V0.0.3

We are happy to release Kùzu 0.0.3. This release comes with the following new main features and improvements:

• Kùzu as a Pytorch Geometric (PyG) Remote Backend: You can now train PyG GNNs and other models directly using graphs (and node features) stored on Kùzu. See this Colab notebook for a demonstrative example.
• Data ingestion from multiple files and numpy files.
• Query optimizer improvements.
• New buffer manager.
• INT32, INT16, FLOAT, and FIXED LIST data types (the latter is particularly suitable to store node features in graph ML applications)
• Query timeout mechanism and interrupting queries from CLI.

See our blog post for more details. Enjoy your reading and don't forget to pip install kuzu!

What's Changed

• Update readme by @andyfengHKU in #1285
• Add internal offset function by @andyfengHKU in #1291
• Decouple group pos between operator and planner by @andyfengHKU in #1290
• Rename config to constants by @ray6080 in #1293
• Unify logger create, get, and drop by @ray6080 in #1294
• Add collect agg function by @acquamarin in #1292
• Add CI and slack badge to README by @ray6080 in #1300
• Fix getRelPropertyNames result by @mewim in #1301
• Fix build issue on Arch Linux by @mewim in #1304
• Add Python API binding for prepared statement by @mewim in #1305
• Factorization rewriter by @andyfengHKU in #1307
• Add Fixed-list DataType to System by @acquamarin in #1298
• Arithmetic function framework refactor by @acquamarin in #1312
• Fix #998 by @ray6080 in #1315
• Fix tensor bug by @acquamarin in #1318
• Hash join build projection push down by @andyfengHKU in #1319
• Replace getRawName with toString by @andyfengHKU in #1324
• Simplify mask struct by @ray6080 in #1320
• Implement float(alias: REAL) dataType by @acquamarin in #1321
• Add twitter badge to readme by @ray6080 in #1326
• Sink projection push down by @andyfengHKU in #1325
• Add INT16,INT32 datatype by @acquamarin in #1328
• Fix logo by @mewim in #1330
• Add remove unnecessary join optimizer by @andyfengHKU in #1329
• Fix issue 1323 by @acquamarin in #1331
• Add CLA by @ray6080 in #1334
• Add logical operator visitor by @andyfengHKU in #1333
• Solve format string vulnerability by @acquamarin in #1335
• Remove csv reader by @mewim in #1341
• Fix min/max agg function on string column BUG by @acquamarin in #1342
• Remove jo connection by @andyfengHKU in #1348
• Fix benchmark query by @acquamarin in #1349
• Fix plan deep copy by @andyfengHKU in #1345
• Patch c str api by @andyfengHKU in #1353
• Update slack link by @ray6080 in #1354
• Refactor FT interface by @acquamarin in #1351
• Change databasePath to positional argument by @acquamarin in #1350
• Fix drop node property bug by @acquamarin in #1356
• Fix empty list bug by @acquamarin in #1358
• Flat schema by @andyfengHKU in #1357
• Bump benchmark threads to 10 by @mewim in #1337
• Asp optimizer rework by @andyfengHKU in #1361
• Add support to copy multiple csv files by @acquamarin in #1363
• Multi key asp by @andyfengHKU in #1364
• Separate FileHandle and BufferManagedFileHandle by @ray6080 in #1365
• Add asp for wcoj by @andyfengHKU in #1366
• Multi label asp by @andyfengHKU in #1370
• Sequential read for StringPropertyList by @rfdavid in #1327
• Add glob support to copycsv by @acquamarin in #1376
• Add test coverage report to CI by @acquamarin in #1378
• Add codecov badge by @mewim in #1380
• Fix codecov link & reorder badge by @mewim in #1381
• Add support to copy from multiple parquet files by @acquamarin in #1377
• Multi scans asp by @andyfengHKU in #1375
• Remove bm resize interface by @ray6080 in #1382
• Rework BM to unify the physical memory usage of BM and MM by @ray6080 in #1368
• Factorization aware dp by @andyfengHKU in #1384
• Fix issue 1385 by @andyfengHKU in #1387
• Implement query interruption by @acquamarin in #1389
• Change cypher comment rule by @acquamarin in #1403
• Add query timeout mechanism by @acquamarin in #1395
• Fix ctrl-D by @acquamarin in #1402
• Filter push down by @andyfengHKU in #1397
• Separate WAL page idx lock from page lock in BMFileHandle by @ray6080 in #1398
• Encode plan when inserting to dp table by @andyfengHKU in #1406
• Add min/max agg function support for more types by @acquamarin in #1410
• Refactor index nested loop join planning by @andyfengHKU in #1411
• Copy node tables from npy files by @mewim in #1396
• BM optimistic reads by @ray6080 in #1409
• Fix hash index slot alignment by @ray6080 in #1414
• Cardinality est by @andyfengHKU in #1415
• Add python APIs by @acquamarin in #1423
• Rework asp optimizer by @andyfengHKU in #1417
• Optimize copy node memory usage by @ray6080 in #1425
• Populate by column by @acquamarin in #1424
• Fix copy invalid filepath by @acquamarin in #1426
• Add limit to TENSOR size by @acquamarin in #1427
• Numerical downcast by @andyfengHKU in #1429
• Fix issue 1428, 1408 by @andyfengHKU in #1430
• Remote backend for PyG by @mewim in #1418
• Add storage version info by @acquamarin in #1436
• change copy from npy keyword by @andyfengHKU in #1438
• Add python bindings for new types by @mewim in #1440
• Fix bm file handle remove pages in eviction queue by @ray6080 in #1431
• Change npy copy grammar by @andyfengHKU in #1441
• Fix copy options case by @acquamarin in #1447
• Fix pyg getAllEdgesForTorchGeometric by @ray6080 in #1442
• Fix hash vector creation in hj by @andyfengHKU in #1449
• Get version from CMakeLists for Python build by @mewim in #1450
• Remove spdlog/fmt from header by @mewim in #1451
• On demand import for PyG by @mewim in #1452
• Remove unused np by @mewim in #1453

New Contributors

• @rfdavid made their first contribution in #1327

Full Changelog: v0.0.2...v0.0.3