Minified version with no table engines

[danthegoodman1]

A lot of binary size comes from the tables engines that may not be relevant for in-process use cases like the merge tree engines, log engines, etc.

Would be great to either have an easy way to compile with engines omitted, or a build that is effectively engine-free (except for some basics like url, s3, file) for a far smaller build. The expectation is that custom engines would be made on top of the url/s3 engines in #52 as sorts of aliases

[auxten]

I did some analysis with bloaty, the AggravateFunctions take up a lot of space.
We can work with clickhouse team to add more compile flags to enable stripping some of them.

compileunits	vmsize	filesize
[section .debug_loc]	0	208584312
[section .text]	151282685	151282685
[section .strtab]	0	59946963
[section .debug_ranges]	0	56290240
[section .rodata]	48926306	48926306
./src/AggregateFunctions/AggregateFunctionSumMap.cpp	7787534	48880276
./src/AggregateFunctions/AggregateFunctionMin.cpp	6657137	45979726
./src/AggregateFunctions/AggregateFunctionMax.cpp	6640377	45884458
[section .dynstr]	38266753	38266753
./src/Interpreters/HashJoin.cpp	3760601	35714954
./src/Interpreters/Aggregator.cpp	3730680	34616722
./src/AggregateFunctions/AggregateFunctionUniqCombined.cpp	5393994	32101843
./src/AggregateFunctions/AggregateFunctionAvgWeighted.cpp	5165493	31640382
./src/AggregateFunctions/AggregateFunctionStatisticsSimple.cpp	6239088	30171734
./src/AggregateFunctions/AggregateFunctionUniq.cpp	3541725	28680035
./src/AggregateFunctions/AggregateFunctionStatistics.cpp	5169543	26427110
./src/Interpreters/ActionsDAG.cpp	219076	23502928
./src/Storages/MergeTree/KeyCondition.cpp	126008	19541606
./src/AggregateFunctions/AggregateFunctionDeltaSumTimestamp.cpp	2681548	19281574
./src/Planner/Planner.cpp	59051	16236022
./src/Planner/PlannerJoins.cpp	40060	16084904
./src/Planner/PlannerJoinTree.cpp	51178	16039921
./contrib/hive-metastore/ThriftHiveMetastore.cpp	2556915	15995512
[section .eh_frame]	15656064	15656064
./src/AggregateFunctions/AggregateFunctionSparkbar.cpp	1801900	15486531
./src/Interpreters/castColumn.cpp	27915	15357640
./src/Storages/StorageReplicatedMergeTree.cpp	1145437	13526552
./src/Columns/ColumnVector.cpp	1582603	13425874
./src/Core/Settings.cpp	1331377	11051219
./src/AggregateFunctions/AggregateFunctionSimpleLinearRegression.cpp	1463071	10388820
./src/Formats/ProtobufSerializer.cpp	571179	9525325
[section .gcc_except_table]	9466872	9466872
./src/Storages/MergeTree/MergeTreeData.cpp	670858	9271290
[section .symtab]	0	8849736
./src/AggregateFunctions/AggregateFunctionQuantile.cpp	1094793	8697684
./contrib/hive-metastore/hive_metastore_types.cpp	1262919	8545635
./src/Core/SettingsEnums.cpp	307120	8250957
./src/AggregateFunctions/AggregateFunctionQuantileExactWeighted.cpp	939254	8015627
./contrib/NuRaft/src/asio_service.cxx	680502	7699920
./src/AggregateFunctions/AggregateFunctionAny.cpp	1000265	7280299
./src/Columns/ColumnDecimal.cpp	806743	7277378
./src/AggregateFunctions/AggregateFunctionQuantileDeterministic.cpp	897465	7270971
./src/AggregateFunctions/AggregateFunctionGroupArrayMoving.cpp	1169946	7142681
./src/Interpreters/InterpreterSelectQuery.cpp	322195	7088294
./src/Interpreters/Context.cpp	400123	6891480
./contrib/libunwind/src/libunwind.cpp	42849	6649674
./src/AggregateFunctions/AggregateFunctionSum.cpp	935631	6490650
./src/AggregateFunctions/AggregateFunctionSequenceMatch.cpp	464328	6480659
./src/AggregateFunctions/AggregateFunctionTopK.cpp	583102	6290272

Full table: clickhouse.sizeinfo.csv

[lmangani]

@alexey-milovidov any input or comments on this intent and related opportunities from the ClickHouse team?

[alexey-milovidov]

We can easily disable particular storages during the build - first try to do it manually and check the difference in the binary size. Then we can introduce some flags for it.

Disabling functions is as easy as removing a .cpp file.

Also, we can get benefits from the removal of the dynamic symbol table: ClickHouse/ClickHouse#47475

[zhanglistar]

Simply strip can reduce about 100MB file size.

[lmangani]

@zhanglistar true, but we already strip libchdb.so which brings it to about ~380MB uncompressed and ~100MB compressed.

[zhanglistar]

Also, disabling some third party libraries can reduce size, like hive or mysql etc.

[auxten]

Also, disabling some third party libraries can reduce size, like hive or mysql etc.

Thanks, I have disabled unnecessary libs for chdb. Hive and MySQL might be useful for chDB users.

[auxten]

@nmreadelf is working on that