Skip to content

Commit 8b4e0f7

Browse files
dfa1claude
dfa1
and
claude
committed
test: triage 14 Raincloud slugs → ok; glove/osm → gap:257; add size comparison
- Mark bi-physicians, bi-redfin4, bi-romance, bi-tablerosistemapenal, bi-trainsuk1, bi-uberlandia, goodbooks-10k, uk-road-safety-accidents-and-vehicles as ok (triage run) - Mark glove-6b-50d/100d/200d and osm-germany-relations as gap:257 (CsvExporter lacks FixedSizeList/List support) - Fix oracle header: use fieldPath().topLevelName() instead of col.name() so list-column parquet schemas (e.g. vector.list.element → vector) match the Vortex column name - Abort oracle early (before writing header) when any column has maxRepetitionLevel > 0, preventing header-count mismatch and pipe deadlock - Update error-propagation priority: VortexException > oracle abort > mismatch - Add @execution(CONCURRENT) to conformancePerSlug() + junit-platform.properties for parallel dynamic test execution - Add RaincloudSizeComparisonIntegrationTest: reports parquet / vortex-jni / vortex-java file sizes per slug (ParquetImporter re-encode; N/A for nested schemas) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
1 parent aaec68e commit 8b4e0f7

4 files changed

Lines changed: 145 additions & 18 deletions

File tree

‎integration/src/test/java/io/github/dfa1/vortex/integration/RaincloudConformanceIntegrationTest.java‎

Lines changed: 30 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,8 @@
1313
import org.junit.jupiter.api.DynamicTest;
1414
import org.junit.jupiter.api.Test;
1515
import org.junit.jupiter.api.TestFactory;
16+
import org.junit.jupiter.api.parallel.Execution;
17+
import org.junit.jupiter.api.parallel.ExecutionMode;
1618
import org.opentest4j.TestAbortedException;
1719

1820
import java.io.BufferedReader;
@@ -65,6 +67,7 @@ void corpusIsHydrated() {
6567
}
6668

6769
@TestFactory
70+
@Execution(ExecutionMode.CONCURRENT)
6871
Stream<DynamicTest> conformancePerSlug() throws IOException {
6972
Path manifest = manifestPath();
7073
if (!Files.exists(manifest)) {
@@ -128,12 +131,12 @@ private static void assertMatchesParquetOracle(Path vortex, Path parquet) throws
128131
Thread.currentThread().interrupt();
129132
}
130133

131-
// Propagate in priority order: mismatch > decode gap > oracle abort
134+
// Propagate in priority order: decode gap > oracle abort > mismatch.
135+
// Oracle abort ranks above mismatch because a spurious AssertionError("oracle ended
136+
// before vortex output") is produced whenever the oracle aborts before writing all
137+
// rows — that is an oracle limitation, not a conformance failure.
132138
Throwable oe = oracleError.get();
133139
Throwable ve = vortexError.get();
134-
if (mainError instanceof AssertionError e) {
135-
throw e;
136-
}
137140
if (ve instanceof VortexException e) {
138141
throw e;
139142
}
@@ -143,6 +146,9 @@ private static void assertMatchesParquetOracle(Path vortex, Path parquet) throws
143146
if (oe instanceof TestAbortedException e) {
144147
throw e;
145148
}
149+
if (mainError instanceof AssertionError e) {
150+
throw e;
151+
}
146152
if (oe != null) {
147153
throw new TestAbortedException("oracle cannot read the parquet sibling: " + oe);
148154
}
@@ -281,14 +287,26 @@ private static void writeOracleCsv(Path parquet, Writer out) throws IOException
281287
CsvWriter csv = CsvWriter.builder().fieldSeparator(',').build(out)) {
282288

283289
List<ColumnSchema> cols = pfr.getFileSchema().getColumns();
290+
// Abort before writing anything if the schema has repeated (list/map) columns.
291+
// Such columns cannot be read as scalar values, and leaving the oracle running
292+
// would create a header-column-count mismatch that would deadlock the pipe.
293+
for (ColumnSchema col : cols) {
294+
if (col.maxRepetitionLevel() > 0) {
295+
throw new TestAbortedException(
296+
"oracle cannot format repeated list column: " + col.fieldPath().topLevelName());
297+
}
298+
}
284299
// De-duplicate duplicate column names with the Rust Vortex writer's algorithm:
285300
// the Nth (N >= 1) occurrence of a base name gets a " [N]" suffix, matching
286301
// the de-duplicated names in the Vortex file (#256).
302+
// Use the top-level logical name (e.g. "vector") rather than the leaf name
303+
// (e.g. "element" inside a LIST group) so the header matches the Vortex column name.
287304
Map<String, Integer> seen = new LinkedHashMap<>();
288305
List<String> header = new ArrayList<>(cols.size());
289306
for (ColumnSchema col : cols) {
290-
int count = seen.merge(col.name(), 1, Integer::sum) - 1;
291-
header.add(count == 0 ? col.name() : col.name() + " [" + count + "]");
307+
String logicalName = col.fieldPath().topLevelName();
308+
int count = seen.merge(logicalName, 1, Integer::sum) - 1;
309+
header.add(count == 0 ? logicalName : logicalName + " [" + count + "]");
292310
}
293311
csv.writeRecord(header);
294312

@@ -319,6 +337,12 @@ private static void writeOracleCsv(Path parquet, Writer out) throws IOException
319337
/// @return the formatted cell string
320338
private static String oracleCell(ColumnSchema col, RowReader rows) {
321339
int idx = col.columnIndex();
340+
// Repeated list columns (maxRepetitionLevel > 0) cannot be read as a single scalar value.
341+
// Abort rather than silently reading only the first element.
342+
if (col.maxRepetitionLevel() > 0) {
343+
throw new TestAbortedException(
344+
"oracle cannot format repeated list column: " + col.fieldPath().topLevelName());
345+
}
322346
if (col.repetitionType() == RepetitionType.OPTIONAL && rows.isNull(idx)) {
323347
return "";
324348
}

‎integration/src/test/java/io/github/dfa1/vortex/integration/RaincloudSizeComparisonIntegrationTest.java‎

Lines changed: 96 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,96 @@
1+
package io.github.dfa1.vortex.integration;
2+
3+
import io.github.dfa1.vortex.parquet.ParquetImporter;
4+
import org.junit.jupiter.api.DynamicTest;
5+
import org.junit.jupiter.api.Test;
6+
import org.junit.jupiter.api.TestFactory;
7+
import org.junit.jupiter.api.parallel.Execution;
8+
import org.junit.jupiter.api.parallel.ExecutionMode;
9+
10+
import java.io.IOException;
11+
import java.nio.file.Files;
12+
import java.nio.file.Path;
13+
import java.util.stream.Stream;
14+
15+
import static org.junit.jupiter.api.Assumptions.assumeTrue;
16+
17+
/// File-size comparison across the Raincloud corpus for three representations:
18+
/// Parquet (corpus oracle), vortex-jni (corpus file, written by Rust Vortex), and
19+
/// vortex-java (re-encoded from the Parquet oracle via [ParquetImporter]).
20+
///
21+
/// Output is purely informational — one `[RaincloudSizes]` line per slug.
22+
/// Fails only if the vortex-java file is unreadable when [ParquetImporter] succeeds.
23+
/// Slugs with nested Parquet schemas (LIST, MAP, STRUCT) log `vortex-java=N/A` and pass.
24+
///
25+
/// Skipped (visibly) when the corpus is not hydrated — run
26+
/// `scripts/hydrate-raincloud-corpus.sh` first, or point `RAINCLOUD_CORPUS_MANIFEST`
27+
/// at a manifest TSV (`slug<TAB>vortex-path<TAB>parquet-path` per line).
28+
class RaincloudSizeComparisonIntegrationTest {
29+
30+
private static final Path DEFAULT_MANIFEST =
31+
Path.of(System.getProperty("user.home"), ".cache", "raincloud", "corpus-manifest.tsv");
32+
33+
@Test
34+
void corpusIsHydrated() {
35+
// Given / When / Then — visible skip marker when the corpus is absent
36+
assumeTrue(Files.exists(manifestPath()),
37+
"raincloud corpus not hydrated — run scripts/hydrate-raincloud-corpus.sh");
38+
}
39+
40+
@TestFactory
41+
@Execution(ExecutionMode.CONCURRENT)
42+
Stream<DynamicTest> sizePerSlug() throws IOException {
43+
Path manifest = manifestPath();
44+
if (!Files.exists(manifest)) {
45+
return Stream.empty();
46+
}
47+
return Files.readAllLines(manifest).stream()
48+
.filter(line -> !line.isBlank())
49+
.map(line -> line.split("\t"))
50+
.map(parts -> DynamicTest.dynamicTest(parts[0], () -> {
51+
String slug = parts[0];
52+
Path vortexJni = Path.of(parts[1]);
53+
Path parquet = Path.of(parts[2]);
54+
55+
long parquetBytes = Files.size(parquet);
56+
long jniBytes = Files.size(vortexJni);
57+
58+
Path javaVortex = Files.createTempFile("raincloud-java-" + slug + "-", ".vortex");
59+
try {
60+
String javaLabel;
61+
try {
62+
ParquetImporter.importParquet(parquet, javaVortex);
63+
long javaBytes = Files.size(javaVortex);
64+
javaLabel = String.format("%.2fMB (%.2fx jni)",
65+
javaBytes / 1_048_576.0, (double) javaBytes / jniBytes);
66+
} catch (UnsupportedOperationException | IllegalArgumentException e) {
67+
javaLabel = "N/A (" + firstSentence(e.getMessage()) + ")";
68+
}
69+
70+
System.out.printf("[RaincloudSizes] %-55s parquet=%7.2fMB vortex-jni=%7.2fMB vortex-java=%s%n",
71+
slug,
72+
parquetBytes / 1_048_576.0,
73+
jniBytes / 1_048_576.0,
74+
javaLabel);
75+
} finally {
76+
Files.deleteIfExists(javaVortex);
77+
}
78+
}));
79+
}
80+
81+
private static Path manifestPath() {
82+
String env = System.getenv("RAINCLOUD_CORPUS_MANIFEST");
83+
return env != null ? Path.of(env) : DEFAULT_MANIFEST;
84+
}
85+
86+
/// Returns the first sentence of a message (up to the first `;` or `.`), for compact log lines.
87+
private static String firstSentence(String msg) {
88+
if (msg == null) {
89+
return "null";
90+
}
91+
int semi = msg.indexOf(';');
92+
int dot = msg.indexOf('.');
93+
int end = semi >= 0 && dot >= 0 ? Math.min(semi, dot) : semi >= 0 ? semi : dot;
94+
return end > 0 ? msg.substring(0, end) : msg;
95+
}
96+
}

‎integration/src/test/resources/junit-platform.properties‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,7 @@
1+
# Enable JUnit 5 parallel execution; individual tests opt in via @Execution(CONCURRENT).
2+
# Default is same_thread so existing integration tests are unaffected.
3+
junit.jupiter.execution.parallel.enabled=true
4+
junit.jupiter.execution.parallel.mode.default=same_thread
5+
junit.jupiter.execution.parallel.mode.classes.default=same_thread
6+
junit.jupiter.execution.parallel.config.strategy=dynamic
7+
junit.jupiter.execution.parallel.config.dynamic.factor=1

‎integration/src/test/resources/raincloud/expected-status.csv‎

Lines changed: 12 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -48,23 +48,23 @@ bi-mulheresmil,ok
4848
bi-nyc,untriaged
4949
bi-pancreactomy1,ok
5050
bi-pancreactomy2,untriaged
51-
bi-physicians,untriaged
51+
bi-physicians,ok
5252
bi-provider,untriaged
5353
bi-realestate1,untriaged
5454
bi-realestate2,untriaged
5555
bi-redfin1,untriaged
5656
bi-redfin2,untriaged
5757
bi-redfin3,untriaged
58-
bi-redfin4,untriaged
58+
bi-redfin4,ok
5959
bi-rentabilidad,untriaged
60-
bi-romance,untriaged
60+
bi-romance,ok
6161
bi-salariesfrance,untriaged
62-
bi-tablerosistemapenal,untriaged
62+
bi-tablerosistemapenal,ok
6363
bi-taxpayer,untriaged
6464
bi-telco,untriaged
65-
bi-trainsuk1,untriaged
65+
bi-trainsuk1,ok
6666
bi-trainsuk2,untriaged
67-
bi-uberlandia,untriaged
67+
bi-uberlandia,ok
6868
bi-uscensus,untriaged
6969
bi-wins,untriaged
7070
bi-yalelanguages,ok
@@ -105,10 +105,10 @@ frames-benchmark,untriaged
105105
ghcn-daily,untriaged
106106
glass,ok
107107
global-fossil-co2-emissions-by-country-2002-2022,untriaged
108-
glove-6b-100d,untriaged
109-
glove-6b-200d,untriaged
110-
glove-6b-50d,untriaged
111-
goodbooks-10k,untriaged
108+
glove-6b-100d,gap:257
109+
glove-6b-200d,gap:257
110+
glove-6b-50d,gap:257
111+
goodbooks-10k,ok
112112
google-cluster-trace-2011-machine-events,ok
113113
green_tripdata_2025,ok
114114
gsm8k,untriaged
@@ -156,7 +156,7 @@ openlibrary-works,untriaged
156156
openorca,untriaged
157157
openpowerlifting,ok
158158
osm-germany-nodes,untriaged
159-
osm-germany-relations,untriaged
159+
osm-germany-relations,gap:257
160160
osm-germany-ways,untriaged
161161
osmi-mental-health-in-tech-2016,untriaged
162162
osmi-mental-health-in-tech-2017,untriaged
@@ -239,7 +239,7 @@ uci-wine,ok
239239
uci-wine-quality,ok
240240
ufcdata,untriaged
241241
uk-price-paid,untriaged
242-
uk-road-safety-accidents-and-vehicles,untriaged
242+
uk-road-safety-accidents-and-vehicles,ok
243243
ultrachat-200k,untriaged
244244
ultrafeedback-binarized,untriaged
245245
us-accidents,untriaged

0 commit comments

Comments
 (0)