Michael Jungmair comments

Results 26 comments of


                                            Michael Jungmair

add devcontainer specification

I usually thest them, by pushing to another branch. However, this is still an issue with your changes, as then wip would still be tagged lates :(

add devcontainer specification

You can rebase,now the dev image is tagged when built on main

Proper lowering of constant db::chars to i32 or util::VarLen32

What about the `ColumnStatisticsMap`, why did you introduce it?

Proper lowering of constant db::chars to i32 or util::VarLen32

There is a performance regression for several queries (e.g., Q12, Q17 for TPC-H). The reason is that now the db.char type is being used more frequently (as previously, char types...

Proper lowering of constant db::chars to i32 or util::VarLen32

Better, but still performance regressions for some TPC-DS queries ╒═════════╤═══════════════╕ │ query │ LingoDB(PR) │ ╞═════════╪═══════════════╡ │ 64.sql │ 3.83x │ ├─────────┼───────────────┤ │ 89.sql │ 3.53x │ ├─────────┼───────────────┤ │ 66.sql...

Proper lowering of constant db::chars to i32 or util::VarLen32

Apparently caused by casts of nullable chars to nullable strings, that are still not handled during the sample evaluation :) `%48 = db.cast %47 : !db.nullable> -> !db.nullable`

Proper lowering of constant db::chars to i32 or util::VarLen32

Performance is now ok :), when the failing CI is fixed, we can merge it :)

Handle `in` expressions with a large right side efficiently

Yes below some limit (20 is maybe a bit high), the current way should remain

try using mimalloc allocator

/benchmark --datasets tpch-10,tpcds-10

try using mimalloc allocator

/benchmark --datasets tpch-10,tpcds-10 --commits 1