Skip to content

StreamFusion 0.1.0-rc4

Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 06 Oct 00:19
· 1652 commits to main since this release
ef3fc3f

What's Changed

  • Fix temporal admission and legacy interval decoding regressions by @jordepic in #80
  • Fix ASCII/CHR parity and extend native decimal JSON and member paths by @liuyongvs in #54
  • Fix decimal overflow and millisecond event-time boundaries by @liuyongvs in #75
  • Preserve calendar watermark semantics with shared native expressions by @liuyongvs in #56
  • Preserve eager IFNULL evaluation inside native islands by @liuyongvs in #111
  • Integrate reviewed native expressions, timestamp windows, and JSON paths by @liuyongvs in #128
  • Keep Flink string-to-boolean semantics inside native Calc by @liuyongvs in #112
  • Admit codepoint-aware INSTR start and occurrence searches by @liuyongvs in #113
  • Preserve exact decimal rounding and integral conversion in native islands by @liuyongvs in #114
  • Keep exact decimal and binary scalar UDF values inside native islands by @liuyongvs in #117
  • Preserve full timestamp values in grouped extrema and recovery by @liuyongvs in #118
  • Keep plain timestamp windows native through assignment and ranking by @liuyongvs in #121
  • Fix TIMESTAMP_LTZ window boundaries and firing in fixed-offset zones by @liuyongvs in #122
  • Support runtime ARRAY indexes and MAP keys with Flink lookup semantics by @liuyongvs in #127
  • Preserve declared types for folded NULL projections by @liuyongvs in #129
  • Fix upstream SQL exception, aggregate recovery, and temporal sort failures by @jordepic in #130
  • Accelerate SQL/JSON paths selecting empty object keys by @jordepic in #131
  • Preserve Flink retained row kinds for retracting OFFSET by @jordepic in #132
  • Preserve Flink nullness and external values in DECIMAL UDF consumers by @jordepic in #133
  • Skip Calc evaluation when filters leave an empty batch by @liuyongvs in #144
  • Fix scalar evaluation semantics and add exact decimal TRY_CAST/TRUNCATE by @liuyongvs in #151
  • Integrate SQL audit coverage and native retracting windows by @jordepic in #150
  • Support exact decimal floating casts and default-safe POWER by @jordepic in #134
  • Support native window COUNT(DISTINCT) with mergeable value sets by @jordepic in #135
  • Preserve JSON string identity through scalar consumers by @jordepic in #137
  • Support negative literal JSON array indexes by @jordepic in #138
  • Support ordered and single-value native grouped aggregates by @jordepic in #140
  • Verify upstream execution routes and retain stalled-test diagnostics by @jordepic in #143
  • Reject updating input in insert-only native window aggregation by @jordepic in #145
  • Evaluate constant FROM_UNIXTIME formats without JVM callbacks by @jordepic in #136
  • Make the public SQL audit corpus portable and coverage accountable by @jordepic in #139
  • Run upstream Delta SQL tests as a required integration suite by @jordepic in #152
  • Support verified escaped JSON member paths by @jordepic in #156
  • Support retracting grouping-only and integer AVG windows by @jordepic in #154
  • Support floating AVG over updating windows by @jordepic in #155
  • Keep OFFSET parity bundles independent of the wall clock by @jordepic in #153
  • Support FLOAT and DOUBLE SUM in retracting windows by @jordepic in #157
  • Support DECIMAL SUM in retracting windows by @jordepic in #158
  • Support partition-derived bounds on retracting Top-N by @jordepic in #159
  • Support escaped Unicode and control characters in JSON paths by @jordepic in #160
  • Support retracting DECIMAL AVG windows by @jordepic in #161
  • Accelerate verified whitespace after SQL/JSON array indexes by @jordepic in #163
  • Preserve shared binary UDF results across complete Calc rows by @jordepic in #162
  • Support conditional IF inside native expression pipelines by @liuyongvs in #169
  • Support Flink-compatible scalar hashes in native pipelines by @liuyongvs in #171
  • Support partition-derived bounds for update-fast Top-N by @jordepic in #170
  • Run SQL/JSON wildcard paths natively by @jordepic in #173
  • Support filtered numeric event-time window aggregates by @jordepic in #174
  • Accelerate SQL/JSON array-index unions by @jordepic in #175
  • Extend SQL/JSON coverage through the batch JVM bridge by @jordepic in #176
  • Support filtered numeric MIN/MAX in event-time windows by @jordepic in #177
  • Support filtered COUNT(DISTINCT) in event-time windows by @jordepic in #178
  • Recognize computed partition keys for variable Top-N bounds by @jordepic in #179
  • Accelerate append-only Top-N with changing bounds by @jordepic in #180
  • Accelerate retracting Top-N with changing bounds by @jordepic in #181
  • Preserve upstream Maven failures when XML reports pass by @jordepic in #190
  • Validate target Flink line before loading StreamFusion payloads by @jordepic in #191
  • Preserve Flink JSON_VALUE exception semantics in the batch JVM route by @jordepic in #192
  • Honor async lookup timeouts and bound outstanding requests by @jordepic in #193
  • Release Arrow vectors when row conversion fails by @jordepic in #195
  • Avoid repeated runtime suites in lake CI jobs by @jordepic in #194
  • Clean up failed native checkpoint preparation and uploads by @jordepic in #196
  • Audit legacy Delta SQL compatibility with released Flink 1.18 by @jordepic in #197
  • Support nested ARRAY and ROW values in the JSON JVM batch bridge by @jordepic in #199
  • Retain changing update-fast Top-N bounds with TTL disabled by @jordepic in #201
  • Preserve Flink's logging provider in deployment artifacts by @jordepic in #203
  • Publish exact upstream execution-audit denominators by @jordepic in #205
  • Make optimized JNI payloads loadable in official Flink images by @jordepic in #204
  • Audit native and fallback routes in unchanged upstream UDF tests by @jordepic in #206
  • Add an isolated Flink 1.18 build with host compatibility adapters by @jordepic in #198
  • Reject mixed Flink payloads before image daemons start by @jordepic in #200
  • Preserve live aggregate state across checkpoint snapshots by @liuyongvs in #222
  • Extend native Calc function coverage with released Flink semantics by @liuyongvs in #220
  • Support narrow integer aggregation and runtime charsets with Flink parity by @liuyongvs in #240
  • Inventory whole-query acceleration across upstream Flink SQL suites by @jordepic in #256
  • Bound INNER JOIN candidate memory and stream output chunks by @liuyongvs in #251
  • Preserve Flink exception order across Calc rows by @liuyongvs in #252
  • Prune unused fields before copying rows into Arrow batches by @liuyongvs in #253
  • Reduce per-row dispatch overhead for generated Flink expressions by @liuyongvs in #255
  • Profile JOIN state write granularity and retain the measured faster layout by @liuyongvs in #254
  • Enable Flink 1.18 RocksDB selection and share upstream test builds by @jordepic in #257
  • Close Flink 1.18 legacy UDF, lookup and INT96 coverage gaps by @jordepic in #258
  • Share SQL/JSON across Flink lines and preserve legacy ENCODE bytes by @jordepic in #259
  • Compare nested SQL result containers by content by @jordepic in #260
  • Keep DECIMAL FLOOR and CEIL inside verified scalar consumers by @jordepic in #262
  • Admit TIMESTAMP_LTZ branches through existing IF execution by @jordepic in #261
  • Preserve COALESCE failure parity and resolve deferred-filter investigation by @jordepic in #269
  • Reduce upstream integration CI build and scheduling overhead by @jordepic in #278
  • Specialize integer, decimal and string DISTINCT state by @jordepic in #267
  • Add combined SQL recovery, rescale and memory-budget matrix by @jordepic in #276
  • Collect SQL invocation work and job evidence for execution audits by @jordepic in #277
  • Support DATE, TIME and BOOLEAN grouped value aggregates by @jordepic in #266
  • Fix recovery failure masking and collect-sink teardown flakes by @jordepic in #279
  • Admit typed two-phase DISTINCT views within exact merge bounds by @jordepic in #270
  • Make empty trim parity independent of prior string decoding by @jordepic in #280
  • Support exact single-phase AVG DISTINCT with persistent membership by @jordepic in #271
  • Support runtime decimal ROUND and TRUNCATE scalar consumers by @jordepic in #265
  • Contain Paimon failure-injection cleanup cascades by @jordepic in #283
  • Support Paimon 1.0.0 on Flink 1.18 by @jordepic in #282
  • Handle temporary sink-file renames in rescaling tests by @jordepic in #284
  • Reduce CI rebuilds and let Flink versions test independently by @jordepic in #285
  • Support two-phase distinct averages with exact membership merging by @jordepic in #272
  • Support Boolean and temporal TRY_CAST in native Calc by @jordepic in #263
  • Support distinct first and last string values with ordered membership by @jordepic in #273
  • Accelerate integer ARRAY_DISTINCT in native Calc by @jordepic in #286
  • Admit verified at-least-once Paimon consumer retention by @jordepic in #287
  • Preserve SINGLE_VALUE cardinality in two-phase string aggregation by @jordepic in #274
  • Document Kafka release pairings and actual validation coverage by @jordepic in #289
  • Reduce CI cache pressure and parallelize Paimon test classes by @jordepic in #288
  • Admit wide-decimal DISTINCT merges and reduce aggregation overhead by @jordepic in #275
  • Reuse native release builds for Paimon 1.0 CI by @jordepic in #290
  • Document and reproduce released Paimon split-commit recovery loss by @jordepic in #291
  • Fix processing-time and Paimon semantics while reducing disk join rewrites by @jordepic in #292
  • Establish Criterion baselines across native operations by @jordepic in #293
  • Deliver validated native benchmarks and copy reductions by @jordepic in #294
  • Add opt-in Linux io_uring reads for native RocksDB state by @jordepic in #295
  • Prepare v0.1.0-rc4 and preserve Paimon updates across recovery by @jordepic in #296
  • Generate publication metadata before validating universal release bundles by @jordepic in #297

Full Changelog: v0.1.0-rc3...v0.1.0-rc4