You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At the moment if you run a query like g.V('1').outE(), for every edge on the vertex '1' GafferPop will run a GetElements operation with only one seed. This can be very problematic for queries that touch a lot of data as each operation has overhead and gets logged so could end up with thousands of operations being called at once.
To fix this we might need to implement custom vertex/edge traversal strategies so that all the seeds are acquired first then added to one big operation rather than lots of smaller ones.
The text was updated successfully, but these errors were encountered:
Add a fold step before each vertex step to collect all vertex ids
so the input to the new VertexStep is a list rather than single ids.
This means we can do a single Gaffer query for all ids.
Note: input seeds are essentially deduped so expected output may change
* Gh-3277 Add custom GafferPop Vertex Step
Add a fold step before each vertex step to collect all vertex ids
so the input to the new VertexStep is a list rather than single ids.
This means we can do a single Gaffer query for all ids.
Note: input seeds are essentially deduped so expected output may change
* Add ListVertexStep
change tab width
* add unit tests
* smells
* code cov
* code cov
* code cov
* iterable
* rename vertex step
* try again
---------
Co-authored-by: wb36499 <166839644+wb36499@users.noreply.github.com>
At the moment if you run a query like g.V('1').outE(), for every edge on the vertex '1' GafferPop will run a GetElements operation with only one seed. This can be very problematic for queries that touch a lot of data as each operation has overhead and gets logged so could end up with thousands of operations being called at once.
To fix this we might need to implement custom vertex/edge traversal strategies so that all the seeds are acquired first then added to one big operation rather than lots of smaller ones.
The text was updated successfully, but these errors were encountered: