Whenever a specific combination of query and indexes leads to in-memory sorting/ordering of quads, the entire result set needs to be buffered. This may lead to crashes if the result set is big enough to make the process run out of memory, even in the absence of an actual bug.
We should have configurable thresholds and failure strategies to make the store behave in a predictable and easy-to-understand fashion. The threshold should be expressed as the maximum number of quads that the store may buffer. The set of strategies could be comprised of:
- crashing the process
- throwing an exception
- print a warning to
stdout or stderr
Whenever a specific combination of query and indexes leads to in-memory sorting/ordering of quads, the entire result set needs to be buffered. This may lead to crashes if the result set is big enough to make the process run out of memory, even in the absence of an actual bug.
We should have configurable thresholds and failure strategies to make the store behave in a predictable and easy-to-understand fashion. The threshold should be expressed as the maximum number of quads that the store may buffer. The set of strategies could be comprised of:
stdoutorstderr