NationalSecurityAgency/ghidra · error · LSHException

Could not perform query on vector

Error message

Could not perform query on vector

What it means

Thrown by ExecutableComparison.processNextVectorInCluster when a QueryNearestVector (similarity search) returns null or a result set whose size is not exactly 1. This indicates the nearest-neighbor query for a given vector failed or returned an unexpected number of result groups.

Source

Thrown at Ghidra/Features/BSim/src/main/java/ghidra/features/bsim/query/client/ExecutableComparison.java:297

	 * @param workList is the current list of IDs yet to be processed for the cluster
	 * @param threshold defines how similar "close" vectors are
	 * @return the VectorResult of the next ID
	 * @throws LSHException if something goes wrong during a query
	 */
	private VectorResult queryVectorForCluster(TreeMap<Long, VectorResult> workList,
		double threshold)
		throws LSHException {
		Entry<Long, VectorResult> entry = workList.pollFirstEntry();
		VectorResult currentVector = entry.getValue();
		baseIds.remove(entry.getKey());		// Look-up vector and remove from to-do list
		queriedIds.add(entry.getKey());		// Mark that this id has been queried
		if (currentVector == null) {
			currentVector = buildSeedVector(entry.getKey());	// If VectorResult isn't present, this must be a seed for the cluster
		}
		QueryNearestVector query = buildVectorQuery(currentVector.vec, threshold);
		ResponseNearestVector response = query.execute(database);
		if (response == null || response.result.size() != 1) {
			throw new LSHException("Could not perform query on vector");
		}
		Iterator<VectorResult> iter = response.result.get(0).iterator();
		while (iter.hasNext()) {
			VectorResult vecResult = iter.next();
			Long curId = vecResult.vectorid;
			if (queriedIds.contains(curId)) {
				continue;	// Already queried
			}
			workList.put(curId, vecResult);
		}
		return currentVector;
	}

	/**
	 * Starting with the first vector in -vectorMap- build the cluster of vectors that are within
	 * a given -threshold- of each other. The cluster is the "connected" component containing the
	 * first vector, where two vectors are "connected" if they are similar to each other within
	 * the threshold. 

View on GitHub (pinned to d5f144c24d)

Solutions

1. Inspect db.getLastError() if the response was null for the specific server error.
2. Verify the vector's LSH parameters (k, L, weights) match the database configuration.
3. Retry on transient failures; reconnect if the session was dropped.
4. Check for data corruption if the error is persistent for a specific vector.

Example fix

// before
VectorResult cur = processNextVectorInCluster(workList, ...); // throws
// after
ResponseNearestVector resp = query.execute(database);
if (resp == null || resp.result.size() != 1) {
    String detail = resp == null ? database.getLastError().message : "unexpected result size";
    log.warn("Nearest vector query failed: {}", detail);
    continue; // skip this vector in the cluster
}

catch (LSHException e) { if (e.getMessage().contains("Could not perform query")) { /* log getLastError, retry once after reconnect, else skip vector */ } }

When it happens

Trigger: Executing a QueryNearestVector for a vector during cluster expansion and the response is null (query failure) or result.size() != 1 (unexpected result structure — should be exactly one result group per queried vector).

Common situations: Connection drop mid-clustering; server-side similarity index error; vector format/length mismatch with the database's LSH configuration; corrupted vector data; transient server overload.

Related errors

• {database.getLastError().message}
• Couldn't open remote program: {} for {}.
• Could not start postgres server process
• Error querying vectorid:
• No functions matching vectorid:

AI-assisted analysis of NationalSecurityAgency/ghidra@d5f144c24d (2026-08-14). Data as JSON: /api/errors/bee586b88244678a. Report an issue: GitHub.