On the EXPLODE_SPLIT page (table-functions/explode-split.md, version-3.x and version-2.1), one example shows an output that contradicts its own ORDER BY:
select k1, e1 from example1 lateral view explode_split(k2, ,) tmp1 as e1 where k1 = 5 order by k1, e1;
+------+------+
| k1 | e1 |
+------+------+
| 5 | 2 |
| 5 | 3 |
| 5 | 1 |
+------+------+
k2 for k1 = 5 is 1,2,3, so explode_split yields 1, 2, 3. With ORDER BY k1, e1 the rows must come back sorted by e1 — i.e. 1, 2, 3 — which is exactly what the current engine returns (3.1.4 and 2.1.11). The documented order 2, 3, 1 is not sorted and cannot be produced by the shown query.
Suggested fix — update the expected output to the sorted order:
| 5 | 1 |
| 5 | 2 |
| 5 | 3 |
Filing separately from the missing example1 setup for this page (handled in a Line-B setup PR); all other examples on the page reproduce correctly.
On the
EXPLODE_SPLITpage (table-functions/explode-split.md, version-3.x and version-2.1), one example shows an output that contradicts its ownORDER BY:k2fork1 = 5is1,2,3, soexplode_splityields1, 2, 3. WithORDER BY k1, e1the rows must come back sorted bye1— i.e.1, 2, 3— which is exactly what the current engine returns (3.1.4 and 2.1.11). The documented order2, 3, 1is not sorted and cannot be produced by the shown query.Suggested fix — update the expected output to the sorted order:
Filing separately from the missing
example1setup for this page (handled in a Line-B setup PR); all other examples on the page reproduce correctly.