Conditional Split in ADF Data Flow issue - no rows in sink

Question

Conditional Split in ADF Data Flow issue - no rows in sink

Ulrike Pick 0

I have a Conditional Split transformation that works fine except one sink results empty although there are source rows matching the split condition.

In my scenario I have - say - 4 conditions, two of which are supposed to end up in the same table sink. Except, one doesn't, and the matching rows don't even show up in the "leftover" bucket.

Any hints would be greatly appreciated.

I suspect that the order of processing might be relevant, leading to the question of how to define the processing order.

Thanks in advance!

1 answer

Your answer

Answer 1

Venkat Reddy Navari 2,975 Microsoft External Staff Moderator

Hi @Ulrike Pick Conditional Splits in ADF can produce unexpected results when multiple output paths converge on a single sink.
Conditional Split Is Top-Down: ADF processes Conditional Split rules in order, from top to bottom. As soon as a row matches a condition, it gets routed to that stream and is not evaluated against the rest. So, if two of your conditions point to the same sink and one is broader or comes first, it could be intercepting rows meant for the other.

Troubleshooting Steps

Review Condition Order Carefully: Even small overlaps can cause a higher condition to intercept rows meant for a lower one. Try moving more specific or narrow rules to the top of the list.
Use Debug Mode + Data Preview: Enable debug mode and preview the data output from each branch of the split. This gives you visibility into where rows are actually going — or not going.
Add a “Catch-All” Rule: Add a final condition with just true as the rule. This captures anything that didn’t match the others and helps identify whether rows are falling through entirely.
Check Data Types and Comparisons: String comparisons may fail due to whitespace, casing, or implicit type mismatches (e.g. string vs. int). Watch out for those subtle issues — they’re easy to overlook.
When Multiple Conditions Should Share a Sink: Instead of sending both directly to the same sink, route them to separate outputs first, then use a Union transformation to merge them before the sink. This gives you clearer diagnostics and better control.

Hope this helps. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Ulrike Pick 0 Reputation points

2025-06-26T16:28:55.23+00:00

Thanks a lot, it will take me some time to try out your suggestions, I'll get back!
Ulrike Pick 0 Reputation points

2025-06-26T17:29:38.3033333+00:00

Hi Venkat,
what's confusing me is that the split doesn't seem to stick to the order of conditions:

The sinks are shown in the right order, but #1/3/5 are processed, while +2/4 are waiting...

Nevertheless I have defined separate sinks now, hoping for a good outcome.

String comparison I coundn't eliminate. Data Preview in Debug mode isn't working (none of the buckets shows data).

EDIT: the sequence order is a real problem, as data is being shoved into the "catch all" bucket before being matched to most of the rules!
Venkat Reddy Navari 2,975 Reputation points Microsoft External Staff Moderator

2025-06-27T10:24:07.5433333+00:00

@Ulrike Pick You're right to focus on condition order. ADF processes Conditional Split rules from top to bottom, not based on how the sinks appear visually. So even if your sink looks like it's in position 2 or 4, what matters is the actual order of the conditions inside the split.

If rows are hitting the catch all bucket too early, it usually means an earlier condition is too broad or misconfigured, especially with string comparisons (like casing or whitespace issues).

You've already done the right thing by separating the outputs. Now you can safely Union them later if needed. Since Debug Preview isn’t working, consider temporarily writing each branch to a file sink to confirm which rows are going where.

Try reordering your rules (most specific first), and normalize strings using trim() and toLower() to avoid silent mismatches.

Hope this helps. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.
Ulrike Pick 0 Reputation points

2025-06-27T10:57:28.4033333+00:00

Thanks again Venkat,

I'm afraid the problem is that I use the visual data flow editor :(
Looking into the code shows a totally different sequence... so editing the code is the appropriate way to handle splits?

EDIT:

Changing the JSON code leads to different appearance in the visual editor, but the code gets changed back into the original order after closing the code window.

Share via

Conditional Split in ADF Data Flow issue - no rows in sink

1 answer

Your answer