Add a demo notebook that demonstrates loading from and saving to neo4j #1184
Conversation
|
Check out this pull request on You'll be able to see Jupyter notebook diff and discuss changes. Powered by ReviewNB. |
|
Code Climate has analyzed commit 4c5bffa and detected 0 issues on this pull request. View more on Code Climate. |
huonw
changed the title
| @@ -0,0 +1,2038 @@ | |||
| { | |||
There was a problem hiding this comment.
We can use single Cypher query to retrieve an identifier for the source and target of each node is sufficient.
"We can use a single Cypher query to retrieve the identifiers of the source and target nodes of each edge."?
ensure the DataFrame columns have the easiest name.
"ensure the DataFrame columns match the defaults"?
Reply via ReviewNB
There was a problem hiding this comment.
Wow, I really mangled that first sentence!
| @@ -0,0 +1,2038 @@ | |||
| { | |||
There was a problem hiding this comment.
One way sort of a subgraph someone might be interested is one where the nodes and/or edges satisfy certain criteria.
"One type of subgraph someone might be interested..."?
Also in general, this Subgraphs sections feels more like a cypher tutorial to me, rather than anything stellargraph specific - I understand that it comes from us believing it would be a common use-case since you probably wouldn't load ALL your neo4j data into memory, but the exact way the user should reduce their graph is probably going to be very specific to their data. Thoughts?
Reply via ReviewNB
There was a problem hiding this comment.
This ~whole notebook is a cypher tutorial, since it's actually not doing anything StellarGraph-y that isn't already explained in loading-pandas.ipynb. I agree that the exact reduction is specific to their data and their use case, but I feel that "filter the elements (in isolation)" and "k-hop subgraphs" are likely to be the most common (or at least, serve as close-enough examples) and so we can spell them out to get people started, and reduce the barriers to working with StellarGraph.
There was a problem hiding this comment.
This ~whole notebook is a cypher tutorial, since it's actually not doing anything StellarGraph-y that isn't already explained in loading-pandas.ipynb.
I guess so, but there's some relationship between the query being run and the way the stellargraph constructor is used that the notebook is demonstrating (e.g. run a query to obtain heterogeneous graph then use it to create a heterogeneous stellargraph), whereas for the subgraph one feels more like it's purely in cypher (i.e. stellargraph doesn't know/care that it's a subgraph). That said, I'm fine with keeping it since I agree it's likely to be a common use-case.
There was a problem hiding this comment.
Ah, I understand the distinction you were seeing, and it makes sense to me now. I'll leave it as is, but we can always rethink.
There was a problem hiding this comment.
This is an excellent tutorial both for StellarGraph and Cypher! I do feel that the k-hop subgraph section goes beyond being a "basic" tutorial, however I think this notebook is the best place to put it for now.
Hopefully we can land #1183 before release to remove the FIXME's.
This adds a notebook to
basics/that demonstrates a few ways one might load graphs and subgraphs from Neo4j into memory (as Pandas DataFrames) and then convert those intoStellarGraphobjects. It follows basically the same structure as the Pandas and NetworkX ones, except has an additional "save results" section at the end too.See: #1034 and #1037