Further optimize filter expressions

[nikolai-mb]

This PR is a continuation of the improvements from PR #7005

By introducing a struct and extracting the parameter manually, it's possible to avoid a closure and lambda expression allocation for each filter operator value.

Note:
The added struct has a single property "p". The name is important as this is the same name that EF Core uses for the SQL Server provider by default when parameterizing values from anonymous closures. Changing the name causes SQL Server snapshot tests to fail, since "@__p_0" in the produced SQL would turn into "@__ChangedName_p"

Commit e9b8238 is not related to the perf improvements below, but should speed up the static class initialization a tiny bit

Benchmarks (commit 9b5b9e1)

The HC14 benchmarks represent the changes made in #7005 and what's currently in the main branch

Method	Mean	Error	StdDev	Ratio	Gen0	Allocated	Alloc Ratio
ContainsHC13	642.49 ns	4.118 ns	3.852 ns	1.00	0.0134	744 B	1.00
ContainsHC14	354.24 ns	2.918 ns	2.729 ns	0.55	0.0095	512 B	0.69
ContainsPR	264.37 ns	1.564 ns	1.386 ns	0.41	0.0072	376 B	0.51
EqualsHC13	382.57 ns	1.079 ns	1.010 ns	1.00	0.0095	488 B	1.00
EqualsHC14	159.52 ns	0.414 ns	0.323 ns	0.42	0.0050	256 B	0.52
EqualsPR	76.30 ns	0.359 ns	0.318 ns	0.20	0.0024	120 B	0.25
GreaterThanHC13	378.11 ns	3.283 ns	3.071 ns	1.00	0.0095	488 B	1.00
GreaterThanHC14	153.46 ns	0.617 ns	0.577 ns	0.41	0.0050	256 B	0.52
GreaterThanPR	75.43 ns	0.243 ns	0.227 ns	0.20	0.0024	120 B	0.25
InHC13	3,836.08 ns	33.165 ns	31.023 ns	1.00	0.1068	5840 B	1.00
InHC14	224.84 ns	0.502 ns	0.445 ns	0.06	0.0057	288 B	0.05
InPR	134.35 ns	0.667 ns	0.623 ns	0.04	0.0029	152 B	0.03
LowerThanHC13	389.97 ns	1.322 ns	1.236 ns	1.00	0.0095	488 B	1.00
LowerThanHC14	154.34 ns	0.502 ns	0.469 ns	0.40	0.0050	256 B	0.52
LowerThanPR	75.47 ns	0.247 ns	0.231 ns	0.19	0.0024	120 B	0.25
NotContainsHC13	648.94 ns	3.856 ns	3.607 ns	1.00	0.0153	792 B	1.00
NotContainsHC14	378.81 ns	1.312 ns	1.228 ns	0.58	0.0110	560 B	0.71
NotContainsPR	281.94 ns	1.398 ns	1.168 ns	0.43	0.0081	424 B	0.54
NotEqualsHC13	381.56 ns	0.809 ns	0.757 ns	1.00	0.0095	488 B	1.00
NotEqualsHC14	152.36 ns	0.502 ns	0.392 ns	0.40	0.0050	256 B	0.52
NotEqualsPR	74.08 ns	0.179 ns	0.158 ns	0.19	0.0024	120 B	0.25
StartsWithHC13	646.54 ns	4.196 ns	3.720 ns	1.00	0.0134	744 B	1.00
StartsWithHC14	348.92 ns	1.124 ns	0.996 ns	0.54	0.0100	512 B	0.69
StartsWithPR	257.99 ns	1.333 ns	1.247 ns	0.40	0.0072	376 B	0.51

[nikolai-mb]

I see there's some raven snapshot tests failing, although not entirely sure why since the EF sql conversions work flawlessly. Will have a look. I'm still having issues with tons of snapshot tests failing due to snapshot containing \n while my machine produces \r\n - is there a known issue / fix for this by any chance?

[nikolai-mb]

On second glance i see that the tests are failing for other PR's as well and not related to this change specifically

[PascalSenn]

@nikolai-mb there are a few flaky tests
@glen-84 do you see anything out of the ordinary here?

[glen-84]

There are currently 4 projects with failing tests:

• Data.Raven.Filters.Tests and Data.Raven.Sorting.Tests
  • Waiting for More ordering bugs ravendb/ravendb#18912 to be fixed.
  • If it's considered urgent, we can switch the search engine to Lucene until the Corax bugs are fixed.
• HotChocolate.AspNetCore.Tests
  • Waiting for an answer from @michaelstaib.
• HotChocolate.Subscriptions.Postgres.Tests
  • Waiting for Pascal to review Fixed ContinuousTask test #7316.

I'm still having issues with tons of snapshot tests failing due to snapshot containing \n while my machine produces \r\n - is there a known issue / fix for this by any chance?

Yes. I have this on my after-hours to-do list. Since the snapshot data can come from various places, it's not always possible to configure the EOL sequence. For this reason, I'm considering proposing an update to Snapshooter that forces a specific sequence to be used while writing the snapshot.

@glen-84 do you see anything out of the ordinary here?

I don't see any new test failures, if that's what you're asking. 🙂

[michaelstaib]

Thank you for your contribution.