iframe-proxy

scanhex12 · 2026-04-08T17:43:47Z

Changelog category (leave one):

New Feature

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

Query condition cache for iceberg tables

Documentation entry for user-facing changes

Documentation is written (mandatory for new features)

Version info

Merged into: 26.5.1.367

scanhex12 · 2026-04-08T17:44:22Z

clickhouse-gh · 2026-04-08T17:44:26Z

Workflow [PR], commit [e80e78d]

Summary: ❌

job_name	test_name	status	info	comment
Finish Workflow		FAIL
	python3 ./ci/jobs/scripts/workflow_hooks/feature_docs.py	FAIL

AI Review

Summary

This PR adds query-condition-cache integration for Iceberg/object-storage reads by propagating table identity (StorageID with UUID), adding condition hashing in FormatFilterInfo, and wiring row-group pruning for Parquet. High-level verdict: ⚠️ Request changes due to two correctness/safety issues still present in cache keying and UUID parsing, plus PR metadata incompleteness.

PR Metadata

⚠️ Changelog category (New Feature) is plausible for this change, but the documentation checkbox is still unchecked.
⚠️ For a New Feature, either provide docs or explicitly justify why none are needed in the PR description.

Findings

❌ Blockers

[src/Storages/ObjectStorage/StorageObjectStorageSource.cpp:674, src/Storages/ObjectStorage/StorageObjectStorageSource.cpp:552] Query-condition-cache lookup/write uses object_info->getFileName() as object identity. Different objects with the same basename in different directories can alias and reuse incorrect row-group masks, causing wrong pruning decisions and potentially incorrect results.
- Suggested fix: key cache entries by a unique object identity (object_info->getPath() or another full unique storage path), not basename.
[src/Databases/DataLake/DatabaseDataLake.cpp:670] parseFromString<UUID>(*catalog_uuid) is called without validation for external catalog metadata. A malformed catalog UUID causes an exception on table open path.
- Suggested fix: use safe parsing (tryParse/guarded parse) and fall back to UUIDHelpers::Nil with a warning when metadata is invalid.

💡 Nits

[PR description] Documentation checkbox for New Feature remains unchecked.

Tests

⚠️ Add a regression test with two files that have the same basename under different directories and distinct row-group matches, then verify cache hits do not cross-contaminate between them.

ClickHouse Rules

Item	Status	Notes
Deletion logging	➖
Serialization versioning	➖
Core-area scrutiny	✅
No test removal	✅
Experimental gate	➖
No magic constants	✅
Backward compatibility	⚠️	Invalid external `table-uuid` metadata now causes exception during table open
`SettingsChangesHistory.cpp`	➖
PR metadata quality	⚠️	`New Feature` doc checkbox is unchecked
Safe rollout	⚠️	Cache key aliases between same basenames can cause unsafe rollout
Compilation time	⚠️	New heavy includes added to `IObjectIterator.h` increase transitive fan-out
No large/binary files	✅

Final Verdict

Status: ⚠️ Request changes
Minimum required actions:
1. Switch query-condition-cache object key from basename to unique full object identity.
2. Make catalog UUID parsing tolerant to malformed external metadata (no hard failure on parse).
3. Complete PR metadata for New Feature documentation requirement.

clickhouse-gh · 2026-04-08T17:50:48Z

+                format_filter_info->filter_actions_dag->dumpNames(),
+                unmatched_ranges,
+                unmatched_ranges.size(),
+                false


getMatchedBuckets is called for every format when condition_hash is set, but the default implementation in IInputFormat throws NOT_IMPLEMENTED.

This means enabling use_query_condition_cache can throw on non-Parquet object-storage formats after EOF (getMatchedBuckets is not implemented for this format).

Please guard this path by format/capability (or make default behavior non-throwing) so unsupported formats keep working.

clickhouse-gh · 2026-04-08T17:52:00Z

+                *format_filter_info->condition_hash,
+                format_filter_info->filter_actions_dag->dumpNames(),
+                unmatched_ranges,
+                unmatched_ranges.size(),


QueryConditionCache::write expects marks_count to be the total number of marks/row-groups in the file, but here unmatched_ranges.size() is passed instead.

This can break correctness and memory safety. Concrete trace: if only row group 7 matches, then unmatched_ranges = [{7,8}] and unmatched_ranges.size() == 1; inside write, std::fill(entry->matching_marks.begin() + 7, ... + 8, false) is applied to a vector of size 1, which is out-of-bounds.

Please pass the full row-group count (and ensure ranges are in the expected coordinate space).

clickhouse-gh · 2026-04-08T17:52:01Z

+                    unmatched_ranges.push_back({UInt64(rg), UInt64(rg + 1)});
+            }
+            auto query_condition_cache = Context::getGlobalContextInstance()->getQueryConditionCache();
+            std::cerr << "write to cache " << storage_id.getNameForLogs() << ' ' << object_info->getFileName() << '\n';


Looks like accidental debug output left in server path:

std::cerr << "write to cache " << ...

Please replace with LOG_TRACE/LOG_DEBUG (or remove). Writing directly to stderr bypasses ClickHouse logging controls and can pollute service logs.

scanhex12 · 2026-04-09T13:02:57Z

Now (I mean after fixes) is much better:

Mac :) 
Mac :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' and hits = 5 settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 1

Query id: d2767502-349b-4e50-bf3f-5113e89e5af3

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 322.735 sec. 

Mac :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' and hits = 5 settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 1

Query id: 709adee9-4aed-412c-baba-312cf45d4806

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 159.676 sec. 

Mac :) 
Mac :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' and hits = 5 settings use_query_condition_cache = false;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = false

Query id: 61700950-fd26-409d-9426-e739a3cb624e

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 325.656 sec.

clickhouse-gh · 2026-04-09T13:38:23Z

+
+                if (object_info->file_bucket_info)
+                {
+                    auto filtered = object_info->file_bucket_info->filterByMatchingRowGroups(matching_row_groups);


Using object_info->getFileName() as the cache key can cause false cache hits between different objects that share the same basename (e.g. .../part=1/data.parquet and .../part=2/data.parquet).

QueryConditionCache::read/write then mixes predicates for different files under the same storage_id.uuid, and matching_row_groups.empty() can skip an unrelated file entirely. That is a correctness issue (rows can be missed).

Please key by a path-unique identifier (for example object_info->getPathOrPathToArchiveIfArchive() or another stable full object path), and use the same key consistently in both read and write paths.

scanhex12 · 2026-04-09T14:02:08Z

Results on my dev machine


.eu-west-1.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE path = 'Main_Page'
SETTINGS use_query_condition_cache = 1

Query id: 420f2f22-5d0e-4ebe-a210-633b56dadbae

   ┌─sum(hits)─┐
1. │ 213832822 │ -- 213.83 million
   └───────────┘

1 row in set. Elapsed: 20.698 sec. Processed 27.87 thousand rows, 4.17 GB (1.35 thousand rows/s., 201.24 MB/s.)
Peak memory usage: 594.37 MiB.

.eu-west-1.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE path = 'Main_Page'
SETTINGS use_query_condition_cache = 1

Query id: 63558620-0c44-43a3-8cd7-9ac35b50902c

   ┌─sum(hits)─┐
1. │ 213832822 │ -- 213.83 million
   └───────────┘

1 row in set. Elapsed: 15.536 sec. Processed 27.87 thousand rows, 4.17 GB (1.79 thousand rows/s., 268.11 MB/s.)
Peak memory usage: 651.54 MiB.

.eu-west-1.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' settings use_query_condition_cache = 0;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE path = 'Main_Page'
SETTINGS use_query_condition_cache = 0

Query id: e2f5931f-c51a-42dd-951c-d60807b9a276

   ┌─sum(hits)─┐
1. │ 213832822 │ -- 213.83 million
   └───────────┘

1 row in set. Elapsed: 19.473 sec. Processed 27.87 thousand rows, 4.17 GB (1.43 thousand rows/s., 213.90 MB/s.)
Peak memory usage: 649.19 MiB.

.eu-west-1.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE path = 'Main_Page'
SETTINGS use_query_condition_cache = 1

Query id: e61b52b0-ee76-4dcc-a2c6-2fabb93e29ad

   ┌─sum(hits)─┐
1. │ 213832822 │ -- 213.83 million
   └───────────┘

1 row in set. Elapsed: 16.898 sec. Processed 27.87 thousand rows, 4.17 GB (1.65 thousand rows/s., 246.49 MB/s.)
Peak memory usage: 647.91 MiB.

.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' settings use_query_condition_cache = 0;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE path = 'Main_Page'
SETTINGS use_query_condition_cache = 0

Query id: 35792fa5-9a2c-4bdf-aa6d-a618e1a76226

   ┌─sum(hits)─┐
1. │ 213832822 │ -- 213.83 million
   └───────────┘

1 row in set. Elapsed: 18.229 sec. Processed 27.87 thousand rows, 4.17 GB (1.53 thousand rows/s., 228.51 MB/s.)
Peak memory usage: 639.19 MiB.

.eu-west-1.compute.internal :) ```

clickhouse-gh · 2026-04-09T14:46:50Z

 #include <Disks/DiskObjectStorage/ObjectStorages/IObjectStorage.h>
 #include <Processors/ISimpleTransform.h>
 #include <Storages/ObjectStorage/StorageObjectStorageConfiguration.h>
+#include <Interpreters/Cache/QueryConditionCache.h>


This header now pulls in QueryConditionCache and FormatFilterInfo types, but this file is included transitively in many object-storage and interpreter paths.

That increases recompilation fan-out and violates the compilation-time guideline for high-fan-out headers. Please forward-declare these types in the header (e.g. class QueryConditionCache; struct FormatFilterInfo;) and keep the heavy includes in IObjectIterator.cpp.

clickhouse-gh · 2026-04-09T17:54:35Z

        ObjectStoragePtr object_storage_,
-        const ContextPtr & context_);
+        const ContextPtr & context_,
+        const StorageID & storage_id_ = StorageID::createEmpty(),


ObjectIteratorSplitByBuckets now accepts storage_id and format_filter_info specifically for query-condition-cache pruning, but both parameters default to empty values.

That makes this change easy to miss at call sites: StorageObjectStorageCluster::getTaskIteratorExtension still calls this ctor with only 4 args, so cache-aware pruning is silently disabled for cluster table-function reads when cluster_table_function_split_granularity='bucket'. Because createReader skips cache lookup when file_bucket_info is already set, those tasks never get pruned later either.

Please pass storage_id and a real FormatFilterInfo here (same as non-cluster object-storage reads), or remove defaults to force all call sites to wire these parameters explicitly.

This seems legitimate, but the downside doesn't seem risky. In the case where we have empty values, we just won't engage the cache, but it is not like we would be doing anything incorrect. Honestly, I think it is fine as-is

alexey-milovidov · 2026-04-09T21:01:45Z

The Can't adjust last granule error in CI is a known issue. The fix is in #101641

alexey-milovidov · 2026-04-09T21:27:02Z

The test_database_iceberg_nessie_catalog and test_database_iceberg_lakekeeper_catalog failures in this PR's CI are false positives caused by a bug in PR #100583 (merged to master on 2026-04-09 at 11:19 UTC). That PR broke the metadata file path resolution in getMetadataLocation for Iceberg catalog tests — it produced a doubled S3 key like table_path/warehouse-rest/table_path/metadata/... instead of table_path/metadata/..., causing MinIO to return 403.

The bug was reverted by PR #102247 (merged at ~14:48 UTC). All nessie/lakekeeper test failures between these two merges are unrelated to this PR.

alexey-milovidov · 2026-04-10T02:27:46Z

The flaky check failure is fixed in #102148, let's update the branch.

clickhouse-gh · 2026-04-13T13:48:58Z

+
+
+@pytest.mark.parametrize("storage_type", ["s3", "azure", "local"])
+def test_query_condition_cache(started_cluster_iceberg_with_spark, storage_type):


QueryConditionCache correctness here depends on object identity. Since this PR uses file-name-based cache keys in object-storage paths, please add a regression case that creates two files in different directories with the same basename (for example .../a/data.parquet and .../b/data.parquet) and validates that cache hits for one file do not affect the other.

Without this, we can miss key-collision regressions while still seeing green cache-hit counters.

scanhex12 · 2026-04-22T13:03:52Z

compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' AND hits = 5 settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 1

Query id: e8b9904e-ccd4-4e7e-86f6-6516f3e09eb0

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 17.293 sec. 

.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' AND hits = 5 settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 1

Query id: 2026a171-85b4-4e11-80e4-fedf76bc93f9

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 9.424 sec. 

compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' AND hits = 5 settings use_query_condition_cache = 0;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 0

Query id: aa729a8f-97e5-4acd-bd91-a2bd0877c067

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 17.252 sec. 
.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' AND hits = 5 settings use_query_condition_cache = 0;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 0

Query id: d9896285-bf78-478e-9aef-5f0fdecf3a1c

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 19.845 sec. 

.compute.internal :) select sum(hits) from wikistat_iceberg2 where path = 'Main_Page' AND hits = 5 settings use_query_condition_cache = 1;

SELECT sum(hits)
FROM wikistat_iceberg2
WHERE (path = 'Main_Page') AND (hits = 5)
SETTINGS use_query_condition_cache = 1

Query id: 08b550da-c5b4-45a0-8ae8-3fe5480cb3a0

   ┌─sum(hits)─┐
1. │      3905 │
   └───────────┘

1 row in set. Elapsed: 9.663 sec.

clickhouse-gh · 2026-04-23T14:12:44Z

    , prewhere_info(std::move(prewhere_info_))
    , column_mapper(column_mapper_)
 {
+    bool use_query_condition_cache = context_->getSettingsRef()[Setting::use_query_condition_cache];


Changelog category is New Feature, but the PR template still has the documentation checkbox unchecked. This change is user-facing (use_query_condition_cache now affects Iceberg/object-storage reads), so please add docs or explicitly explain why no documentation update is needed.

clickhouse-gh · 2026-05-06T19:11:16Z

LLVM Coverage Report

Metric	Baseline	Current	Δ
Lines	84.10%	84.10%	+0.00%
Functions	91.10%	91.10%	+0.00%
Branches	76.60%	76.60%	+0.00%

Changed lines: 61.25% (147/240) | lost baseline coverage: 8 line(s) · Uncovered code

Full report · Diff report

clickgapai · 2026-05-07T04:24:49Z

Hi @scanhex12 @grantholly-clickhouse — while reviewing this PR I found the following:

Bug (P1): Cluster mode with cluster_table_function_split_granularity=BUCKET poisons the query condition cache by writing 'no match' for row groups assigned to other workers — [CI Proof] Cluster mode with cluster_table_function_split_granularity=BUCKET poisons the query condition cache by writing 'no match' for row groups assigned to other workers #104278

Happy to discuss — close anything that's wrong or already addressed.

clickgapai · 2026-05-07T09:34:02Z

Update on the earlier review: CI on the proof PR #104278 confirmed the suspected bug via Integration tests (amd_tsan, 1/6). Filed as a separate Issue so it isn't lost when the proof PR closes:

Cluster mode with cluster_table_function_split_granularity=BUCKET poisons the query condition cache by writing 'no match' for row groups assigned to other workers — Cluster mode with cluster_table_function_split_granularity=BUCKET poisons the query condition cache by writing 'no match' for row groups assigned to other workers #104288

(Automated follow-up. The proof PR has been closed; please track the bug on the Issue.)

clickgapai · 2026-05-07T11:13:53Z

Correction — false-positive confirmation. My earlier comment at 11:01 UTC on this PR claimed the QCC bucket-poisoning bug was confirmed by CI on proof PR #104278. That was wrong — both the previous Issue #104288 (closed as not_planned) and this confirmation were the same false-positive shape: a Python ValueError: Mismatch in fields in the proof test's pyiceberg setup, mis-classified by the bot as a sanitizer-detected bug.

Two distinct bot bugs combined: (1) the proof test had a real schema mismatch (int32 schema vs int64 dataframe), and (2) when re-evaluating the reopened proof PR, the bot picked up the stale CI run's log instead of the latest run, locking in the same wrong classification.

Dropping this finding. The bug claim about cluster_table_function_split_granularity=BUCKET poisoning the QCC remains plausible but unverified — a maintainer can re-investigate with a working test if interested.

…terInfo Bot review on PR ClickHouse#104225 surfaced that the `isQueryConditionCacheWritable` predicate added in commits 0eb3d0e9 / 1860c81 / bfa2364 was wired into the three `MergeTree` write paths but missed a fourth, very recently added one: the object-storage QCC write in `StorageObjectStorageSource::generate` (introduced by PR ClickHouse#102115 -- "Query condition cache for iceberg tables", merged 2026-05-07). That site populates the cache from `Iceberg`/`S3`/`Azure` queries when the format reader produces matched-bucket information, keyed by `FormatFilterInfo::condition_hash` -- which until now was created based only on `use_query_condition_cache`, with no writability check. So a query against an object-storage table running with `allow_suspicious_low_cardinality_types = 1` or any non-`Production` `allow_experimental_*` setting could still poison the cache for a subsequent default-settings query, leaving issue ClickHouse#104203 only partially closed for the engine as a whole. Mirror the existing `MergeTreeReaderSettings::query_condition_cache_writable` flag on `FormatFilterInfo`. Set it from the same predicate (`Settings::isQueryConditionCacheWritable`) at the same site where `condition_hash` is computed. Gate the `query_condition_cache->write` call in `StorageObjectStorageSource::generate` on the new flag. Reads through `condition_hash` stay open -- the lookup-side path remains valid because entries written from "writable" contexts are correct, and the same context that wrote nothing can still benefit from cache hits produced by ordinary queries. Coverage: extended `tests/integration/test_storage_iceberg_with_spark/test_query_condition_cache.py` with `test_query_condition_cache_not_writable_with_relaxed_settings`, parametrized across (`s3`, `azure`, `local`) and three triggering settings (`allow_suspicious_low_cardinality_types`, `allow_experimental_funnel_functions` [`Experimental` tier], `allow_experimental_database_iceberg` [`Beta` tier]). The test asserts that `system.query_condition_cache` stays empty after a filtered query in any of those contexts. Stateless test `04208_query_condition_cache_writable_104203` continues to exercise the `MergeTree` write paths and is unaffected.

PedroTadim · 2026-05-11T11:20:42Z

…solution in next commit) --- Original cherry-pick message follows: Merge pull request ClickHouse#102115 from scanhex12/iceberg_qcc Query condition cache for iceberg tables # Conflicts: # src/Databases/DataLake/DatabaseDataLake.cpp # src/Processors/Formats/Impl/ParquetV3BlockInputFormat.cpp # src/Processors/Formats/Impl/ParquetV3BlockInputFormat.h # src/Storages/ObjectStorage/DataLakes/Iceberg/ManifestFileIterator.cpp # src/Storages/ObjectStorage/StorageObjectStorageSource.cpp

scanhex12 marked this pull request as draft April 8, 2026 17:43

scanhex12 changed the title ~~Query condition cache for iceberg tables~~ [WIP] Query condition cache for iceberg tables Apr 8, 2026

clickhouse-gh Bot added the pr-feature Pull request with new product feature label Apr 8, 2026

clickhouse-gh Bot reviewed Apr 8, 2026

View reviewed changes

nikitamikhaylov mentioned this pull request Apr 9, 2026

Implement Query Condition Cache for Iceberg #89104

Closed

clickhouse-gh Bot reviewed Apr 9, 2026

View reviewed changes

scanhex12 changed the title ~~[WIP] Query condition cache for iceberg tables~~ Query condition cache for iceberg tables Apr 9, 2026

scanhex12 marked this pull request as ready for review April 9, 2026 14:05

clickhouse-gh Bot reviewed Apr 9, 2026

View reviewed changes

clickhouse-gh Bot reviewed Apr 10, 2026

View reviewed changes

Comment thread src/Storages/ObjectStorage/StorageObjectStorageSource.h

scanhex12 force-pushed the iceberg_qcc branch from e55b962 to ef2726a Compare April 11, 2026 20:32

clickhouse-gh Bot reviewed Apr 13, 2026

View reviewed changes

Comment thread src/Storages/ObjectStorage/StorageObjectStorageSource.cpp Outdated

scanhex12 force-pushed the iceberg_qcc branch from 0059440 to 96fdfe7 Compare April 13, 2026 13:21

clickhouse-gh Bot reviewed Apr 13, 2026

View reviewed changes

Comment thread src/Storages/ObjectStorage/StorageObjectStorageSource.cpp Outdated

scanhex12 force-pushed the iceberg_qcc branch from 9c290d7 to eeef06f Compare April 13, 2026 13:34

clickhouse-gh Bot reviewed Apr 13, 2026

View reviewed changes

CheSema self-assigned this Apr 22, 2026

grantholly-clickhouse reviewed Apr 22, 2026

View reviewed changes

Comment thread tests/integration/test_storage_iceberg_with_spark/test_query_condition_cache.py

clickhouse-gh Bot reviewed Apr 23, 2026

View reviewed changes

grantholly-clickhouse approved these changes May 6, 2026

View reviewed changes

grantholly-clickhouse added this pull request to the merge queue May 6, 2026

scanhex12 removed this pull request from the merge queue due to a manual request May 6, 2026

scanhex12 added this pull request to the merge queue May 7, 2026

Merged via the queue into ClickHouse:master with commit 69167ef May 7, 2026
323 of 326 checks passed

scanhex12 deleted the iceberg_qcc branch May 7, 2026 02:39

robot-clickhouse added the pr-synced-to-cloud The PR is synced to the cloud repo label May 7, 2026

clickgapai mentioned this pull request May 7, 2026

Cluster mode with cluster_table_function_split_granularity=BUCKET poisons the query condition cache by writing 'no match' for row groups assigned to other workers #104288

Open

groeneai mentioned this pull request May 7, 2026

Refuse QueryConditionCache writes from contexts with relaxed/experimental settings #104225

Closed

1 task

Diskein mentioned this pull request May 12, 2026

Performance regression in Iceberg manifest reading: exception hot-loop, eager log serialization, and std::filesystem::path in resolve() #104169

Open

zvonand mentioned this pull request May 15, 2026

Antalya 26.3: Query condition cache for iceberg tables Altinity/ClickHouse#1804

Merged

1 task

groeneai mentioned this pull request May 19, 2026

Fix crash when reading Iceberg tables with ORC files and PREWHERE enabled #101206

Merged

1 task

il9ue pushed a commit to il9ue/ClickHouse that referenced this pull request May 20, 2026

Resolve conflicts in cherry-pick of ClickHouse#102115

9966fee

groeneai mentioned this pull request May 21, 2026

Fix flaky test_insert_select_from_cluster_with_partition_pruning #105525

Open

1 task

subkanthi pushed a commit to Altinity/ClickHouse that referenced this pull request May 26, 2026

Resolve conflicts in cherry-pick of ClickHouse#102115

c783d78

zvonand mentioned this pull request Jun 25, 2026

Antalya 26.5: Query condition cache for iceberg tables Altinity/ClickHouse#1964

Open

1 task

clickgapai mentioned this pull request Jun 25, 2026

Logical Error: Source column is not Map #107024

Closed

This was referenced Jun 29, 2026

Antalya 26.5: Antalya 26.3 port - Fix view over iceberg (+ cluster-requests prereq #1687) Altinity/ClickHouse#1955

Open

Antalya 26.5: 26.1 Antalya port - improvements for cluster requests Altinity/ClickHouse#1953

Open

This was referenced Jul 3, 2026

Support Query Condition Cache for local Parquet files #109247

Open

Query Condition Cache: key object-storage entries by ETag #109310

Open



		@pytest.mark.parametrize("storage_type", ["s3", "azure", "local"])
		def test_query_condition_cache(started_cluster_iceberg_with_spark, storage_type):

Sunbelt Computer Software

PL/B Language Development and Support

Uh oh!

Conversation

scanhex12 commented Apr 8, 2026 • edited by robot-clickhouse Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changelog category (leave one):

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

Documentation entry for user-facing changes

Version info

Uh oh!

scanhex12 commented Apr 8, 2026

Uh oh!

clickhouse-gh Bot commented Apr 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

AI Review

Summary

PR Metadata

Findings

Tests

ClickHouse Rules

Final Verdict

Uh oh!

clickhouse-gh Bot Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

clickhouse-gh Bot Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

clickhouse-gh Bot Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

scanhex12 commented Apr 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clickhouse-gh Bot Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

scanhex12 commented Apr 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clickhouse-gh Bot Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

clickhouse-gh Bot Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

grantholly-clickhouse Apr 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alexey-milovidov commented Apr 9, 2026

Uh oh!

alexey-milovidov commented Apr 9, 2026

Uh oh!

alexey-milovidov commented Apr 10, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

clickhouse-gh Bot Apr 13, 2026

Choose a reason for hiding this comment

Uh oh!

scanhex12 commented Apr 22, 2026

Uh oh!

Uh oh!

clickhouse-gh Bot Apr 23, 2026

Choose a reason for hiding this comment

Uh oh!

clickhouse-gh Bot commented May 6, 2026

LLVM Coverage Report

Uh oh!

Uh oh!

Uh oh!

clickgapai commented May 7, 2026

Uh oh!

clickgapai commented May 7, 2026

Uh oh!

clickgapai commented May 7, 2026

Uh oh!

scanhex12 commented Apr 8, 2026 •

edited by robot-clickhouse

Loading

clickhouse-gh Bot commented Apr 8, 2026 •

edited

Loading

scanhex12 commented Apr 9, 2026 •

edited

Loading

scanhex12 commented Apr 9, 2026 •

edited

Loading

grantholly-clickhouse Apr 26, 2026 •

edited

Loading