initial version of neo4j driver adapter #1924
Conversation
| * @param unmasked The string to mask | ||
| * @return The masked string | ||
| */ | ||
| protected static String maskDigits(String unmasked) { |
There was a problem hiding this comment.
#nit I think this would be a good candidate to be extracted into some common NB utility so that all drivers can utilize it.
There was a problem hiding this comment.
sounds good. what place do we think if appropriate for this?
adapter-neo4j/src/main/java/io/nosqlbench/adapter/neo4j/Neo4JAdapterUtils.java
Show resolved
Hide resolved
| for (int i = 0; i < n; i++) { | ||
| try { | ||
| idx = records[i].index(fieldName); | ||
| values[i] = records[i].get(idx).asObject(); |
There was a problem hiding this comment.
If any one record does not have the required field, then the call to this method fails for all and exits the loop throwing an exception. Just verifying that this is the desired behavior as opposed to eg setting values[i] for that i to null or something and continuing iteration.
There was a problem hiding this comment.
right, this was the intended behavior for this parsing of the Record array.
I figured if the user is leveraging this method, they are actually expecting the field from all Records that Neo4J returned to them. So if we throw an exception upon not getting the field, this forces the user to go back and check their data and/or data model.
Should we add a tryToGetFieldForAllRecords method here, that just populates an index in values with null,as you've suggested, if it's not found?
|
For the failing test, you can remove the offending maven dependency and test. This was a step towards symbolic verification of our floating point rounding error, but it will have to be added back in a more suitable way when we need it. |
|
agreed, I saw Mark already removed it in his PR |
|
I pushed up a set of changes to reconcile with (more recent changes to) main so it should merge cleanly. As long as it goes green, we'll merge it. |
Apologies in advance for making this PR a single commit. This adds a Neo4J Java driver adapter to NoSQLBench. There are three op types added for running Cypher queries:
The full workload of resetting schema, creating schema, ingesting data, running ANN searches was run against a free-tier Neo4J Aura database, and all of the above appear to be working properly. One thing to keep in mind in the future is whether we are doing what is recommended for deletion of graph nodes during schema reset (a reference to support writeup is provided in the workload yaml)
In the case that we don't have the driver configured optimally, we can add in the future.