neo4j-labs / graph Goto Github PK

Any aspirations to add PCSR, maybe as an option?

From the article:
"...PCSR was orders of magnitude faster for inserts and updates than CSR
and adjacency list while maintaining similar graph traversal times."

http://supertech.csail.mit.edu/papers/WheatmanXu18.pdf

IMHO, space (memory) is easier to solve than the long time it takes to generate a graph from a large dataset.

Thoughts?

I just saw your FOSDEM presentation and am curious if you think partitioning is in-scope for this project, and if so, if you've thought about how/where it should go? My interests are in k-way partitioning similar to packages like METIS and SCOTCH, with graph coarsening and KL/FM refinement, as well as the closely related problem of computing nested dissection orderings for sparse direct solvers.

See GDS client for some inspiration:

https://github.com/neo4j/graph-data-science-client/blob/b02dbb4000cc408ebca906fe45af1fc0dd3ff141/graphdatascience/graph/nx_loader.py#L47-L129

Is it possible to alter a graph after its creation, or is a graph read-only?
I.e. is something like this possible?
https://docs.rs/dendron/0.1.5/dendron/node/struct.Node.html#method.create_as_first_child

It would make live for devs much easier, trying to figure out which library to use for their needs.

... including at least petgraph.

Hello,

I have some trouble with your librairy with graph like this one :

Node(0).
Node(1).
Node(2).
Node(3).
Node(4).
Node(5).
Node(6).
Node(7).
Node(8).
Node(9).
Node(10).
Node(11).
Node(12).
Node(13).
Node(14).
edge(12,9).
edge(11,6).
edge(5,4).
edge(12,7).

As you can see some nodes exist but aren't linked to any nodes.
I can't add them to the graph by the edges() function.
I'm also try to compute the page rank of this graph.
Is there is a trick i can do like when a node is alone it has always the same ranking or something, so that i don't have to had it it the graph ?
Or can we find a solution about that ?

Thank you

Hi there, quick question: do you have any plans to release graph-mate as conda package on conda-forge? I know I could just pip install it in a conda environment, but for my use-case a native conda package would be a better fit.

If not, no worries, I might give it a go myself. Just never packaged a package with rust bindings.

Hi there,
is there a way of building a graph where the highest node index in the edge list does not correspond to the number of vertices; thus: there are isolated vertices with higher node indices? Maybe I missed it but I did not find such a way in the documentation.

Binaries in graph_app use either u32 or usize. We should replace the usages of usize with u64 to make it more explicit.

I'm looking at graph for backing a RVSDG implementation which is centered around edge "kinds" that hold various data, ex a value edge or a control flow edge. The ability to store data within edges and to query against it effectively (sometimes you just want to see state nodes from a given source to destination) is essential for that, so that'd be a really great addition.
This also requires the ability to have multiple (possibly identical) edges going between nodes and the multiplicity of the edges being retained, if that's a concern.
Thanks!

as suggested by @knutwalker

• to_undirected in the Rust API accepts an optional layout, which is not exposed in the Python API
• Additionally, allow relabelling in the same step
• Maybe we can extend the Layout enum to smth like { Unsorted, SortedById, SortedByDegree, DeduplicatedById, DeduplicatedByDegree } ?