Use cache-max-... settings for the root operation, also when it's lazy
#1709
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #1709 +/- ##
==========================================
+ Coverage 89.92% 89.94% +0.01%
==========================================
Files 393 393
Lines 37601 37615 +14
Branches 4231 4232 +1
==========================================
+ Hits 33814 33834 +20
+ Misses 2486 2483 -3
+ Partials 1301 1298 -3 ☔ View full report in Codecov by Sentry. |
hannahbast
changed the title
cache_max_... settings for root operation, also when it's lazy
hannahbast
changed the title
cache_max_... settings for root operation, also when it's lazycache_max_size... settings for the root operation, also when it's lazy
hannahbast
changed the title
cache_max_size... settings for the root operation, also when it's lazycache_max_size_... settings for the root operation, also when it's lazy
hannahbast
changed the title
cache_max_size_... settings for the root operation, also when it's lazycache-max-... settings for the root operation, also when it's lazy
There was a problem hiding this comment.
This looks good, thanks a lot for the quick fix! I tried it with the example query from #1638, and that works fine now. I have a small question about the test coverage but that might be void.
| isRoot ? cache.getMaxSizeSingleEntry() | ||
| : std::min(RuntimeParameters().get<"lazy-result-max-cache-size">(), | ||
| cache.getMaxSizeSingleEntry()); |
There was a problem hiding this comment.
In the modified tests, isRoot is always false, so I wonder why CodeCov ist not complaining about incomplete coverage here.
|
So the reason why the test coverage is fine is because there are probably dozens of tests for classes inheriting from |
|
@hannahbast I added the proper unit tests |
Conformance check passed ✅No test result changes. |
|
Since #1638, QLever has the runtime parameter
lazy-result-max-cache-size, with a deliberately small default value of 5 MB. However, the final result (of the root operation) should be cached not depending on that size, but depending oncache-max-size,cache-max-size-single-entry, andcache-max-num-entries, even if the last operation is (output-)lazy. This is now indeed so. Fixes #1701 (which provides a good example of a query, where the old behavior is weird and the new behavior makes much more sense).