Revert Fix half up in round #332

main changes: 1. remove website folder. 2. add compile script. 3. fix CmakeList. 4. clean up github workflow.

main changes: 1. Support multiple HDFS endpoints. 2. Add utility method MemoryUsageTracker::highUsage().

main changes: 1. add expand node. 2. add kCastIntAllowDecimal.

main changes: 1. implement vread for input stream. 2. add caseSensitive in ReaderOptions. 3. add ORC suuport. 4. make the DataBuffer grow ratio configurable.

main changes: 1. use companion function to support mixed aggregation steps. 2. add expand operator. 3. Set partial full when cardinality is high or memory exceeds. 4. Config option "join_spill_memory_threshold" (kJoinSpillMemoryThreshold) not working. 5. fix hashjoin runtime issue.

main changes: 1. Decimal relative code. 2. add Add mapping from named_struct to row_constructor.

main changes: 1. Fix half up in round 2. Clear null for partial avg results. 3. Support float & double types in pmod function. 4. Support date/decimal type hash function. 5. Optimize might contain 6. Let function validation fail for lookaround pattern in RE2-based implementation. 7. Register lpad/rpad functions for Spark SQL.

main changes: 1. Decimal relative.

main changes: 1. Support Decimal (type parse/function mapping/getPrecisionAndScale) 2. Support JoinRel and choose Merge join or Hash join by advanced_extension. 3. Support reading partition columns. 4. Support selecting a subfield from struct. 5. Support kstruct/kBinary/kDecimal literal. 6. Support switch/ kSingularOrList expression. 7. Support ExpandRel/WindowRel. 8. Filter pushdown. 9. SubstraitToVeloxPlanValidator.

main changes: 1. Decimal relative. 2. Use a config to control the tokenizing of dot. 3. Use companion function to support mixed aggregation steps

main changes: 1. Make varchar and varbinary compatible. 2. Add ValueStreamNode operator

* Add back not node. * Fix NOT handling in metadata filter. --------- Co-authored-by: Jimmy Lu <[email protected]>

Updated dependencies: libhdfs3: Try find_package first gflags: Allow static library xsimd, gtest: Allow finding from system Arrow, Parquet: Prefer external build if -D<package>_DIR=/path/to/arrow/install is defined flex 2.5.13 -> 2.6.0 (2.6.0 removed register keyword) Compat boost >= 1.79

velox cpp file use "velox/vector/arrow/Abi.h" velox header file use external arrow/Abi.h

add decimal column reader support. --------- Co-authored-by: zuochunwei <[email protected]> Co-authored-by: zhejiangxiaomai <[email protected]>

Add a drafted timestamp reader. Support timestamp in substrait, and add some validations for unsupported cases.

…ject#245) Align First/Last aggregate function's intermediate result with Spark. Support decimal.

Co-authored-by: yangyimin <[email protected]>

Support serialize and deserialize RowVector.

xpand timestamps in page reader so getValues of SelectiveIntegerColumnReader can be reused.

* fix segmentation fault in createStringFlatVector

Co-authored-by: zhli1142015 <[email protected]>

Further improved SelectiveLongDecimalColumnReader. Added decoder for int128_t. --------- Co-authored-by: yangyimin <[email protected]>

…ct#274)

…ct#268)

…ject#287) * Initial commit * Fix compile issue * Cherry pick PR 4510 * Fix issue in upstream PR * Fix bugs for unique sort key * Fix bugs for repeated sort key * Add more test cases * Fix int type issue * Handle null * Remove some commented code * Remove check null * Fix velox ut failure for rows frame * Format the code

…-project#291)

In Ubuntu, thrift will be installed manually in setup scripts, but Arrow still compile thrift, then Velox use system thrift and pre-build arrow/parquet. In Centos, Velox could not found system thrift, so it will compile Arrow and thrift twice. Since Arrow will compile thrift in all environments, lets use these pre-build shared libs to save time and keep toolchain consistency. Support native dependency could be overrided by env variables.