Graphql refactor + rolling/expanding #2090
Conversation
# Conflicts: # raphtory-graphql/src/model/graph/graph.rs # raphtory-graphql/src/model/graph/nodes.rs # raphtory/src/core/utils/errors.rs
# Conflicts: # python/tests/test_base_install/test_filters/semantics/test_edge_property_filter_semantics.py # python/tests/test_base_install/test_filters/semantics/test_node_property_filter_semantics.py # python/tests/test_base_install/test_filters/test_edge_composite_filter.py # python/tests/test_base_install/test_filters/test_edge_filter.py # python/tests/test_base_install/test_filters/test_edge_property_filter.py # python/tests/test_base_install/test_filters/test_node_composite_filter.py # python/tests/test_base_install/test_filters/test_node_filter.py # python/tests/test_base_install/test_filters/test_node_property_filter.py
…as bigger than the window
miratepuffin
changed the title
python/tests/test_base_install/test_graphql/test_rolling_expanding.py
Outdated
Show resolved
Hide resolved
| self_clone.graph.write_updates()?; | ||
| let self_clone_2 = self.clone(); | ||
|
|
||
| let nodes: Vec<Result<NodeView<GraphWithVectors>, GraphError>> = |
There was a problem hiding this comment.
We should just change the type here to a HashSet instead to fix the FIXME below :)
There was a problem hiding this comment.
Also swap the Result and Vec to get early failure? Right now it will run through all the updates even if there is failures and add everything that works (maybe that is what we want?) but then the embedding part will fail early on the first broken update which doesn't make sense if we did actually want the current behaviour. We probably actually want to return the list of all failures in this case.
| impl<'graph, T: TimeOps<'graph> + Clone + 'graph> ExactSizeIterator for WindowSet<'graph, T> { | ||
| //unfortunately because Interval can change size, there is no nice divide option | ||
| fn len(&self) -> usize { | ||
| let mut cursor = self.cursor; | ||
| let mut count = 0; | ||
| while cursor < self.end + self.step { | ||
| let window_start = self.window.map(|w| cursor - w); | ||
| if let Some(start) = window_start { | ||
| if start >= self.end { | ||
| break; | ||
| } | ||
| } | ||
| count += 1; | ||
| cursor = cursor + self.step; | ||
| } | ||
| count | ||
| } |
There was a problem hiding this comment.
if we can't do better, we shouldn't implement this as it is no better than count on the iterator but I think we could do better (the months would need some special handling). At least implement the fast option when the interval is simple?
What changes were proposed in this pull request?
Removed functions/API changes
New functions
Refactor
Bug fixes
Testing
Why are the changes needed?
Improvements in the UI timeline and also useful for Graphql users.
Multiple queries being blocked as queries were being run on the main thread.
Are there any further changes required?
The spawn blocking is probably more of a patch than a final solution, but we need to get some benchmarks going first to know what works best.