will go live in a near-term release. The grounding model uses the bundled Fusion metadata catalog so suggestions reference real tables and columns, not hallucinations.
Bring your own key
Plug in your own API key for OpenAI, Anthropic, or Gemini. No prompts or schemas leave your machine without your consent.
Run locally with Ollama
Prefer to keep everything on-device? Point Data Collage at a local Ollama endpoint and run a quantised model on your hardware.
Metadata-grounded prompts
The 9,000+ table/column catalog is used as retrieval context — not stuffed into every prompt — so token usage stays sane and suggestions stay accurate.
Ready to put Data Collage on your desk?
v1 is in early access. Pricing on request.