Skip to content

Commit 5b0ae4a

Browse files
mstrathmanclaude
mstrathman
and
claude
committed
test: re-point the C soak drivers to the new surface (#18, #19)
`soak.c` and `soak_onnx.c` still exercised the removed old TVF forms (`predict(train,apply)`, `backtest('query',h)`), so `make test-asan` and `make test-asan-onnx` failed at runtime ("no such table: predict"). Re-point: - old `predict` TVF -> `predict_batch` (the batched / onnx / in-context serve path); added scalar `predict()` + `fit()` coverage (register, blob-via-subquery, proba, and the wrong-arity / unknown-model error paths). - old `backtest('query',h)` TVF -> the `backtest(ts,value,h)` aggregate + `backtest_rows()`; dropped the query-string-validation error cases since there is no inner query to validate now. Both soak drivers pass clean under ASan/UBSan. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
1 parent 25cf87a commit 5b0ae4a

2 files changed

Lines changed: 80 additions & 48 deletions

File tree

‎tests/soak.c‎

Lines changed: 70 additions & 38 deletions
Original file line numberDiff line numberDiff line change
@@ -129,10 +129,25 @@ int main(void) {
129129
"\"epochs\":50}')",
130130
1);
131131

132+
/* fit() aggregate: train + register native students (gbt and tree) once
133+
* here (a second register would hit MODEL_EXISTS); the scalar predict()
134+
* inside the loop serves them per row. */
135+
if (run_discard(db,
136+
"SELECT fit(f1, f2, label,"
137+
" '{\"kind\":\"gbt\",\"register\":\"soak_fit\"}') FROM tab",
138+
1) ||
139+
run_discard(db,
140+
"SELECT fit(f1, f2, label,"
141+
" '{\"kind\":\"tree\",\"register\":\"soak_fit_tree\"}')"
142+
" FROM tab",
143+
1))
144+
goto done_fail;
145+
132146
int fails = 0;
133147
for (int i = 0; i < 50; i++) {
134148
/* serving loop: the aggregate forms (the one calling convention for
135-
* forecast/detect_anomalies) + predict, across models and options */
149+
* forecast/detect_anomalies) + batched predict_batch, across models
150+
* and options */
136151
if (run_discard(db, "SELECT forecast(ts, value, 6) FROM series", 1) ||
137152
run_discard(db,
138153
"SELECT grp, forecast(ts, value, 4) FROM series"
@@ -145,36 +160,49 @@ int main(void) {
145160
" '{\"model\":\"sub-pca\"}') FROM series",
146161
1) ||
147162
run_discard(db,
148-
"SELECT * FROM predict("
163+
"SELECT * FROM predict_batch("
149164
"'SELECT f1, f2, label FROM tab WHERE id < 100',"
150165
"'SELECT id, f1, f2 FROM tab WHERE id >= 100',"
151166
" '{\"target\":\"label\"}')",
152167
1) ||
153-
run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
168+
run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
154169
" '{\"model\":\"soak_student\"}')",
155170
1))
156171
goto done_fail;
157172

158-
/* gbt-student (forest runtime) + tree/forest error paths */
159-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
173+
/* scalar predict(): serve the fit()-registered students per row, plus
174+
* proba, a fit() blob via subquery, and the arity/unknown-model errors */
175+
fails += run_discard(db, "SELECT id, predict('soak_fit', f1, f2) FROM tab", 1);
176+
fails += run_discard(db, "SELECT id, predict('soak_fit_tree', f1, f2) FROM tab", 1);
177+
fails += run_discard(db, "SELECT id, predict('soak_fit', f1, f2,"
178+
" '{\"proba\":true}') FROM tab",
179+
1);
180+
fails += run_discard(db, "SELECT id, predict((SELECT fit(f1, f2, label)"
181+
" FROM tab), f1, f2) FROM tab",
182+
1);
183+
fails += run_discard(db, "SELECT predict('soak_fit', f1) FROM tab", 0);
184+
fails += run_discard(db, "SELECT predict('nope', f1, f2) FROM tab", 0);
185+
186+
/* gbt-student (forest runtime) + tree/forest error paths, served batched */
187+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
160188
" '{\"model\":\"soak_gbt\"}')",
161189
1);
162-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
190+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
163191
" '{\"model\":\"soak_knn\"}')",
164192
1);
165-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
193+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
166194
" '{\"model\":\"soak_soft\"}')",
167195
1);
168-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
196+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
169197
" '{\"model\":\"soak_mlp\"}')",
170198
1);
171-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
199+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
172200
" '{\"model\":\"soak_bad\"}')",
173201
0);
174-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
202+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
175203
" '{\"model\":\"soak_gbt_bad\"}')",
176204
0);
177-
fails += run_discard(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM tab',"
205+
fails += run_discard(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM tab',"
178206
" '{\"model\":\"soak_mlp_bad\"}')",
179207
0);
180208
fails += run_discard(db,
@@ -185,8 +213,8 @@ int main(void) {
185213
" '{\"student_id\":\"nope\"}')",
186214
0);
187215

188-
/* auto selection, conformal intervals, backtest() (grouped, gapped) --
189-
* exercises the rolling-origin backtest scratch allocations */
216+
/* auto selection, conformal intervals, backtest() aggregate (grouped,
217+
* gapped) -- exercises the rolling-origin backtest scratch allocations */
190218
fails += run_discard(db, "SELECT forecast(ts, value, 6,"
191219
" '{\"model\":\"auto\"}') FROM series",
192220
1);
@@ -198,15 +226,19 @@ int main(void) {
198226
"\"conformal\",\"folds\":8,\"gap\":2}')"
199227
" FROM series GROUP BY grp",
200228
1);
201-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value FROM series', 6,"
202-
" '{\"folds\":10}')",
229+
fails += run_discard(db, "SELECT backtest(ts, value, 6, '{\"folds\":10}')"
230+
" FROM series",
203231
1);
204-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value FROM series', 6,"
232+
fails += run_discard(db, "SELECT backtest(ts, value, 6,"
205233
" '{\"model\":\"auto\",\"interval_method\":\"conformal\","
206-
"\"folds\":12}')",
234+
"\"folds\":12}') FROM series",
235+
1);
236+
fails += run_discard(db, "SELECT grp, backtest(ts, value, 5, '{\"gap\":3}')"
237+
" FROM series GROUP BY grp",
207238
1);
208-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value, grp FROM series',"
209-
" 5, '{\"group_cols\":[\"grp\"],\"gap\":3}')",
239+
/* backtest_rows expansion round-trip */
240+
fails += run_discard(db, "SELECT * FROM backtest_rows((SELECT backtest(ts,"
241+
" value, 6, '{\"folds\":8}') FROM series))",
210242
1);
211243
/* option error + edge paths */
212244
fails += run_discard(db, "SELECT forecast(ts, value, 6,"
@@ -215,11 +247,11 @@ int main(void) {
215247
fails += run_discard(db, "SELECT forecast(ts, value, 6, '{\"folds\":0}')"
216248
" FROM series",
217249
0);
218-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value FROM series', 6,"
219-
" '{\"model\":\"nope\"}')",
250+
fails += run_discard(db, "SELECT backtest(ts, value, 6, '{\"model\":\"nope\"}')"
251+
" FROM series",
220252
0);
221-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value FROM series', 6,"
222-
" '{\"gap\":100000}')",
253+
fails += run_discard(db, "SELECT backtest(ts, value, 6, '{\"gap\":100000}')"
254+
" FROM series",
223255
1);
224256

225257
/* tsb + auto candidate sets (statistical models; the student-candidate
@@ -262,8 +294,8 @@ int main(void) {
262294
0);
263295

264296
/* error paths every iteration too: aggregate misuse and option
265-
* rejection, plus backtest's collect_series failure branches, so
266-
* valgrind sees the partial-series cleanup under load */
297+
* rejection, plus the backtest aggregate's partial-group cleanup under
298+
* load, so valgrind sees the failure branches free everything */
267299
fails += run_discard(db, "SELECT forecast('SELECT ts FROM series', value, 3)"
268300
" FROM series",
269301
0);
@@ -287,35 +319,34 @@ int main(void) {
287319
0);
288320
fails += run_discard(db, "SELECT forecast(ts, value, 0) FROM series", 0);
289321
fails += run_discard(db, "SELECT forecast(ts, value, 4) FROM series WHERE 0", 1);
290-
fails += run_discard(db, "SELECT * FROM backtest('DELETE FROM series', 3)", 0);
291-
fails += run_discard(db, "SELECT * FROM backtest('NOT SQL', 3)", 0);
292-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts FROM series', 3)", 0);
293-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value FROM series', 3,"
294-
" '{\"time_col\":\"nope\"}')",
322+
/* backtest aggregate error + edge paths (no inner query to validate now) */
323+
fails += run_discard(db, "SELECT backtest(ts, value, 0) FROM series", 0);
324+
fails += run_discard(db, "SELECT backtest(ts, value, 3, '{\"bogus\":1}')"
325+
" FROM series",
295326
0);
296-
fails += run_discard(db, "SELECT * FROM backtest('SELECT ts, value, grp FROM series',"
297-
" 3, '{\"group_cols\":[\"nope\"]}')",
327+
fails += run_discard(db, "SELECT backtest(ts, value, 3, '{\"time_col\":"
328+
"\"nope\"}') FROM series",
298329
0);
330+
fails += run_discard(db, "SELECT backtest(ts, value, 3) FROM series WHERE 0", 1);
299331
/* duplicate option keys (the CI fuzzer's leak): last-wins, no leak */
300332
fails += run_discard(db, "SELECT forecast(ts, value, 3,"
301333
" '{\"model\":\"theta-classic\",\"model\":"
302334
"\"stub-seasonal-naive\"}') FROM series",
303335
1);
304336
fails += run_discard(db,
305-
"SELECT * FROM backtest('SELECT ts, value, grp FROM series',"
306-
" 3, '{\"model\":\"theta-classic\",\"model\":"
307-
"\"stub-seasonal-naive\",\"time_col\":\"ts\",\"time_col\":"
308-
"\"ts\",\"group_cols\":[\"grp\"],\"group_cols\":[\"grp\"]}')",
337+
"SELECT grp, backtest(ts, value, 3,"
338+
" '{\"model\":\"theta-classic\",\"model\":"
339+
"\"stub-seasonal-naive\"}') FROM series GROUP BY grp",
309340
1);
310341
fails += run_discard(db,
311-
"SELECT * FROM predict('SELECT f1, f2, label FROM tab',"
342+
"SELECT * FROM predict_batch('SELECT f1, f2, label FROM tab',"
312343
" 'SELECT id, f1, f2 FROM tab',"
313344
" '{\"target\":\"label\",\"task\":\"classify\",\"task\":"
314345
"\"classify\",\"model\":\"knn5-incontext\",\"model\":"
315346
"\"knn5-incontext\"}')",
316347
1);
317348
fails += run_discard(db,
318-
"SELECT * FROM predict('SELECT f1, f2, label FROM tab',"
349+
"SELECT * FROM predict_batch('SELECT f1, f2, label FROM tab',"
319350
" 'SELECT id, f1 FROM tab', '{\"target\":\"label\"}')",
320351
0);
321352
fails += run_discard(db, "SELECT predict_ulid('not a time')", 0);
@@ -329,6 +360,7 @@ int main(void) {
329360
1);
330361
fails += run_discard(db, "SELECT * FROM forecast_rows('not json')", 0);
331362
fails += run_discard(db, "SELECT * FROM anomaly_rows('[1,2]')", 0);
363+
fails += run_discard(db, "SELECT * FROM backtest_rows('not json')", 0);
332364
fails += run_discard(db, "SELECT * FROM forecast_rows(NULL)", 1);
333365
}
334366

‎tests/soak_onnx.c‎

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -144,29 +144,29 @@ int main(int argc, char **argv) {
144144

145145
for (int i = 0; i < 20; i++) {
146146
/* vector: success (multi-batch: 1501 rows) */
147-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
147+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
148148
" apply',json_object('model','clf'))", 1);
149149
/* introspected model + positional features */
150-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
150+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
151151
" apply',json_object('model','bare'))", 1);
152152
/* in_context: success (train context + 1501-row query, multi-batch) */
153-
fails += run(db, "SELECT * FROM predict('SELECT f1, f2, label FROM tr',"
153+
fails += run(db, "SELECT * FROM predict_batch('SELECT f1, f2, label FROM tr',"
154154
"'SELECT id, f1, f2 FROM apply',json_object('model','knn1'))", 1);
155155
/* every error branch, both layouts */
156-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
156+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
157157
" apply',json_object('model','clf','device','banana'))", 0);
158-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
158+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
159159
" apply',json_object('model','clf','device','cuda'))", 0);
160-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
160+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
161161
" apply',json_object('model','clf','precision','fp16'))", 0);
162-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1 FROM apply',"
162+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1 FROM apply',"
163163
"json_object('model','clf'))", 0);
164-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
164+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
165165
" apply',json_object('model','ghost'))", 0);
166166
/* in_context error branches: no train, missing target, bad label */
167-
fails += run(db, "SELECT * FROM predict(NULL,'SELECT id, f1, f2 FROM"
167+
fails += run(db, "SELECT * FROM predict_batch(NULL,'SELECT id, f1, f2 FROM"
168168
" apply',json_object('model','knn1'))", 0);
169-
fails += run(db, "SELECT * FROM predict('SELECT f1, f2 FROM tr',"
169+
fails += run(db, "SELECT * FROM predict_batch('SELECT f1, f2 FROM tr',"
170170
"'SELECT id, f1, f2 FROM apply',json_object('model','knn1'))", 0);
171171
/* two-head forecast (aggregate form): reconstruction success +
172172
* point/interval, then the fail-loud single-output flip declaration */

0 commit comments

Comments
 (0)