Troubleshooting & Common Pitfalls

A round-up of the things that most commonly trip people up in Data Collage, grouped by area. Each entry follows the same shape: symptom → likely cause → what to try.

If you hit something that isn’t covered here, drop us a note at datacollage@newarcinc.com with as much detail as you can.

Connections and setup

Test connection fails

The three usual causes, in roughly the order to check them:

Wrong Fusion URL. Should be the home URL, like https://fa-xxxx-saasfaprod1.fa.ocs.oraclecloud.com — no trailing slash, no /fscmUI/, no /sales/ path. Confirm by loading the URL in your browser; the Fusion home page should appear.
Wrong or expired password. Try logging into Fusion in your browser with the same credentials to confirm. If you’re forced to reset, come back and update the password in Data Collage.
Missing BI Publisher access. Your Fusion user needs to be able to open BI Publisher reports from the BI Publisher Catalog. Test that in your browser first.

See §4 — Step 2 for the full 6-step test breakdown.

Usually a corporate browser/security policy blocking the embedded webview Data Collage uses for SSO. Things to try:

Retry from a different network (off-VPN, off-corporate-proxy) to confirm whether your network is in the way.
Restart Data Collage and try once more — sometimes the webview just needs a clean start.
If neither helps, contact your IT team about whitelisting embedded webviews for desktop applications.

SSO keeps re-prompting mid-session

That’s expected behavior when Fusion’s session token expires or is rejected by BI Publisher. Log in again when the window pops; your query is retried automatically once authentication succeeds.

”Save” button stays disabled in the connection dialog

Test connection (or Verify SSO login) hasn’t run successfully against the current values in the form. Any edit invalidates the previous Test result — re-run Test after changing anything, and Save will activate.

SQL execution errors

”Statement is not allowed” / read-only validator rejection

Data Collage only allows SELECT and WITH statements. INSERT, UPDATE, DELETE, and DDL are blocked at the validator before your query ever leaves your machine. This is by design — see §1. Modifying Fusion data should be done through Fusion’s own configuration UIs, FBDI loaders, or REST APIs — not via SQL through Data Collage.

Common Oracle error codes

When a query fails, the error panel shows a short summary and any ORA- codes returned by Fusion. The most common ones:

Code	What it usually means	What to try
`ORA-00942`	Table or view does not exist	Almost always missing access, not a missing table. Confirm you can query it in BI Publisher first.
`ORA-00904`	Invalid identifier	A typo in a column name, or referring to a column from the wrong table alias.
`ORA-01017`	Invalid username/password	Re-open the connection dialog and re-test; the Fusion password may have changed.
`ORA-00936`	Missing expression	A trailing comma, an unclosed parenthesis, or an empty `SELECT` list.
`ORA-00933`	SQL command not properly ended	A stray `;` inside a single statement, or the wrong statement under your cursor.

SOAP timeouts on large extracts

If a query has been running for a long time and eventually fails with a timeout-shaped error, your result set is probably too large for BI Publisher’s synchronous gateway. Try:

Narrow the date range in your WHERE clause.
Add ROWNUM <= 10000 (or similar) to cap the result while you iterate.
For a true bulk extract, export from BIP directly rather than through Data Collage.

Bind variable dialog pops up unexpectedly

You have a stray :name-style placeholder somewhere in your SQL — often inside a string literal that has a : in it (e.g. a time format like 'HH24:MI:SS'). Either properly quote/escape, or rename the placeholder.

Formula columns

Result shows `∅` even though the inputs look fine

NULL is propagating through your formula. Any NULL input makes the whole expression NULL — this is SQL semantics, not Excel semantics. Wrap any column that might be NULL with NVL(col, 0):

amount_due = invoice_amount - NVL(amount_paid, 0)

See §8 — Working with NULLs for the longer explanation.

Save rejected with a “cycle” error

Two of your formulas reference each other (directly or transitively). For example, A = B + 1 and B = A * 2. Untangle the chain — at least one formula needs to depend only on raw columns, not on another formula that depends on it.

Formula won’t accept a column you expected to use

You’re referring to a non-numeric column (typically a VARCHAR2 or DATE). Formula columns operate only on numeric values in V1. Options:

Cast the column in your SQL: TO_NUMBER(col).
Use a different (already-numeric) column.

A formula disappeared after re-running the SQL

The column it referenced was renamed or removed from your query. Data Collage auto-drops formulas whose references can’t be resolved and shows a toast when it does. To fix:

Either edit the formula to use the new column name.
Or add the original column back to your SELECT list.

Result grid

Find Duplicates does nothing

You opened the Duplicates dropdown but didn’t tick any columns to define the duplicate key. Tick one or more columns first, then click Find duplicates.

Top N button is disabled

Your current result set has no numeric columns — Top N needs at least one numeric column to rank by. Add a numeric column to your SELECT (or use a formula column to compute one) and try again.

Columns “disappeared” from my grid

They’re hidden, not gone. Open the Columns picker on the toolbar and look for unticked entries — or just click Show all if you don’t remember what you hid.

Sort or filter state didn’t survive Save Analysis

By design. Saved Analyses store SQL, formulas, column formats, hidden columns, column order, and bind defaults — but not sort, filters, Find Duplicates, or Top N state. Those are presentational only.

If you need persistent filtering, do it in the SQL with a WHERE clause.

Saved Analyses

Analysis won’t run after opening

No connection is selected — either no pinned default is set, or you’ve opened the analysis in a session where no connection is active. Open the Connections panel and click a connection to make it active.

The current tab isn’t backed by a .dcanalysis file yet. Use Save as Analysis… the first time to create the file; afterwards, Save Analysis is enabled and writes back to that same file.

Bind defaults didn’t pre-fill

The analysis was saved before you supplied bind values (or you cleared the values). Run the query once, fill in the bind dialog, then Save Analysis — the values you entered are now persisted with the analysis and will pre-fill on the next open.

Exports

Excel cells show numbers as text

Either Apply column formats was off during export, or the column wasn’t detected as numeric. Things to try:

Open the Export menu and confirm Apply column formats is checked.
If a column should be numeric but isn’t being recognized, add an explicit cast in your SQL (TO_NUMBER(col) or CAST(col AS NUMBER)).

A large export seems to hang

Wait for the success toast before doing anything else with the file. Tens of thousands of rows can take several seconds to encode and write to disk — opening the file in Excel mid-write produces unpredictable results.

If you’ve waited 30+ seconds and nothing has happened, try a smaller export to confirm the basics are working, then come back to the large one.

Exported file has more rows than I expected

Grid filters — Find Duplicates, Top N, per-column filters — and sort state do not carry over to exports. The export is always the full result set Fusion returned, in original row order.

To export a filtered subset, do the filtering in SQL with a WHERE clause (and the sort with ORDER BY) and re-run.

Reporting bugs

If you hit something not covered here — or you’ve tried the suggestions and they didn’t help — please write to us at datacollage@newarcinc.com.

The more of the following you can include, the faster we can help:

Data Collage version — find it in Settings → About.
Windows version — Windows 10 or 11, and the build number if you know it.
What you were trying to do — short description of the goal.
Exact steps that reproduce the issue — even if it feels obvious.
What you expected vs what actually happened.
The exact error message, if any — copy/paste rather than paraphrase.
A screenshot if the issue is visual or hard to describe in words.

We read every message, and detailed reports turn into fixes faster than vague ones — your effort here directly speeds up the next release.