Very poor performance

[bbkr]

I'd love to use Raku in more demanding projects but currently performance of DBIish is far from being acceptable. I've created simple benchmark - execute 100K prepared queries that accepts one argument and returns single row as Hash.

Results on my machine:

• Perl - https://gist.github.com/bbkr/d9d175c981c0a233af5bd3a4c4a3ebe0 - 10.7s
• Python - https://gist.github.com/bbkr/bb6bd31dab1654dc6fb35023788c6196 - 15.3s *
• DotNet - https://gist.github.com/bbkr/3a8f3b1c72eaf4d29bd8f6b21da192eb - 11.0s
• Raku - https://gist.github.com/bbkr/44c61fdfd3bdc410146e3dea8ab7493c - 51.7s

This is general issue about lack of optimizations in DBIish. I do not have any ready to implement advice here. But being 3-5 times slower than competition in databases area is serious blocker in wider Raku adoption.

* Python was not using server-side prepared statements

[rbt]

There have been some improvements on this over the past year, mostly (I think) from the new dispatch mechanism being used by NativeCalls.

With PostgreSQL using your test code (driver/username/password changed) I get:

• Perl: 1.9s
• Raku: 7.3s

Nearly 1.4s of that is creating :hash'd results. Using $statement.row() for array results it's 5.9s.

Some of the remaining gap will be closed with expected performance improvements to LEAVE {} handling in a near-future version of Raku.