duckdb/arrow: CWE-407 scan — 4 DuckDB defects, 2 Arrow defects

DuckDB (C++ query engine):
- duckdb-0001: Binder::AddCorrelatedColumn vector dedup O(C²) MEDIUM 200x
- duckdb-0002: HasCorrelatedExpressions vector scan O(N×M) MEDIUM 100x
- duckdb-0003: ComputeOverlappingBindings vector scan O(N×H) MEDIUM 219x
- duckdb-0004: Deliminator group-join binding check O(G×J) MEDIUM 125x

Apache Arrow (C++ analytics):
- arrow-0001: AsofJoin IsTimeOrKeyColumn vector scan O(F×K) MEDIUM 114x
- arrow-0002: Scanner AddFieldsNeededForFilter vector dedup O(F×C) MEDIUM 250x

All 6/6 unit tests PASS.
This commit is contained in:
russell@unturf.com 2026-03-30 10:10:27 -04:00
parent 0772cf539a
commit 6b975a3b9e
13 changed files with 959 additions and 184 deletions

View file

@ -0,0 +1,136 @@
import java.util.*;
/**
* CWE-407 simulation tests for Apache Arrow defects.
*
* arrow-0001: cpp/src/arrow/acero/asof_join_node.cc
* IsTimeOrKeyColumn: std_has(key_col_index_, i) O(K) inside InitSrcToDstMapping
* loop over F fields -> O(F×K). Fix: unordered_set for O(1) lookup.
* Also MakeOutputSchema: std_has(by_field_ix, i) O(K) per field -> O(F×K).
*
* arrow-0002: cpp/src/arrow/dataset/scanner.cc
* AddFieldsNeededForFilter: std::find in options->columns O(C) per referenced
* field -> O(F×C). Fix: unordered_set for O(1) dedup.
*/
public class ArrowTest {
// =========================================================
// arrow-0001: AsofJoin IsTimeOrKeyColumn O(F×K) vs O(F)
// =========================================================
/** DEFECTIVE: linear scan of key columns for each field */
static long isTimeOrKeyDefective(int numFields, List<Integer> keyCols, int timeCol) {
long ops = 0;
for (int i = 0; i < numFields; i++) {
if (i == timeCol) continue;
// std_has: linear scan
for (int k : keyCols) {
ops++;
if (k == i) break;
}
}
return ops;
}
/** FIXED: HashSet for O(1) key membership */
static long isTimeOrKeyFixed(int numFields, List<Integer> keyCols, int timeCol) {
Set<Integer> keySet = new HashSet<>(keyCols);
long ops = keyCols.size(); // build set
for (int i = 0; i < numFields; i++) {
if (i == timeCol) continue;
ops++; // O(1) lookup
keySet.contains(i);
}
return ops;
}
static boolean testArrow0001() {
System.out.println("=== arrow-0001: AsofJoin IsTimeOrKeyColumn O(F*K) vs O(F+K) ===");
// Simulate wide schema with many key columns
int F = 500;
List<Integer> keyCols = new ArrayList<>();
for (int i = 0; i < 200; i++) keyCols.add(i + 300); // 200 key columns at high indices
long opsDefective = isTimeOrKeyDefective(F, keyCols, 0);
long opsFixed = isTimeOrKeyFixed(F, keyCols, 0);
double ratio = (double) opsDefective / opsFixed;
System.out.printf(" F=%d K=%d defective_ops=%,d fixed_ops=%,d ratio=%.1fx%n",
F, keyCols.size(), opsDefective, opsFixed, ratio);
boolean pass = ratio > 5.0;
System.out.println(" " + (pass ? "PASS" : "FAIL"));
return pass;
}
// =========================================================
// arrow-0002: Scanner AddFieldsNeededForFilter O(F×C) vs O(F+C)
// =========================================================
static class FieldPath {
final int index;
FieldPath(int i) { this.index = i; }
@Override public boolean equals(Object o) {
return o instanceof FieldPath && ((FieldPath) o).index == this.index;
}
@Override public int hashCode() { return Integer.hashCode(index); }
}
/** DEFECTIVE: std::find in columns vector per referenced field */
static long addFieldsDefective(List<FieldPath> existingColumns, List<FieldPath> fieldsReferenced) {
long ops = 0;
List<FieldPath> columns = new ArrayList<>(existingColumns);
for (FieldPath fp : fieldsReferenced) {
// linear scan for dedup
boolean found = false;
for (FieldPath c : columns) {
ops++;
if (c.equals(fp)) { found = true; break; }
}
if (!found) columns.add(fp);
}
return ops;
}
/** FIXED: unordered_set for O(1) dedup */
static long addFieldsFixed(List<FieldPath> existingColumns, List<FieldPath> fieldsReferenced) {
Set<FieldPath> existing = new HashSet<>(existingColumns);
long ops = existingColumns.size(); // build set
List<FieldPath> columns = new ArrayList<>(existingColumns);
for (FieldPath fp : fieldsReferenced) {
ops++; // O(1) lookup
if (existing.add(fp)) {
columns.add(fp);
}
}
return ops;
}
static boolean testArrow0002() {
System.out.println("=== arrow-0002: Scanner AddFieldsNeededForFilter O(F*C) vs O(F+C) ===");
int N = 500;
List<FieldPath> existingColumns = new ArrayList<>();
for (int i = 0; i < N; i++) existingColumns.add(new FieldPath(i));
List<FieldPath> fieldsReferenced = new ArrayList<>();
for (int i = N / 2; i < N + N / 2; i++) fieldsReferenced.add(new FieldPath(i)); // half overlap
long opsDefective = addFieldsDefective(existingColumns, fieldsReferenced);
long opsFixed = addFieldsFixed(existingColumns, fieldsReferenced);
double ratio = (double) opsDefective / opsFixed;
System.out.printf(" F=%d C=%d defective_ops=%,d fixed_ops=%,d ratio=%.1fx%n",
N, N, opsDefective, opsFixed, ratio);
boolean pass = ratio > 10.0;
System.out.println(" " + (pass ? "PASS" : "FAIL"));
return pass;
}
// =========================================================
public static void main(String[] args) {
int pass = 0, fail = 0;
if (testArrow0001()) pass++; else fail++;
if (testArrow0002()) pass++; else fail++;
System.out.printf("%nArrow CWE-407: %d/%d PASS%n", pass, pass + fail);
if (fail > 0) System.exit(1);
}
}