Engine Merge obeys row policy #50209

ilejn · 2023-05-24T22:03:38Z

Changelog category (leave one):

Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Engine Merge filters the records according to the row policies of the underlying tables.

robot-ch-test-poll2 · 2023-05-24T22:11:12Z

This is an automated comment for commit c2816ec with description of existing statuses. It's updated for the latest CI running

❌ Click here to open a full report in a separate page

Successful checks

Check name	Description	Status
AST fuzzer	Runs randomly generated queries to catch program errors. The build type is optionally given in parenthesis. If it fails, ask a maintainer for help	✅ success
ClickHouse build check	Builds ClickHouse in various configurations for use in further steps. You have to fix the builds that fail. Build logs often has enough information to fix the error, but you might have to reproduce the failure locally. The cmake options can be found in the build log, grepping for cmake. Use these options and follow the general build process	✅ success
Compatibility check	Checks that clickhouse binary runs on distributions with old libc versions. If it fails, ask a maintainer for help	✅ success
Docker image for servers	The check to build and optionally push the mentioned image to docker hub	✅ success
Fast test	Normally this is the first check that is ran for a PR. It builds ClickHouse and runs most of stateless functional tests, omitting some. If it fails, further checks are not started until it is fixed. Look at the report to see which tests fail, then reproduce the failure locally as described here	✅ success
Flaky tests	Checks if new added or modified tests are flaky by running them repeatedly, in parallel, with more randomization. Functional tests are run 100 times with address sanitizer, and additional randomization of thread scheduling. Integrational tests are run up to 10 times. If at least once a new test has failed, or was too long, this check will be red. We don't allow flaky tests, read the doc	✅ success
Install packages	Checks that the built packages are installable in a clear environment	✅ success
Mergeable Check	Checks if all other necessary checks are successful	✅ success
Push to Dockerhub	The check for building and pushing the CI related docker images to docker hub	✅ success
SQLTest	There's no description for the check yet, please add it to tests/ci/ci_config.py:CHECK_DESCRIPTIONS	✅ success
SQLancer	Fuzzing tests that detect logical bugs with SQLancer tool	✅ success
Sqllogic	Run clickhouse on the sqllogic test set against sqlite and checks that all statements are passed	✅ success
Stateful tests	Runs stateful functional tests for ClickHouse binaries built in various configurations -- release, debug, with sanitizers, etc	✅ success
Style Check	Runs a set of checks to keep the code style clean. If some of tests failed, see the related log from the report	✅ success
Unit tests	Runs the unit tests for different release types	✅ success
Upgrade check	Runs stress tests on server version from last release and then tries to upgrade it to the version from the PR. It checks if the new server can successfully startup without any errors, crashes or sanitizer asserts	✅ success

Check name	Description	Status
CI running	A meta-check that indicates the running CI. Normally, it's in success or pending state. The failed status indicates some problems with the PR	⏳ pending
Integration tests	The integration tests report. In parenthesis the package type is given, and in square brackets are the optional part/total tests	❌ failure
Performance Comparison	Measure changes in query performance. The performance test report is described in detail here. In square brackets are the optional part/total tests	❌ failure
Stateless tests	Runs stateless functional tests for ClickHouse binaries built in various configurations -- release, debug, with sanitizers, etc	❌ failure
Stress test	Runs stateless functional tests concurrently from several clients to detect concurrency-related errors	❌ failure

ilejn · 2023-06-14T17:40:59Z

I believe that test failures in 00002_log_and_exception_messages_formatting are irrelevant, although AST fuzzer (ubsan) issue IS relevant.
Fuzzer generated a query
SELECT rand(NULL) % NULL FROM merge(currentDatabase(), '02763_merge_merge_1') PREWHERE (rand(1.1920928955078125e-7) % -0.) % rand(NULL) ORDER BY rand(2) % -1 DESC NULLS FIRST, x DESC NULLS FIRST SETTINGS optimize_move_to_prewhere = 0
which is wrong and produces
Code: 59. DB::Exception: Received from localhost:9000. DB::Exception: Illegal type Float64 of column for filter. Must be UInt8 or Nullable(UInt8) or Const variants of them.: While executing MergeTreeInOrder. (ILLEGAL_TYPE_OF_COLUMN_FOR_FILTER)
on my laptop.
Looking into why it crashes in CI environment.

Enmk

Please address the comments

Enmk · 2023-07-10T08:19:27Z

src/Storages/StorageMerge.cpp

+                        required_columns.begin(), required_columns.end(),
+                        std::inserter(filter_columns, filter_columns.begin()));
+
+                    source_step_with_filter->addFilter(filter_dag_ptr, filter_columns.front());


Why only with columns from RBAC that are not required columns? Would filter work properly if there are some required columns in expression?

Why only 1 filter, what if there are many columns?

The 'extra' column here is a complex condition that may contain logical operations, something like
'field1 > x and field2 > y'.
That's why only one filter is needed.
I'll try to simulate collision between the column and what is listed in required column, this is an interesting point.

It might be beneficial to put that info into comment for this piece of code

Enmk · 2023-07-10T08:24:44Z

src/Storages/StorageMerge.cpp

+                    Names filter_columns;
+
+                    std::set_difference(fa_actions_columns_sorted.begin(), fa_actions_columns_sorted.end(),
+                        required_columns.begin(), required_columns.end(),


Suggested change

required_columns.begin(), required_columns.end(),

required_columns_sorted.begin(), required_columns_sorted.end(),

Yes, required_columns_sorted must be used, thanks.

Enmk · 2023-07-10T08:26:32Z

src/Storages/StorageMerge.cpp

+
+
+        std::set_difference(fa_actions_columns_sorted.begin(), fa_actions_columns_sorted.end(),
+            required_columns.begin(), required_columns.end(),


Same here, required_columns_sorted ?

Enmk · 2023-07-10T08:30:10Z

src/Storages/StorageMerge.cpp

+    if (row_policy_filter)
+    {
+        ASTPtr expr = row_policy_filter->expression;
+
+        auto syntax_result = TreeRewriter(local_context).analyze(expr, pipe_columns);
+        auto expression_analyzer = ExpressionAnalyzer{row_policy_filter->expression, syntax_result, local_context};
+
+        auto actions_dag = expression_analyzer.getActionsDAG(true, false);
+        auto filter_actions = std::make_shared<ExpressionActions>(actions_dag, ExpressionActionsSettings::fromContext(local_context, CompileExpressions::yes));
+        auto required_columns = filter_actions->getRequiredColumns();
+
+        LOG_TRACE(&Poco::Logger::get("ReadFromMerge::convertinfSourceStream"), "filter_actions_dag: {},<> {}, <> {}",
+            filter_actions->getActionsDAG().dumpNames(), filter_actions->getActionsDAG().dumpDAG(), filter_actions->getSampleBlock().dumpStructure());
+
+
+        auto fa_actions_columns_sorted = filter_actions->getSampleBlock().getNames();
+        std::sort(fa_actions_columns_sorted.begin(), fa_actions_columns_sorted.end());
+
+        Names required_columns_sorted = required_columns;
+        std::sort(required_columns_sorted.begin(), required_columns_sorted.end());
+
+        Names filter_columns;
+
+
+        std::set_difference(fa_actions_columns_sorted.begin(), fa_actions_columns_sorted.end(),
+            required_columns.begin(), required_columns.end(),
+            std::inserter(filter_columns, filter_columns.begin()));


This looks almost identical to the code block you've added above, maybe factor it out into a function?

Enmk · 2023-07-10T08:35:28Z

tests/queries/0_stateless/02763_row_policy_storage_merge.sql.j2

Maybe also add RBAC based on 2 or more columns?

Enmk · 2023-07-17T11:38:17Z

src/Storages/StorageMerge.cpp

    }

    return builder;
 }

+void ReadFromMerge::RowPolicyData::init(RowPolicyFilterPtr row_policy_filter_ptr_,
+    const std::shared_ptr<DB::IStorage> storage,


Please remove const here, it breaks the build.

Jul 14 14:16:37 /build/src/Storages/StorageMerge.cpp:391:9: error: parameter 2 is const-qualified in the function declaration; const-qualification of parameters only has an effect in function definitions [readability-avoid-const-params-in-decls,-warnings-as-errors]

https://s3.amazonaws.com/clickhouse-builds/PRs/50209/9739ed2bb4eabf2b20083bc324186d319faf34b7/binary_tidy/build_log.log

Enmk · 2023-07-17T11:39:03Z

src/Storages/StorageMerge.cpp

@@ -961,8 +1129,14 @@ void ReadFromMerge::convertingSourceStream(
    const Aliases & aliases,
    ContextPtr local_context,
    QueryPipelineBuilder & builder,
-    const QueryProcessingStage::Enum & processed_stage)
+    const QueryProcessingStage::Enum & processed_stage,


Please remove const here too

ilejn · 2023-09-12T09:17:18Z

Integration test failures are not related.

Please, review proposed change.

ilejn · 2023-09-27T17:19:58Z

Hello @vitlibar ,
could you please review this PR and/or attract someone's attention to it?

src/Storages/StorageMerge.cpp

vitlibar · 2023-09-29T20:19:48Z

src/Storages/StorageMerge.cpp

+
+            if (row_policy_data.needCare())
+            {
+                if (auto * source_step_with_filter = dynamic_cast<SourceStepWithFilter*>((plan.getRootNode()->step.get())))


What if that step is not derived from SourceStepWithFilter?

Please add a comment here that it's an optimization if the storage can use filtering while reading (e.g. MergeTree).

The conditions can be combined to be shorter:

if (plan.isInitialized() and row_policy_data) { if (auto * source_step_with_filter = dynamic_cast<SourceStepWithFilter*>((plan.getRootNode()->step.get()))) row_policy_data->addStorageFilter(source_step_with_filter); }

It is not obvious that "The conditions can be combined to be shorter".
Could you explain why you are sure that going further with non initialized plan is safe and makes sense?

vitlibar · 2023-09-29T20:21:11Z

src/Storages/StorageMerge.cpp

 {
+    if (row_policy_data.needCare())


needCare() is not a very clear name. row_policy_data.hasRowPolicy() seems better.

Renamed, although it is usually considered bad practice if a method name contain class name (something like Vector::vector_size).
Nevertheless, let it be 'hasRowPolicy'.

Better remove the method at all, see #50209 (comment)

vitlibar · 2023-09-29T20:24:07Z

tests/queries/0_stateless/02763_row_policy_storage_merge.sql.j2

+DROP ROW POLICY IF EXISTS 02763_filter_4 ON 02763_merge_merge_1;
+DROP ROW POLICY IF EXISTS 02763_filter_5 ON 02763_merge_fancycols;
+DROP ROW POLICY IF EXISTS 02763_filter_6 ON 02763_merge_fancycols;
+


Could you please write in the description to this PR a very simple example how it worked before your PR and how it's going to work now?

I have made it by myself:

CREATE TABLE tbl1 (x UInt32) ENGINE=MergeTree ORDER BY tuple(); INSERT INTO tbl1 SELECT number AS x FROM numbers(10); CREATE ROW POLICY r1 ON tbl1 USING x%2==0 TO ALL; SELECT * FROM tbl1; -- applies the row policy SELECT * FROM merge('default', 'tbl1'); -- doesn't apply the row policy

Original behavior is described in the issue
#50163
which is referred.

My example is shorter :)

ilejn · 2023-10-24T21:44:51Z

Hello @vitlibar , I think I've addressed your comments, except binary search.
I've switched to NamesAndTypesList::getDifference in RowPolicyData ctor , which probably made code more natural.

ilejn · 2023-11-01T10:50:23Z

Hello @vitlibar , do you have ideas/suggestions what should be improved further?
Should I rebase onto fresh master?

vitlibar · 2023-11-01T18:37:11Z

src/Storages/StorageMerge.cpp

+    auto storage_columns = storage_metadata_snapshot->getColumns();
+    auto needed_columns = storage_columns.getAllPhysical();
+
+    auto syntax_result = TreeRewriter(local_context).analyze(expr, needed_columns);


Is it ok to use here storage_columns.getAllPhysical() as needed_columns? I mean what if a storage has an alias column and a row policy filter involves that alias? Can you please check that case?

Is it ok to use here storage_columns.getAllPhysical() as needed_columns? I mean what if a storage has an alias column and a row policy filter involves that alias? Can you please check that case?

Very good point, quite possible that it is a mistake. Will check.

Indeed we have an issue here.
But the problem is deeper, Engine Merge is not very good at aliases in general, they seem do not work via function https://fiddle.clickhouse.com/4648fd3d-3c1a-462e-919b-90dafcafc131

The problem is even deeper ;) , looks like Engine Merge does not work with tables with different structures. And 'does not work' means gives wrong result.
https://fiddle.clickhouse.com/27a3d1d1-51a4-491c-b6ad-5421c11d1f0b
Sorry, my bad.

Summing up issue with alias based row policy.
I've fixed it for cases where Engine Merge is used directly, not via table function merge(). Works ok. Not pushed yet.
I haven't fixed it for merge() yet.
How it looks https://pastila.nl/?00d38392/d68be2ab03fb64572f5a3f1b3bd383b2#2RTThmq3iM6cf1C1nLQVUg==

Alternatively we can

expose alias via merge table function - trivial fix. Can be treated as improvement, but breaks current behavior.

go with this PR further as is and fix alias based row policy for merge() in a separated PR.

expose alias via merge table function - trivial fix. Can be treated as improvement, but breaks current behavior.

I suppose ideally the merge() function should return aliases only if that alias is requested explicitly; or if asterisk is used and asterisk_include_alias_columns is set. Because it's better to be consistent with how we treat aliases for normal tables. But of course if an alias is used in a row policy then it should work no matter if we're going to show that alias to a user or not.

I assume that it is easier to have both

merge() properly returning aliases and respecting asterisk_include_alias_columns

alias based row policies working

implemented,
than (2) but not (1).
Working on it.

Of course a better solution is always better, I just suggested to split work between two PRs. But ok, no problem, do it the way you like.

The last push brought many changes in terms of lines of code, but actually changes are not significant.
I've introduced processAliases method because I have to add new names (extendNames) before taking care of aliases and it is too much clutter for initializePipeline().
Besides this, I renamed convertingSourceStream to convertAndFilterSourceStream , because filtering is needed in-between.

asterisk_include_alias_columns is left out of the scope despite my initial intention (and claim).

src/Storages/StorageMerge.cpp

vitlibar · 2023-11-13T10:41:03Z

src/Storages/StorageMerge.h

+        const StorageWithLockAndName & storage_with_lock,
+        Aliases & aliases,
+        const Block & sample_block,
+        ContextMutablePtr modified_context);


What exactly does the function processAliases() do? I know it's named processAliases but how it processes them? A comment here would be useful.

I've moved this part of code as is, but Ok, I'll try to add some comments.

vitlibar · 2023-11-13T10:42:21Z

src/Storages/StorageMerge.cpp

+    }
+
+    Aliases aliases;
+    processAliases(real_column_names, storage_with_lock, aliases, sample_block, modified_context);


Maybe it's better to rename processAliases to extractAliases() and make it return Aliases?

vitlibar · 2023-11-13T10:45:53Z

src/Storages/StorageMerge.cpp

+    }
+    if (!column_names_as_aliases.empty())
+    {
+        real_column_names = column_names_as_aliases;


So it replaces real_column_names only if !column_names_as_aliases.empty()
Such a complicated logic.

Let's make the function processAliases() return column_names_as_aliases as is, and do replace read_column_names not here but in the function which calls processAliases()?

processAliases should return both Aliases and column_names, that's why non constant references are used.
I think that a more complex approach with std::pair or something does not make real sense here.
If you don't mind I would like to keep processAliases in its current form and to avoid further interference in its logic.

You've made an interference already, it doesn't add too much interference if we pass arguments this way or another. I'm ok with passing results via reference parameters. For me it's just better to avoid parameters which are input and output at the same time - just to make the life easier. So why can't we make column_names_as_aliases a pure output parameter and do this logic in createSource():

QueryPipelineBuilderPtr ReadFromMerge::createSources(...) { ... Aliases aliases; Names column_names_as_aliases; processAliases(context, storage_with_lock, sample_block, modified_context, aliases, column_names_as_aliases); if (!column_names_as_aliases.empty()) real_column_names = column_names_as_aliases; ...

May be because this logic is a part of processing aliases ? ;)

vitlibar · 2023-11-15T17:02:25Z

Please resolve the conflict

ilejn · 2023-11-15T20:35:26Z

Please resolve the conflict

The changes are rather significant.

merge_row_policy: changes because of refactoring in master

vitlibar · 2023-11-21T16:35:25Z

tests/analyzer_tech_debt.txt

@@ -42,6 +42,7 @@
 02003_WithMergeableStateAfterAggregationAndLimit_LIMIT_BY_LIMIT_OFFSET
 02404_memory_bound_merging
 02725_agg_projection_resprect_PK
+02763_row_policy_storage_merge_alias


Basically because of

$ grep merge_alias analyzer_tech_debt.txt 01214_test_storage_merge_aliases_with_where 01925_test_storage_merge_aliases

Aliases in Merge do not work if analyzer enabled.

vitlibar · 2023-11-21T17:05:59Z

The build is broken (see). Can you please fix it?

ilejn · 2023-11-21T21:45:22Z

The build is broken (see). Can you please fix it?

Yes, sure.

ilejn · 2023-11-22T08:11:17Z

@vitlibar
I've got rid of processAliases because in refactored version there fewer reasons for it.
And no problem of non-const references as a bonus.
To keep balance ;) I've introduced rvalue reference for real_column_names which might be strange but seems the best option to me.

ilejn · 2023-11-23T08:03:26Z

Test failures are not caused by the proposed changes.

vitlibar · 2023-11-25T09:34:03Z

Test failures:
01280_ttl_where_group_by - see issue 49699
02152_http_external_tables_memory_tracking - fixed in #57130
test_dictionaries_postgresql/test.py::test_postgres_schema - see issue 56627
test_backup_restore_on_cluster/test_concurrency.py::test_create_or_drop_tables_during_backup[Replicated-ReplicatedMergeTree] - fixed in #57007
test_system_clusters_actual_information/test.py::test - see issue 55168
test_storage_kafka/test.py::test_system_kafka_consumers_rebalance_mv - see issue 56043

ilejn mentioned this pull request May 24, 2023

Merge tables should additionally check row policies for all underlying tables. #50163

Closed

robot-ch-test-poll2 added the pr-bugfix Pull request with bugfix, not backported by default label May 24, 2023

alexey-milovidov removed the pr-bugfix Pull request with bugfix, not backported by default label Jun 1, 2023

robot-ch-test-poll1 added the pr-improvement Pull request with some product improvements label Jun 6, 2023

ilejn force-pushed the merge_row_policy branch from f59e7df to 4df253a Compare June 14, 2023 21:37

ilejn changed the title ~~WIP Engine Merge obeys row policy~~ Engine Merge obeys row policy Jun 15, 2023

ilejn force-pushed the merge_row_policy branch from 4df253a to 4d8ad1a Compare June 25, 2023 21:24

Enmk suggested changes Jul 10, 2023

View reviewed changes

ilejn force-pushed the merge_row_policy branch from 4d8ad1a to 34a787c Compare July 12, 2023 14:43

Enmk reviewed Jul 17, 2023

View reviewed changes

ilejn force-pushed the merge_row_policy branch from 9739ed2 to 7379fa4 Compare July 19, 2023 08:41

ilejn marked this pull request as ready for review September 18, 2023 08:39

vitlibar self-assigned this Sep 28, 2023

vitlibar reviewed Sep 29, 2023

View reviewed changes

src/Storages/StorageMerge.cpp Outdated Show resolved Hide resolved

vitlibar reviewed Sep 29, 2023

View reviewed changes

ilejn force-pushed the merge_row_policy branch from 79c7c4f to 0554018 Compare October 4, 2023 09:09

ilejn added 7 commits October 5, 2023 08:23

merge_row_policy: initial

020d76a

merge_row_policy: tiny cleanup

1e31859

merge_row_policy: with QueryProcessingStage::Complete

0464b4f

merge_row_policy - extra debug

2acc4c2

merge_row_policy: original behavior restored + extra debug

eebdff4

merge_row_policy: FilterTransform

b6f682d

merge_row_policy: row_policy_storage_merge test

657c39c

ilejn added 2 commits October 16, 2023 22:16

merge_row_policy: get rid of RequiredSourceColumnsVisitor

786183e

merge_row_policy: cleanup

0d27150

vitlibar reviewed Nov 1, 2023

View reviewed changes

src/Storages/StorageMerge.cpp Outdated Show resolved Hide resolved

vitlibar reviewed Nov 1, 2023

View reviewed changes

src/Storages/StorageMerge.cpp Outdated Show resolved Hide resolved

vitlibar reviewed Nov 1, 2023

View reviewed changes

src/Storages/StorageMerge.cpp Outdated Show resolved Hide resolved

ilejn added 4 commits November 6, 2023 15:18

merge_row_policy: aliases slightly work

d900f8a

merge_row_policy: processAliases introduced

6e189b1

merge_row_policy: cleanup, traces removed

f1af834

merge_row_policy: style fix, cleanup

9c5f9f7

vitlibar reviewed Nov 13, 2023

View reviewed changes

merge_row_policy: some comments

3e6a5d8

vitlibar approved these changes Nov 15, 2023

View reviewed changes

ilejn added 3 commits November 17, 2023 09:19

Merge remote-tracking branch 'origin/master' into merge_row_policy

9eeae56

merge_row_policy: changes because of refactoring in master

merge_row_policy: cleanup after merge

a8abec6

merge_row_policy: alias test separated out and disabled if analyzer

f226fa6

vitlibar reviewed Nov 21, 2023

View reviewed changes

merge_row_policy: make clang-tidy happy

c2816ec

vitlibar merged commit 2e7f314 into ClickHouse:master Nov 25, 2023
319 of 336 checks passed

	required_columns.begin(), required_columns.end(),
	required_columns_sorted.begin(), required_columns_sorted.end(),



		std::set_difference(fa_actions_columns_sorted.begin(), fa_actions_columns_sorted.end(),
		required_columns.begin(), required_columns.end(),

Engine Merge obeys row policy #50209

Engine Merge obeys row policy #50209

Conversation

ilejn commented May 24, 2023 • edited by alexey-milovidov Loading

Changelog category (leave one):

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

robot-ch-test-poll2 commented May 24, 2023 • edited by robot-ch-test-poll3 Loading

ilejn commented Jun 14, 2023

Enmk left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Enmk Jul 10, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Enmk Jul 17, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ilejn commented Sep 12, 2023

ilejn commented Sep 27, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar Sep 29, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ilejn commented Oct 24, 2023

ilejn commented Nov 1, 2023

vitlibar Nov 1, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ilejn Nov 6, 2023 • edited Loading

Choose a reason for hiding this comment

ilejn Nov 6, 2023 • edited Loading

Choose a reason for hiding this comment

vitlibar Nov 6, 2023 • edited Loading

Choose a reason for hiding this comment

ilejn Nov 6, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ilejn Nov 9, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar Nov 13, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar Nov 13, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar Nov 13, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar Nov 15, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar commented Nov 15, 2023

ilejn commented Nov 15, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vitlibar commented Nov 21, 2023

ilejn commented Nov 21, 2023

ilejn commented Nov 22, 2023

ilejn commented Nov 23, 2023

vitlibar commented Nov 25, 2023

ilejn commented May 24, 2023 •

edited by alexey-milovidov

Loading

robot-ch-test-poll2 commented May 24, 2023 •

edited by robot-ch-test-poll3

Loading

Enmk Jul 10, 2023 •

edited

Loading

Enmk Jul 17, 2023 •

edited

Loading

vitlibar Sep 29, 2023 •

edited

Loading

vitlibar Nov 1, 2023 •

edited

Loading

ilejn Nov 6, 2023 •

edited

Loading

ilejn Nov 6, 2023 •

edited

Loading

vitlibar Nov 6, 2023 •

edited

Loading

ilejn Nov 6, 2023 •

edited

Loading

ilejn Nov 9, 2023 •

edited

Loading

vitlibar Nov 13, 2023 •

edited

Loading

vitlibar Nov 13, 2023 •

edited

Loading

vitlibar Nov 13, 2023 •

edited

Loading

vitlibar Nov 15, 2023 •

edited

Loading