Amazon Neptune Engine version 1.3.2.1 (2024-06-20)

openCypher fixes

• A bug was detected in the query plan cache feature for parameterized queries that contain an inner WITH clause having SKIP and LIMIT as parameters. The SKIP/LIMIT values were not properly parameterized, and as a result, subsequent executions of the same cached query plan with different parameter values would still return the same results as the first execution. This has been fixed.

  # insert some nodes
  UNWIND range(1, 10) as i CREATE (s {name: i}) RETURN s
  
  # sample query
  MATCH (p) 
  WITH p ORDER BY p.name SKIP $s LIMIT $l 
  RETURN p.name as res
  
  # first time executing with {"s": 2, "l": 1}
  {
    "results" : [ {
      "res" : 3
    } ]
  }
  
  # second time executing with {"s": 2, "l": 10}
  # due to bug, produces
  {
    "results" : [ {
      "res" : 3
    } ]
  }
  # with fix, produces correct results: 
  {
    "results" : [ {
      "res" : 3
    }, {
      "res" : 4
    }, {
      "res" : 5
    }, {
      "res" : 6
    }, {
      "res" : 7
    }, {
      "res" : 8
    }, {
      "res" : 9
    }, {
      "res" : 10
    } ]
  }%

• Fixed a bug where parameterized mutation queries throw an InternalFailureException when the parameter that was passed is not already present in the database.

• Fixed a bug where parameterized Bolt queries get stuck after hitting a race condition during query resource cleanup.

General improvements

• Support for TLS version 1.3 including cipher suites TLS_AES_128_GCM_SHA256 and TLS_AES_256_GCM_SHA384. TLS 1.3 is an option - TLS 1.2 is still the minimum.

• openCypher extended support for dateime format is in lab_mode for this version. We encourage you to test it.

Gremlin improvements

• TinkerPop 3.7.x upgrade

  • Provides a large expansion of the Gremlin language.

    • New steps for processing strings, lists and dates.

    • New syntax for specifying cardinality withing the mergeV() step.

    • union() can now be used as a start step.

    • To learn more about the changes in 3.7.x, see the TinkerPop upgrade documentation.

  • When upgrading client Gremlin language drivers for Java, note that the serializer classes have ungone some renaming. You will need to update package and class naming in your configuration files and in code, if specified.

• StrictTimeoutValidation (only when enabled via labmode StrictTimeoutValidation by including StrictTimeoutValidation=enabled): When the StrictTimeoutValidation parameter has a value of enabled, a per-query timeout value specified as a request option or a query hint cannot exceed the value set globally in the parameter group. In such a case, Neptune will throw a InvalidParameterException. This setting can be confirmed in a response on the /status endpoint when the value is disabled, and in Neptune versions 1.3.2.0 and 1.3.2.1 the default value of this parameter is Disabled.

openCypher improvements

• Low latency queries and throughput performance improvement: Overall performance improvements for low latency openCypher queries. The new version also improves the throughput for such queries. The improvements are more significant when parameterized queries are used.

• Support for Query Plan Cache: When a query is submitted to Neptune, the query string is parsed, optimized, and transformed into a query plan, which then gets executed by the engine. Applications are often backed by common query patterns that are instantiated with different values. Query plan cache can reduce the overall latency by caching the query plans and thereby avoiding parsing and optimization for such repeated patterns.

• Performance Improvement for DISTINCT aggregation queries.

• Performance improvement for joins involving nullable variables.

• Performance improvement for queries involving not equals to id(node/relationship) predicate.

• Extended support for datetime functionality (Only enabled via lab mode DatetimeMillisecond by including DatetimeMillisecond=enabled. For more information, see Temporal support in the Neptune openCypher implementation (Neptune Analytics and Neptune Database 1.3.2.0 and above).

General improvements

• Updated the NeptuneML error message when validating access to Graphlytics buckets.

Gremlin fixes

• Fixed missing label information in DFE query translation, for scenarios where non-path contributing steps contain labels. For example:

  g.withSideEffect('Neptune#useDFE', true).
    V().
    has('name', 'marko').
    has("name", TextP.regex("mark.*")).as("p1").
    not(out().has("name", P.within("peter"))).
    out().as('p2').
    dedup('p1', 'p2')

• Fixed a NullPointerException bug in the DFE query translation, which occurs when a query is executed in two DFE fragments, and the first fragment is optimized to an unsatisfiable node. For example:

  g.withSideEffect('Neptune#useDFE', true).
    V().
    has('name', 'doesNotExists').
    has("name", TextP.regex("mark.*")).
    inject(1).
    V().
    out().
    has('name', 'vadas')

• Fixed a bug where Neptune could throw an InternalFailureException when a query contains ValueTraversal inside by() modulator and its input is Map. For example:

  g.V().
    hasLabel("person").
    project("age", "name").by("age").by("name").
    order().by("age")

openCypher fixes

• Improved UNWIND operations (e.g. expand a list of values into individual values) to help prevent out of memory (OOM) situations. For example:

  MATCH (n)-->(m)
  WITH collect(m) AS list
  UNWIND list AS m
  RETURN m, list

• Fixed custom id optimization in case of multiple MERGE operations where id is injected via UNWIND. For example:

  UNWIND [{nid: 'nid1', mid: 'mid1'}, {nid: 'nid2', mid: 'mid2'}] as ids
  MERGE (n:N {`~id`: ids.nid})
  MERGE (m:M {`~id`: ids.mid})

• Fixed memory explosion while planning for complex queries with property access and multiple hops with bi-directional relationships. For example:

  MATCH (person1:person)-[:likes]->(res)-[:partOf]->(group)-[:knows]-(:entity {name: 'foo'}), 
         (person1)-[:knows]->(person2)-[:likes]-(res2), (comment)-[:presentIn]->(:Group {name: 'barGroup'}), 
         (person1)-[:commented]->(comment2:comment)-[:partOf]->(post:Post), (comment2)-[:presentIn]->(:Group {name: 'fooGroup'}), 
         (comment)-[:contains]->(info:Details)-[:CommentType]->(:CommentType {name: 'Positive'}),
         (comment2)-[:contains]->(info2:Details)-[:CommentType]->(:CommentType {name: 'Positive'}) 
  WHERE datetime('2020-01-01T00:00') <= person1.addedAfter <= datetime('2023-01-01T23:59') AND comment.approvedBy = comment2.approvedBy 
  MATCH (comment)-[:contains]->(info3:Details)-[:CommentType]->(:CommentType {name: 'Neutral'})
  RETURN person1, group.name, info1.value,  post.ranking, info3.value

• Fixed aggregation queries with null as group by variables. For example:

  MATCH (n)
  RETURN null AS group, sum(n.num) AS result

SPARQL fixes

• Fixed the SPARQL parser to improve parsing time for large queries like INSERT DATA containing many triples and large tokens.

• Gremlin earliest version supported: 3.6.2

• Gremlin latest version supported: 3.7.1

• openCypher version: Neptune-9.0.20190305-1.0

• SPARQL version: 1.1