Skip to content

Releases

v0.7.0

Choose a tag to compare

@github-actions github-actions released this 11 Nov 07:00
6cf6df2

What's Changed

  • Refuse compiled Cypher on a SQLite older than 3.53.2, and keep plain SQL and the schema manager working there (#565) by @jasperblues in #735
  • Wait for an index to be built before ensure returns, on Neo4j and FalkorDB, so a write can't crash Neo4j as the index comes online by @jasperblues in #736
  • Add the drivine CLI's skeleton, built as a native image, with its own CI job on every store (#706) by @jasperblues in #740
  • Make the index ONLINE before the faulty wait in the await test, so it can't race the build by @jasperblues in #746
  • Read each self-loop from its one node in the write-loop checks, and say to analyze a Postgres graph after a reload by @jasperblues in #747
  • Pin the Cypher compiler to 0.36.0 and run the Cypher spec at v0.36.0 by @jasperblues in #748
  • Close a Postgres session out of its pool after it builds an index, so its shared memory goes with it by @jasperblues in #749
  • Set a Postgres session's parallel workers back as they were after an index build with none by @jasperblues in #752
  • Join the application's Spring or JTA transaction, rather than commit it partway, in a Drivine transaction on its connection by @jasperblues in #753
  • Take the store layout's DDL from the shared module's generators, and run the spec at v0.37.0 (liberation-data/drivine-cypher-spec#105) by @jasperblues in #754
  • Take SQLite's write lock before a write outside a transaction reads, so writers on many connections wait their turn rather than run out of tries by @jasperblues in #756
  • Take the _DrivineSchema marker, the store stamp and the node stamp from the shared module, and run the spec at v0.38.0 (liberation-data/drivine-cypher-spec#106) by @jasperblues in #759
  • drivine4j onto compiler v0.39.0: the spec at v0.39.0, the module's schema types, and the tests for #517, #154 and #156 by @jasperblues in #755
  • Run the Cypher spec's search.yaml on every store, and read the ranking corpus from the spec by @jasperblues in #772
  • Take full-text search from the shared module: query parsing, BM25, search SQL and graph-engine heads (drivine-cypher-spec#100, step 3) by @jasperblues in #773
  • Say what is known of Memgraph's several-properties hold: seen on the arm64 box only, cause under investigation (#762) by @jasperblues in #774
  • Give a query's columns and its nodes, relationships and paths with their kinds: PersistenceManager.queryResult (#741) by @jasperblues in #776
  • Run Memgraph's search spec on a container of its own and hold its several-properties steps everywhere, naming the cause (memgraph/memgraph#5353) by @jasperblues in #775
  • drivine CLI: print columns, nodes, relationships and paths from queryResult (#741, step 3) by @jasperblues in #778
  • drivine CLI: run a script's statements, from -f, stdin or the argument, split by a lexer pass, with --continue (#709, slice 1) by @jasperblues in #782
  • Drive schema management with the shared module's planner, so drivine4j runs each round and the module decides (liberation-data/drivine-cypher-spec#135) by @jasperblues in #783
  • drivine CLI: :begin/:commit/:rollback, :param, :params, :source and -P, with --continue inside a transaction (#709, slice 2) by @jasperblues in #786
  • Run the spec at v0.42.0 on the 0.42.0 compiler, and search a label that needs escaping by @jasperblues in #787
  • Release v0.7.0 by @jasperblues in #788

Full Changelog: v0.6.2...v0.7.0

Part of Drivine 2026.11.6, with drivine 5.0.0, drivine-python 0.8.0 and Cypher spec v0.42.0.

v0.6.2

Choose a tag to compare

@github-actions github-actions released this 09 Nov 23:48
04e77ca

What's Changed

  • Don't take a supertype's private @nodeid as a subclass's node id, as the code generator doesn't (#628) by @jasperblues in #674
  • Split CI's tests over parallel jobs, and move the slow suites to a nightly by @jasperblues in #672
  • Build the Cypher compiler from drivine-cypher-spec's compiler module, in place of drivine4j's own copy by @jasperblues in #662
  • Run the Cypher spec at v0.31.0, the compiler release drivine4j pins by @jasperblues in #691
  • Pin the Cypher compiler to 0.32.0 and run the Cypher spec at v0.32.0 by @jasperblues in #705
  • Run the Cypher spec's host.yaml by @jasperblues in #671
  • Run host.yaml's regexp cases: allow the name, and pair them by SQL that holds the REGEXP operator by @jasperblues in #715
  • Make SQLite's graph ids AUTOINCREMENT so none is used again, and refuse a file an earlier Drivine made (#713) by @jasperblues in #721
  • Bind an Iterator or once-only Sequence nested in a compiled statement's parameter as its items (#722) by @jasperblues in #723
  • Commit a statement Drivine runs outside a transaction on a pool whose connections commit nothing, and leave a Spring transaction's connection to Spring (#395) by @jasperblues in #725
  • Write a @caseInsensitive text's copy as a CLOB on Oracle, so a long one saves, fills and is found (#532) by @jasperblues in #728
  • Pin the Cypher compiler to 0.35.0 and run the Cypher spec at v0.35.0 by @jasperblues in #731
  • Release v0.6.2 by @jasperblues in #733

Full Changelog: v0.6.1...v0.6.2

Part of Drivine 2026.11.5, with drivine 4.0.0, drivine-python 0.6.0 and Cypher spec v0.35.0.

v0.6.1

Choose a tag to compare

@github-actions github-actions released this 08 Nov 01:48
471a82b

What's Changed

  • Count a term written again at once each time, not every other time, on the SQL stores (#607) by @jasperblues in #619
  • Score a prefix term the same for each node that holds it, as Lucene's prefix query does, on the SQL stores (#612) by @jasperblues in #623
  • Read a query's words as the text's expression does, Alphabetic letters, decimal digits and underscores (#616) by @jasperblues in #627
  • Find an override's node id through its supertypes, not KSP2's findOverridee, which at times found nothing (#621) by @jasperblues in #629
  • Read a * inside quotes as text, as Lucene's parser does, so "of sou*" is the phrase "of sou" (#622) by @jasperblues in #632
  • Say that the SQL stores split a dotted name into words, where Neo4j's default analyzer keeps it one token by @jasperblues in #634
  • Parse Cypher with ANTLR's two-stage SLL-then-LL prediction, so a compile spends less time in the parser (#655) by @jasperblues in #656
  • Run each schema DDL command on a SQL store on one connection, not one for each catalog read by @jasperblues in #660
  • Release v0.6.1 by @jasperblues in #664

Full Changelog: v0.6.0...v0.6.1

Part of Drivine 2026.11.4, with drivine 3.6.0, drivine-python 0.5.0 and Cypher spec v0.30.0.

v0.6.0

Choose a tag to compare

@github-actions github-actions released this 06 Nov 23:08
9c9a0ea

What's Changed

  • Plan a compiled write's per-row loop on Postgres without materializing, so a node table whose statistics say it is near empty isn't read whole for each row (#631) by @jasperblues in #637
  • Read a node a pattern finds by a row's or a parameter's value through the property's index on Postgres by @jasperblues in #641
  • Let a keyword name a variable, as Neo4j does, so APOC's MATCH (end:...) compiles by @jasperblues in #640
  • Run Neo4j schema DDL through the schema manager on SQL stores (#636) by @jasperblues in #642
  • Read a row's node or relationship back by a probe of its key on Postgres (#644) by @jasperblues in #645
  • Release v0.6.0 by @jasperblues in #647

Full Changelog: v0.5.0...v0.6.0

Part of Drivine 2026.11.3, with drivine 3.5.1, drivine-python 0.4.1 and Cypher spec v0.30.0.

v0.5.0

Choose a tag to compare

@github-actions github-actions released this 06 Nov 02:23
0086aaa

What's Changed

  • Read a property of a node or a map a function gives by a filtered path on Oracle, so a text past 32 KB over two rows doesn't close the session (#531) by @jasperblues in #535
  • Number a year before 1 as Neo4j does on Postgres, Oracle and SQLite by @jasperblues in #487
  • Read the kind and take out the text reverse() gives through its text on Oracle, so a text past 32 KB over two rows doesn't close the session (#538) by @jasperblues in #546
  • Make NaN and the infinities floats in compiled Cypher on Postgres, Oracle and SQLite (#295, #409) by @jasperblues in #539
  • Keep a WITH that passes a path on apart on Oracle, so each row reads its own path's nodes and relationships (#537) by @jasperblues in #542
  • Refuse a float a function, arithmetic, a CASE, an UNWIND or a WITH gives as a text function's start or length, as Neo4j does (#285) by @jasperblues in #549
  • Document that Cypher of your own keeps no @caseInsensitive copy by @jasperblues in #568
  • Wrap the @caseInsensitive KDoc example's statement in a CypherStatement by @jasperblues in #569
  • Publish the openCypher TCK scores in the Cypher on a relational database docs by @jasperblues in #470
  • Join compiled Cypher's sources with JOIN … ON TRUE on SQLite, so it may reorder them and a step no longer scans every pair of nodes (#571) by @jasperblues in #572
  • Give NaN and the infinities from a division by a float zero, toFloat of Java's names, and sum() over both infinities on Postgres and SQLite (#408, #544, #545) by @jasperblues in #563
  • Refuse a MERGE that would make a property null, or NaN on Oracle, as Neo4j does (#543) by @jasperblues in #564
  • Read an undirected relationship both ways round and join on equalities on Postgres and Oracle, not an OR (#570) by @jasperblues in #574
  • Read an id compared by = or IN through the primary key on SQLite, Postgres and Oracle (#575) by @jasperblues in #577
  • Compare a node with a node so Oracle keeps both joins, and finds a node with more than one label (#576) by @jasperblues in #579
  • Read a node set equal to another node by its key on Oracle, testing its labels by their text (#578) by @jasperblues in #582
  • Renumber a year BC on Oracle by two functions the store keeps, so a statement of temporal arithmetic is planned in seconds again (drivine#572) by @jasperblues in #584
  • Read an empty map of properties in a pattern as none, and refuse pattern values Neo4j refuses (#583) by @jasperblues in #586
  • Leave a CI runner room in the hard-parse bounds, as Drivine for TypeScript's and for Python's do by @jasperblues in #589
  • Delete the rows of a compiled DELETE as a set, so MATCH (n) DETACH DELETE n is one statement, not one for each node (#587) by @jasperblues in #590
  • Say in AGENTS.md that releases go out as trains by @jasperblues in #454
  • Plan a node read by its key without statistics feedback on Oracle, so a new cursor's second and third runs keep the keyed plan (#585) by @jasperblues in #596
  • Test labels by their text in a set-based DELETE's subquery on Oracle, which crashed the session planning it from the index over labels by @jasperblues in #597
  • Keep a compiled write loop's rows in a list the store extends in place, so the loop takes time in step with its rows, not their square by @jasperblues in #600
  • Read a labelled node found by its id, or as a relationship's end, by its key on Oracle, not through the index over labels for each id or edge (#580, #581) by @jasperblues in #603
  • Test the labels of a node a MATCH finds after a grouping WITH by their text on Oracle, which crashed the session building its cursor (#601) by @jasperblues in #606
  • Score full-text search by BM25 on Postgres, SQLite and Oracle without Oracle Text, as Neo4j does (#604) by @jasperblues in #608
  • Plan a compiled write's per-row reads as nested loops on Postgres, so a MERGE over a table vacuumed while empty isn't quadratic (#609) by @jasperblues in #613
  • Check that the failing-loop settings test fails inside the loop, so a failure before it can't pass for the restore (#614) by @jasperblues in #615
  • Pin drivine4j-sample's Testcontainers to 1.21.4 over the Boot BOM's 1.21.3, whose Docker API a Docker Engine 29 refuses (#618) by @jasperblues in #620
  • Release v0.5.0 by @jasperblues in #630

Full Changelog: v0.4.1...v0.5.0

Part of Drivine 2026.11.2, with drivine 3.5.1, drivine-python 0.4.1 and Cypher spec v0.27.0.

v0.4.1

Choose a tag to compare

@github-actions github-actions released this 02 Nov 04:51
6ab4af2

What's Changed

  • Move the README's documentation into docs/, one file per section, and keep a short README by @jasperblues in #355
  • Write a list literal of more than 100 items, or a map of more than 50 entries, in parts on Postgres and SQLite by @jasperblues in #362
  • Reverse a text past 4,000 bytes on Oracle, where it was null by @jasperblues in #376
  • Work an aggregate's argument out first where it would bind a value, so temporal arithmetic in an aggregate compiles to kilobytes by @jasperblues in #359
  • Keep the temporal values a property of an unwound item may hold, so a list of them a WITH takes out is read by its places on Oracle by @jasperblues in #381
  • Order two lists item by item in <, <=, > and >=, and refuse toString of a list or a map, on the SQL stores by @jasperblues in #361
  • Give OracleTablesOnFirstConnectionTest's pools two sessions each, so CI's listener doesn't turn connections away by @jasperblues in #390
  • Bind a property's texts beside a long operand on Oracle, so a text past 32 KB a text function or a sum reads beside one leaves the session open by @jasperblues in #387
  • Hand out plain-SQL connections whatever the graph name, and commit the graph's tables on a pool that commits nothing (#364, #365) by @jasperblues in #396
  • Work a sort key beside a variable-length walk out first, so it compiles to kilobytes, not megabytes by @jasperblues in #392
  • Test IN against a long list literal of several kinds as one list, so -1 IN 1,001 items compiles to kilobytes on Oracle by @jasperblues in #407
  • Read a CALL's imported variables from the call's own rows where it aggregates, so Oracle's sum() and collect() see them by @jasperblues in #406
  • Work an aggregate out first where an operation over its result would bind a value, so texts and text functions over an aggregate compile to kilobytes by @jasperblues in #394
  • Make randomUUID() Oracle's UUID() and need Oracle 23.9, so a stamp's token doesn't come round again by @jasperblues in #402
  • Release: skip the CI gate when the same tree already passed CI (port of drivine#431) by @jasperblues in #431
  • Read a grouping key in an order term over an aggregate as the GROUP BY reads it, so Oracle sorts by me.age + count(you.age) again by @jasperblues in #429
  • Work out keys() on each row on Oracle, so every row has its own keys and not the first row's by @jasperblues in #430
  • Turn Oracle's JSON table transformation off in a query of two tables or more, so a node keeps its label test by @jasperblues in #432
  • Run the openCypher TCK on each store, held nightly to a per-scenario baseline by @jasperblues in #436
  • Find a text past 32 KB with CONTAINS on Oracle, where INSTR refused it by @jasperblues in #441
  • Read an item of a list through its text on Oracle, so one past 32 KB read by its place over two rows doesn't crash the session by @jasperblues in #443
  • Credit the TCK scenarios main now passes by @jasperblues in #445
  • Test five more shapes of an aggregate in a CALL by @jasperblues in #446
  • Add the Apache 2.0 license file by @jasperblues in #444
  • Release v0.4.1 by @jasperblues in #438
  • Serve a case-insensitive lookup from an index: @caseInsensitive keeps a lowercased copy by @jasperblues in #149
  • Work out a list that holds a subquery in a derived table on Oracle, so x IN it finds each row's items in a WHERE by @jasperblues in #447
  • Refuse a value of a known kind where an operator or a function takes another, as Neo4j does by @jasperblues in #448
  • Show the explainer video in the README, as the landing page does by @jasperblues in #442
  • Make a graph's tables again on the next connection when the first fails, and run the block again after a deadlock (#450) by @jasperblues in #453
  • Compare a text past 32 KB as a text on Oracle, so =, IN and < with one aren't refused and don't crash the session by @jasperblues in #458
  • Refuse a variable used as two kinds, or made where it is bound already, as Neo4j does by @jasperblues in #451
  • Give an integer sum past a long's range as a float on Oracle and Postgres, by its total (#339) by @jasperblues in #455
  • Test only the nodes a related field loads in a condition on it, so a view isn't found through a node of another label by @jasperblues in #457
  • Make a randomUUID() inside a larger expression anew for each row on the SQL stores (#412) by @jasperblues in #463
  • Give a CALL only the variables it imports, so a name it doesn't import is a new variable in it (#410) by @jasperblues in #464
  • Keep the compiler's hidden per-row variables in a CALL that imports only what it names by @jasperblues in #471
  • Truncate to a millennium, century, decade, week-based year or microsecond, and with a map of fields, on the SQL stores by @jasperblues in #468
  • Make the store identity's constraint again after a deadlock, and write a stamp the constraint refuses again by @jasperblues in #477
  • Test every node within a field's depth in a condition on it, as the load reads them by @jasperblues in #474
  • Join a DSL condition's related rows in the EXISTS on Postgres, so it can be a semi-join by @jasperblues in #480
  • Name the compiler's per-row UUID variables as no name of the statement's is, so a quoted #anewN stays the statement's (#479) by @jasperblues in #481
  • Pass on what a CALL imports with a leading WITH *, and nothing with a WITH * where nothing is bound (#472) by @jasperblues in #484
  • Gather a pattern comprehension whose body reads none of its pattern's variables from a derived table of its matches (#467) by @jasperblues in #488
  • Refuse an aggregate in a WHERE, and one beside a variable no item groups by in a body or an ORDER BY by @jasperblues in #490
  • Ensure the store key's constraint again when another manager made it meanwhile by @jasperblues in #493
  • Read a key that is a variable's property in an aggregating item, as Neo4j does by @jasperblues in #496
  • Make a date or a moment from a week, an ordinal day or a quarter, and read its day of the quarter, on the SQL stores by @jasperblues in #498
  • Filter a WITH's groups by an aggregate in its WHERE, as Neo4j does, where the WITH aggregates or is DISTINCT by @jasperblues in #501
  • Group, keep distinct and sort a map's null and a missing key as one null by @jasperblues in #503
  • Read a grouping key in a nested body, a body over no aggregate, any(y IN … WHERE y = e) and an order term, as Neo4j does by @jasperblues in #507
  • Make a date, a time or a moment from another temporal value, with the map's fields laid over its own, on the SQL stores by @jasperblues in #509
  • Run the Cypher spec at v0.18.0, and drop the local holds it now covers by @jasperblues in #511
  • Mark a CALL's UNION column as maybe JSON's null where any query's value is, not only the first's by @jasperblues in #514
  • Read any/none(y IN list WHERE y = e) as e IN list where e's only y is one a comprehension or reduce binds of its own by @jasperblues in #522
  • Run the nightly's FalkorDB as ghcr.io/falkordb/falkordb:latest, since Docker Hub's latest is a 4.x backport by @jasperblues in #523
  • Fix the Neo4j store-identity deadlock flake: look for the winner's key constraint for longer after a lost race by @jasperblues in #525
  • Run the Cypher spec at v0.21.0, and hol...
Read more

v0.4.0

Choose a tag to compare

@github-actions github-actions released this 28 Oct 07:50
a6aa1ed

What's Changed

  • Make all(), any(), none() and single() three-valued on the SQL stores, as Cypher's are by @jasperblues in #283
  • Read a start or length of substring(), left() and right() on the SQL stores as Neo4j reads it by @jasperblues in #280
  • Round to 0 places as round(x) does, half toward +infinity, where Neo4j does so by @jasperblues in #286
  • Run the Cypher spec at v0.6.9 by @jasperblues in #287
  • Show how to run Cypher on Postgres, Oracle and SQLite in the README by @jasperblues in #297
  • Compare and sort texts by code point in Cypher on Postgres, not by the database's collation by @jasperblues in #291
  • Run the Cypher spec at v0.6.10 by @jasperblues in #300
  • Compare a text past 32 KB in a list that holds a null on Oracle (#293) by @jasperblues in #302
  • Fix round() on a null, zero and tiny numbers, and pick its rounding when compiling by @jasperblues in #306
  • Walk a parameter's list with an item too long to write in the statement, so Oracle doesn't refuse its literal by @jasperblues in #310
  • Tell the sign of the least positive numbers on Oracle, whose SIGN gives 0 for them by @jasperblues in #315
  • Read a property a RETURN or a WITH gives through its text on Oracle, so a long one over two rows doesn't crash the session by @jasperblues in #316
  • Run the Cypher spec at v0.6.13: a property past 32 KB read on two nodes by @jasperblues in #317
  • Sort values of several kinds in Cypher's order on Oracle and SQLite by @jasperblues in #308
  • Make no adaptive plan in an Oracle session, so a join filtered by a label and a property keeps its rows by @jasperblues in #318
  • Give the test report step a 4 GB Node heap by @jasperblues in #321
  • Fail round() by a precision below 0 or past 2147483647 only on a row that rounds a number by it, as Neo4j does by @jasperblues in #324
  • Read an unwound item's text through text where a write works out what it writes on Oracle, so a long one over two rows doesn't crash the session by @jasperblues in #326
  • Measure, cut, change and search a text past 4,000 bytes on Oracle, where it was null by @jasperblues in #314
  • Say Postgres's exact decimal covers sum() and avg(), and that it is more accurate by @jasperblues in #329
  • Compare lists that hold a null item by item on Oracle in a statement that writes and in a reduce step by @jasperblues in #323
  • Keep a sum of floats a float on Oracle when its total is whole, as Neo4j does by @jasperblues in #334
  • Say SQLite adds floats as doubles, correcting for rounding as it goes by @jasperblues in #335
  • Open the concurrent function test's sessions once, so CI's Oracle listener doesn't refuse them by @jasperblues in #337
  • Sort a list of zoned values, and a list of lists of them, by their moments in Cypher on the SQL stores by @jasperblues in #319
  • Make the graph's tables on a SQL store's first connection, so Cypher through a persistence manager works on a new store by @jasperblues in #328
  • Log each statement in full in tests only when asked, so CI's JUnit report can read the Oracle job's results by @jasperblues in #342
  • Read a value a write's subquery takes from its row from what a UNION or an OPTIONAL CALL gives, not from the row by @jasperblues in #340
  • Join an OPTIONAL CALL or an OPTIONAL MATCH on 1 = 1 on Oracle, so one in a COUNT { } runs by @jasperblues in #343
  • Read an unwound item's text through its text on Oracle where a RETURN or a WITH gives it, so one past 32 KB over two rows doesn't crash the session by @jasperblues in #344
  • Run the Cypher spec at v0.6.16 by @jasperblues in #345
  • Release to Maven Central from a pushed vX.Y.Z tag, through CI by @jasperblues in #348
  • Read a list of temporal values item by item on Oracle, so a list worked out with temporal arithmetic parses at once, and sort zoned lists by their moments there too by @jasperblues in #346
  • Release a version bump when it merges, and add a Prepare release workflow by @jasperblues in #350
  • Run the Cypher spec at v0.6.17 by @jasperblues in #351
  • Run the Cypher spec at v0.6.18 by @jasperblues in #354
  • Don't keep a compiled Cypher statement in Oracle's statement cache, so a few dozen of them don't run the instance out of PGA (#313) by @jasperblues in #352
  • Read a list that holds temporal values at any depth item by item on Oracle, a coalesce of one and lists or maps of them too, so they parse at once by @jasperblues in #353
  • Prepare release: bump the sample's model file and the README's install snippets with build.gradle by @jasperblues in #356
  • Release v0.4.0 by @jasperblues in #357

Full Changelog: 0.3.0...v0.4.0

Part of Drivine 2026.10, with drivine 3.4.0, drivine-python 0.3.0 and Cypher spec v0.6.18.