MQE Absent function #10523

lamida · 2025-01-27T14:35:50Z

What this PR does

Implement absent function. Added a new Operator: Absent. We also need to modify InstantVectorFunctionOperatorFactory to pass parser.Expressions object needed to evaluate the labels of the absent argument.

Benchmark

Step to create the benchmark:

Add test in pkg/streamingpromql/benchmarks/benchmarks.go
cd tools/benchmark-query-engine
go run . -bench=Query/absent -count=6 | tee absent-benchmark.txt
./compare.sh absent-benchmark.txt
rm absent-benchmark.txt

goos: darwin
goarch: arm64
pkg: github.com/grafana/mimir/pkg/streamingpromql/benchmarks
cpu: Apple M1 Max
                                                           │  Prometheus   │                Mimir                │
                                                           │    sec/op     │    sec/op     vs base               │
Query/absent(a_1),_instant_query-10                           144.0µ ±  4%   144.2µ ±  8%        ~ (p=0.699 n=6)
Query/absent(a_1),_range_query_with_100_steps-10              154.3µ ±  5%   147.4µ ±  2%   -4.44% (p=0.004 n=6)
Query/absent(a_1),_range_query_with_1000_steps-10             250.2µ ± 12%   199.0µ ±  3%  -20.47% (p=0.002 n=6)
Query/absent(a_100),_instant_query-10                         752.1µ ±  5%   731.9µ ±  3%   -2.69% (p=0.015 n=6)
Query/absent(a_100),_range_query_with_100_steps-10            1.347m ±  3%   1.250m ±  1%   -7.21% (p=0.002 n=6)
Query/absent(a_100),_range_query_with_1000_steps-10           6.264m ±  1%   5.454m ±  1%  -12.93% (p=0.002 n=6)
Query/absent(a_2000),_instant_query-10                       10.416m ±  2%   9.895m ±  9%        ~ (p=0.065 n=6)
Query/absent(a_2000),_range_query_with_100_steps-10           21.42m ± 13%   19.24m ± 11%  -10.16% (p=0.015 n=6)
Query/absent(a_2000),_range_query_with_1000_steps-10         146.07m ±  1%   97.98m ±  3%  -32.92% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_instant_query-10                     127.0µ ± 15%   117.8µ ±  9%   -7.25% (p=0.026 n=6)
Query/absent(a_1_>_Inf),_range_query_with_100_steps-10        176.2µ ±  3%   128.1µ ±  6%  -27.30% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_range_query_with_1000_steps-10       578.7µ ±  4%   190.6µ ±  4%  -67.07% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_instant_query-10                   768.5µ ±  7%   734.2µ ±  6%   -4.46% (p=0.041 n=6)
Query/absent(a_100_>_Inf),_range_query_with_100_steps-10      1.500m ±  3%   1.305m ±  2%  -13.03% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_range_query_with_1000_steps-10     7.671m ±  4%   5.881m ±  1%  -23.34% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_instant_query-10                  10.49m ± 12%   10.05m ±  2%   -4.19% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_100_steps-10     23.52m ±  9%   20.16m ±  2%  -14.28% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_1000_steps-10    171.2m ±  5%   106.4m ±  5%  -37.86% (p=0.002 n=6)
geomean                                                       2.370m         1.929m        -18.58%

                                                           │  Prometheus   │                Mimir                │
                                                           │     B/op      │     B/op      vs base               │
Query/absent(a_1),_instant_query-10                           22.23Ki ± 0%   20.85Ki ± 0%   -6.22% (p=0.002 n=6)
Query/absent(a_1),_range_query_with_100_steps-10              25.65Ki ± 0%   21.65Ki ± 0%  -15.58% (p=0.002 n=6)
Query/absent(a_1),_range_query_with_1000_steps-10             51.70Ki ± 0%   26.31Ki ± 0%  -49.12% (p=0.002 n=6)
Query/absent(a_100),_instant_query-10                         194.9Ki ± 0%   150.5Ki ± 0%  -22.78% (p=0.002 n=6)
Query/absent(a_100),_range_query_with_100_steps-10            253.7Ki ± 0%   206.9Ki ± 0%  -18.44% (p=0.002 n=6)
Query/absent(a_100),_range_query_with_1000_steps-10           697.5Ki ± 0%   632.3Ki ± 0%   -9.35% (p=0.002 n=6)
Query/absent(a_2000),_instant_query-10                        3.536Mi ± 0%   2.577Mi ± 0%  -27.11% (p=0.002 n=6)
Query/absent(a_2000),_range_query_with_100_steps-10           4.660Mi ± 0%   3.679Mi ± 0%  -21.04% (p=0.002 n=6)
Query/absent(a_2000),_range_query_with_1000_steps-10          12.80Mi ± 1%   11.85Mi ± 1%   -7.49% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_instant_query-10                     25.20Ki ± 0%   21.42Ki ± 0%  -15.02% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_range_query_with_100_steps-10        56.87Ki ± 0%   22.16Ki ± 0%  -61.03% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_range_query_with_1000_steps-10      338.08Ki ± 0%   26.59Ki ± 0%  -92.14% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_instant_query-10                   198.0Ki ± 0%   153.3Ki ± 0%  -22.54% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_range_query_with_100_steps-10      285.2Ki ± 0%   209.7Ki ± 0%  -26.49% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_range_query_with_1000_steps-10     984.6Ki ± 0%   631.3Ki ± 0%  -35.88% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_instant_query-10                  3.542Mi ± 0%   2.655Mi ± 0%  -25.03% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_100_steps-10     4.694Mi ± 0%   3.758Mi ± 0%  -19.95% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_1000_steps-10    13.05Mi ± 1%   11.89Mi ± 1%   -8.90% (p=0.002 n=6)
geomean                                                       473.8Ki        314.8Ki       -33.56%

                                                           │  Prometheus  │               Mimir                │
                                                           │  allocs/op   │  allocs/op   vs base               │
Query/absent(a_1),_instant_query-10                            401.0 ± 0%    347.0 ± 0%  -13.47% (p=0.002 n=6)
Query/absent(a_1),_range_query_with_100_steps-10               511.0 ± 0%    354.0 ± 0%  -30.72% (p=0.002 n=6)
Query/absent(a_1),_range_query_with_1000_steps-10             1437.0 ± 0%    380.0 ± 0%  -73.56% (p=0.002 n=6)
Query/absent(a_100),_instant_query-10                         2.413k ± 0%   2.252k ± 0%   -6.67% (p=0.002 n=6)
Query/absent(a_100),_range_query_with_100_steps-10            2.830k ± 0%   2.661k ± 0%   -5.97% (p=0.002 n=6)
Query/absent(a_100),_range_query_with_1000_steps-10           6.344k ± 0%   5.275k ± 0%  -16.85% (p=0.002 n=6)
Query/absent(a_2000),_instant_query-10                        40.87k ± 0%   38.80k ± 0%   -5.08% (p=0.002 n=6)
Query/absent(a_2000),_range_query_with_100_steps-10           47.06k ± 0%   46.87k ± 0%   -0.42% (p=0.002 n=6)
Query/absent(a_2000),_range_query_with_1000_steps-10         100.04k ± 0%   98.94k ± 0%   -1.10% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_instant_query-10                      462.0 ± 0%    361.0 ± 0%  -21.86% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_range_query_with_100_steps-10         970.5 ± 0%    368.0 ± 0%  -62.08% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_range_query_with_1000_steps-10       5499.0 ± 0%    394.0 ± 0%  -92.84% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_instant_query-10                   2.474k ± 0%   2.269k ± 0%   -8.29% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_range_query_with_100_steps-10      3.289k ± 0%   2.679k ± 0%  -18.55% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_range_query_with_1000_steps-10    10.406k ± 0%   5.293k ± 0%  -49.14% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_instant_query-10                  40.93k ± 0%   38.82k ± 0%   -5.15% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_100_steps-10     47.50k ± 0%   46.88k ± 0%   -1.31% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_1000_steps-10   104.09k ± 0%   98.99k ± 0%   -4.90% (p=0.002 n=6)
geomean                                                       6.007k        4.035k       -32.82%

                                                           │  Prometheus   │                Mimir                │
                                                           │       B       │      B        vs base               │
Query/absent(a_1),_instant_query-10                           67.13Mi ± 1%   66.93Mi ± 1%        ~ (p=0.240 n=6)
Query/absent(a_1),_range_query_with_100_steps-10              67.01Mi ± 2%   66.65Mi ± 1%        ~ (p=0.686 n=6)
Query/absent(a_1),_range_query_with_1000_steps-10             64.52Mi ± 2%   65.21Mi ± 1%   +1.07% (p=0.041 n=6)
Query/absent(a_100),_instant_query-10                         62.32Mi ± 1%   61.69Mi ± 1%        ~ (p=0.093 n=6)
Query/absent(a_100),_range_query_with_100_steps-10            62.62Mi ± 1%   62.02Mi ± 2%        ~ (p=0.132 n=6)
Query/absent(a_100),_range_query_with_1000_steps-10           64.91Mi ± 1%   61.78Mi ± 1%   -4.83% (p=0.002 n=6)
Query/absent(a_2000),_instant_query-10                        63.10Mi ± 1%   63.23Mi ± 1%        ~ (p=0.786 n=6)
Query/absent(a_2000),_range_query_with_100_steps-10           71.98Mi ± 2%   62.62Mi ± 1%  -13.00% (p=0.002 n=6)
Query/absent(a_2000),_range_query_with_1000_steps-10         127.16Mi ± 3%   68.32Mi ± 2%  -46.27% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_instant_query-10                     67.08Mi ± 2%   67.46Mi ± 1%        ~ (p=0.193 n=6)
Query/absent(a_1_>_Inf),_range_query_with_100_steps-10        65.09Mi ± 1%   67.28Mi ± 2%   +3.36% (p=0.002 n=6)
Query/absent(a_1_>_Inf),_range_query_with_1000_steps-10       63.39Mi ± 1%   65.66Mi ± 1%   +3.57% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_instant_query-10                   62.13Mi ± 1%   62.17Mi ± 1%        ~ (p=1.000 n=6)
Query/absent(a_100_>_Inf),_range_query_with_100_steps-10      62.62Mi ± 1%   61.56Mi ± 1%   -1.68% (p=0.002 n=6)
Query/absent(a_100_>_Inf),_range_query_with_1000_steps-10     65.15Mi ± 0%   61.96Mi ± 0%   -4.89% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_instant_query-10                  63.38Mi ± 3%   63.34Mi ± 2%        ~ (p=0.732 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_100_steps-10     71.33Mi ± 1%   62.75Mi ± 1%  -12.03% (p=0.002 n=6)
Query/absent(a_2000_>_Inf),_range_query_with_1000_steps-10   129.14Mi ± 3%   67.95Mi ± 2%  -47.39% (p=0.002 n=6)
geomean                                                       70.26Mi        64.32Mi        -8.45%

Which issue(s) this PR fixes or relates to

Fixes #

Checklist

Tests updated.
Documentation added.
CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX].
about-versioning.md updated with experimental features.

pkg/streamingpromql/operators/absent.go

pkg/streamingpromql/functions.go

lamida · 2025-02-13T20:30:18Z

@charleskorn I have addressed all of your comments from the previous review. Please help to do another review. 🙏

pkg/streamingpromql/testdata/ours/functions.test

pkg/streamingpromql/operators/functions/absent.go

pkg/streamingpromql/functions_test.go

pkg/streamingpromql/testdata/ours/functions.test

jhesketh

Sorry if this has already been discussed/considered (if so, please disregard this comment).

However, I wonder if given the uniqueness of this operator if it makes sense to handle it as a special case instead of as an InstantVectorFunctionOperatorFactory?

Eg in convertFunctionCallToInstantVectorOperator in query.go we check if it's the absent function and then create the operator and pass it the full inner expression instead of looking it up in instantVectorFunctionOperatorFactories.

This way the absent operator can reason about the inner series. For example, if I'm understand the function correctly, there is nothing to do if the inner expression is something other than a selector (eg there's no point evaluating a sum).

Sorry if this has already been discussed or derails things too much!

charleskorn · 2025-02-17T00:08:09Z

This way the absent operator can reason about the inner series. For example, if I'm understand the function correctly, there is nothing to do if the inner expression is something other than a selector (eg there's no point evaluating a sum).

absent can be run over any expression that produces an instant vector. The only difference in behaviour between the case where absent is run over a instant vector selector and some other expression is the labels it will return:

if the argument is an instant vector selector, absent will infer the output labels from the selector
if the argument isn't an instant vector selector, absent will always produce a series with no labels (ie. {})

jhesketh · 2025-02-17T03:45:38Z

This way the absent operator can reason about the inner series. For example, if I'm understand the function correctly, there is nothing to do if the inner expression is something other than a selector (eg there's no point evaluating a sum).

absent can be run over any expression that produces an instant vector. The only difference in behaviour between the case where absent is run over a instant vector selector and some other expression is the labels it will return:
* if the argument is an instant vector selector, `absent` will infer the output labels from the selector

* if the argument isn't an instant vector selector, `absent` will always produce a series with no labels (ie. `{}`)

Right, I see. However it may still be neater to have the operator called specifically in query.go to avoid passing both an evaluated and non-evaluated expression to all the functions (which feels odd).

pkg/streamingpromql/operators/functions/absent.go

charleskorn · 2025-02-17T23:36:15Z

However it may still be neater to have the operator called specifically in query.go to avoid passing both an evaluated and non-evaluated expression to all the functions (which feels odd).

Are you referring to passing the expression to the function factory? If so, I don't mind this so much, but I don't hold this opinion strongly. Making absent a special case also feels a bit odd to me.

jhesketh · 2025-02-18T00:50:05Z

However it may still be neater to have the operator called specifically in query.go to avoid passing both an evaluated and non-evaluated expression to all the functions (which feels odd).

Are you referring to passing the expression to the function factory? If so, I don't mind this so much, but I don't hold this opinion strongly. Making absent a special case also feels a bit odd to me.

Yes, that's what I was referring to.
It feels odd to me to pass both the parsed arguments, and unparsed expressions. Especially since the expressions are only used by absent.
I don't hold this strongly either, but find of the two oddities, handling absent in a special case is my personal preference.

lamida · 2025-02-25T17:33:54Z

It feels odd to me to pass both the parsed arguments, and unparsed expressions. Especially since the expressions are only used by absent.

@jhesketh I will consider to refactor this once I am making progress with absent_over_time. Probably with that absent_over_time, it will be more obvious, whether we should pull up passing the unparsed arguments, or instead we should just keep this way. But for now, I prefer just to keep this and have the PR merged once all other concerns are addressed.

But you have some valid points.

pkg/streamingpromql/operators/functions/absent.go

pkg/streamingpromql/testdata/ours/functions.test

charleskorn

LGTM.

Could you please add a benchmark for this and post a comparison with Prometheus' engine? Assuming the results look OK, I'm happy for this to be merged, but if there's any significant regression with MQE, then let's discuss before merging.

(let me know if you need some help with this)

charleskorn · 2025-02-28T00:51:01Z

pkg/streamingpromql/operators/functions/absent.go

+	output := types.InstantVectorSeriesData{}
+	defer func() { a.called = true }()
+
+	if a.called {
+		return output, types.EOS
+	}


[nit] I think this can be simplified a bit:

Suggested change

output := types.InstantVectorSeriesData{}

defer func() { a.called = true }()

if a.called {

return output, types.EOS

}

if a.called {

return output, types.EOS

}

a.called = true

output := types.InstantVectorSeriesData{}

Also, I think a name like exhausted or used might be clearer - called is a little ambiguous (what was called?)

Renamed to exhausted and refactor as suggested by removing defer. Except that output initiliasation still need on the top of the line before the first return.

c1536e7

charleskorn

(meant to approve with the comment above 🤦)

lamida · 2025-02-28T02:39:27Z

Could you please add a benchmark for this and post a comparison with Prometheus' engine? Assuming the results look OK, I'm happy for this to be merged, but if there's any significant regression with MQE, then let's discuss before merging.

Sure I will do the benchmarking shortly. Thanks for checking.

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Co-authored-by: Charles Korn <charleskorn@users.noreply.github.com>

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Co-authored-by: Charles Korn <charleskorn@users.noreply.github.com>

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

lamida marked this pull request as ready for review January 27, 2025 14:57

lamida requested a review from a team as a code owner January 27, 2025 14:57

lamida requested a review from charleskorn January 27, 2025 17:31

charleskorn reviewed Jan 28, 2025

View reviewed changes

lamida force-pushed the lamida/mqe-absent-function branch 2 times, most recently from d7818b2 to 51762d5 Compare February 10, 2025 03:21

lamida force-pushed the lamida/mqe-absent-function branch from 66776aa to 607c848 Compare February 13, 2025 19:47

lamida requested a review from charleskorn February 13, 2025 20:29

charleskorn reviewed Feb 13, 2025

View reviewed changes

charleskorn reviewed Feb 14, 2025

View reviewed changes

pkg/streamingpromql/testdata/ours/functions.test Outdated Show resolved Hide resolved

lamida requested a review from charleskorn February 14, 2025 10:42

jhesketh reviewed Feb 16, 2025

View reviewed changes

jhesketh reviewed Feb 17, 2025

View reviewed changes

pkg/streamingpromql/operators/functions/absent.go Outdated Show resolved Hide resolved

pkg/streamingpromql/operators/functions/absent.go Show resolved Hide resolved

charleskorn reviewed Feb 17, 2025

View reviewed changes

charleskorn mentioned this pull request Feb 18, 2025

Mimir Query Engine #10067

Open

lamida force-pushed the lamida/mqe-absent-function branch from c9de184 to c5a1d77 Compare February 25, 2025 07:30

lamida requested review from jhesketh and charleskorn February 25, 2025 18:38

charleskorn reviewed Feb 25, 2025

View reviewed changes

pkg/streamingpromql/operators/functions/absent.go Outdated Show resolved Hide resolved

pkg/streamingpromql/operators/functions/absent.go Outdated Show resolved Hide resolved

pkg/streamingpromql/operators/functions/absent.go Outdated Show resolved Hide resolved

charleskorn reviewed Feb 26, 2025

View reviewed changes

pkg/streamingpromql/operators/functions/absent.go Outdated Show resolved Hide resolved

pkg/streamingpromql/operators/functions/absent.go Show resolved Hide resolved

pkg/streamingpromql/testdata/ours/functions.test Show resolved Hide resolved

lamida requested a review from charleskorn February 27, 2025 17:37

charleskorn reviewed Feb 28, 2025

View reviewed changes

charleskorn approved these changes Feb 28, 2025

View reviewed changes

lamida and others added 24 commits February 28, 2025 11:25

Add more tests that handles multiple input series

9488007

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Absent don't need dedup

14584ef

Co-authored-by: Charles Korn <charleskorn@users.noreply.github.com>

Fix missing absent function call in test

3ae7de8

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Show range selector result for multiseries in the test

4ba13ef

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Fix wrong test

6eeef91

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Track sample presence

a03c891

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Fix lint

b2d7b76

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Replace presence map with slice

8532b13

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Use 6m step and load in our tests

bbdf90e

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Remove range evaluation from absent test

eacd0e8

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Simplify initilise presence slice

687d861

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Optimise setting presence slice

8338e99

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Allocate bool slice only if we know we have point to add

2feb0ce

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

No need to check presence nil slice

4607100

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Return everytime there is non nil err

4ca22a0

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

No need index to initialise slice pool

b0d7bd7

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Fix incorrect defer

1356821

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Refactor presence check

ec761d0

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Update pkg/streamingpromql/operators/functions/absent.go

e55c37a

Co-authored-by: Charles Korn <charleskorn@users.noreply.github.com>

Enfore NextSeries to be called only once

c5c14b8

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Fix whitespace

b07a9ab

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Fix lint

3a91fef

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Add AGPL header

f4f1ba8

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Rename called to exhausted and remove defer

c1536e7

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

lamida force-pushed the lamida/mqe-absent-function branch from 3a3f44a to c1536e7 Compare February 28, 2025 03:25

lamida added 2 commits February 28, 2025 11:30

The rebase was not correclty applied

ff1121f

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

Update the benchmark for absent

3cc7a85

Signed-off-by: Jon Kartago Lamida <me@lamida.net>

lamida enabled auto-merge (squash) February 28, 2025 04:00

lamida merged commit 1065d86 into main Feb 28, 2025
28 checks passed

lamida deleted the lamida/mqe-absent-function branch February 28, 2025 04:08

MQE Absent function #10523

MQE Absent function #10523

Uh oh!

Conversation

lamida commented Jan 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What this PR does

Benchmark

Which issue(s) this PR fixes or relates to

Checklist

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lamida commented Feb 13, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jhesketh left a comment

Choose a reason for hiding this comment

Uh oh!

charleskorn commented Feb 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jhesketh commented Feb 17, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

charleskorn commented Feb 17, 2025

Uh oh!

jhesketh commented Feb 18, 2025

Uh oh!

lamida commented Feb 25, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

charleskorn left a comment

Choose a reason for hiding this comment

Uh oh!

charleskorn Feb 28, 2025

Choose a reason for hiding this comment

Uh oh!

lamida Feb 28, 2025

Choose a reason for hiding this comment

Uh oh!

charleskorn left a comment

Choose a reason for hiding this comment

Uh oh!

lamida commented Feb 28, 2025

Uh oh!

Uh oh!

Uh oh!

lamida commented Jan 27, 2025 •

edited

Loading

charleskorn commented Feb 17, 2025 •

edited

Loading