Fix handling of OPTIONAL with empty or unconnected patterns
#1847
Conversation
src/engine/NeutralOptional.cpp
Outdated
| size_t NeutralOptional::getCostEstimate() { | ||
| // Cloning is expensive, so we estimate the cost as the cost of the child. | ||
| // Does not apply to the lazy case. | ||
| return std::max<size_t>(tree_->getCostEstimate(), 1); |
There was a problem hiding this comment.
Not sure if this is the right thing to do here.
There was a problem hiding this comment.
I don't understand the comment about the cloning etc.
There also is nothing to clone here?
And like this, the operation is free, as the cost estimate always has to include the costs for the children.
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #1847 +/- ##
==========================================
+ Coverage 90.20% 90.24% +0.03%
==========================================
Files 400 401 +1
Lines 38514 38616 +102
Branches 4319 4330 +11
==========================================
+ Hits 34743 34850 +107
+ Misses 2477 2473 -4
+ Partials 1294 1293 -1 ☔ View full report in Codecov by Sentry. |
src/engine/NeutralOptional.cpp
Outdated
| size_t NeutralOptional::getCostEstimate() { | ||
| // Cloning is expensive, so we estimate the cost as the cost of the child. | ||
| // Does not apply to the lazy case. | ||
| return std::max<size_t>(tree_->getCostEstimate(), 1); |
There was a problem hiding this comment.
I don't understand the comment about the cloning etc.
There also is nothing to clone here?
And like this, the operation is free, as the cost estimate always has to include the costs for the children.
…mpty-pattern-planning
Conformance check passed ✅No test result changes. |
|
Fixes #1761, by ensuring that the
OPTIONALclause will always return at least a single value, so the cartesian product will never be empty.