GH-120: optimise created parquet files#138
Merged
Merged
Conversation
…files' into GH-120/optimise-created-parquet-files # Conflicts: # crates/icegate-ingest/src/shift/plan_runner.rs # crates/icegate-ingest/src/shift/planner.rs # crates/icegate-ingest/src/shift/planner_partitioning.rs
# Conflicts: # crates/icegate-ingest/src/cli/commands/run.rs # crates/icegate-ingest/src/shift/iceberg_storage.rs # crates/icegate-ingest/src/shift/mod.rs
WalkthroughReplaces grouped segment-based planning with a row-group–level planner and ExtractField extraction API; adds planner partitioning, clustering/bin-packing, byte-based Iceberg rollover sizing, WAL boundary metadata/name tracking, oversized-cluster metrics, and wide test and wiring updates across queue, shift, WAL, storage, helm, and dashboard configs. ChangesShift planning, queue extraction, WAL metadata, and wiring
Sequence Diagram(s)sequenceDiagram
participant Q as QueueReader (plan_segments)
participant E as Extractor (ResolvedField)
participant P as Planner (plan_row_groups)
participant S as Scheduler (schedule_shift_tasks)
participant X as Executor (ShiftExecutor / IcebergStorage)
Q->>E: resolve extractors per segment
E-->>Q: per-row-group ExtractedValue map
Q-->>P: SegmentsPlan.entries (RowGroupPlanEntry...)
P->>P: parse boundary payloads & derive partition buckets
P->>P: stable-sort, swept-line clustering, bin-pack → chunks
P-->>S: PlannedChunks + PlannerStats
S->>S: group row_groups by wal_offset → SegmentToRead
S-->>X: submit ShiftInput(segments)
X->>X: read segments, merge, write Parquet (rollover by bytes)
X-->>S: write acknowledgements / file artifacts
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related issues
Possibly related PRs
Suggested reviewers
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (3)
crates/icegate-ingest/src/shift/config.rs (1)
164-188:⚠️ Potential issue | 🟠 Major | ⚡ Quick winDon't silently drop the legacy
max_input_bytes_per_tasksetting.Because
ShiftReadConfigis still deserialized with defaults, older configs can keep parsing while theirmax_input_bytes_per_taskoverride is ignored and replaced by the new 64/128 MiB defaults. That turns an upgrade into a silent runtime behavior change for chunk sizing. Please either map the legacy key into the new bounds or reject it explicitly with a clear config error.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/icegate-ingest/src/shift/config.rs` around lines 164 - 188, ShiftReadConfig currently ignores a legacy max_input_bytes_per_task key when deserializing, causing silent behavior changes; update deserialization or Default handling so legacy overrides are honored or rejected: detect the presence of max_input_bytes_per_task during config load and either (1) map it to set both lower_bound_input_bytes_per_task and upper_bound_input_bytes_per_task (or compute appropriate bounds) or (2) return a clear error refusing the legacy key; implement this mapping/rejection in the config deserialization path that constructs ShiftReadConfig (or in Default::default() wrapper used during deserialization) to ensure max_input_bytes_per_task is not silently dropped.crates/icegate-ingest/src/wal/sorter.rs (1)
47-56:⚠️ Potential issue | 🟠 Major | 🏗️ Heavy liftReplace the hard-coded partition column here.
The planner path is now driven by
PlannerPartitionSpec, but WAL sorting still assumes"tenant_id"at Line 50. If the partition spec changes, these two paths will drift and the sorter will start producing row groups that are not homogeneous for the configured partition key. Please thread the shared schema/partition-spec source of truth into this path instead of using a string literal.As per coding guidelines,
**/*.rs: Only use Iceberg schema from crates/icegate-common/src/schema.rs, never hardcode columns in working code.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/icegate-ingest/src/wal/sorter.rs` around lines 47 - 56, The sorter currently hardcodes "tenant_id" when locating the partition column (see usage around tenant_idx, batch.schema(), and tenant_id_column); replace that literal with the partition key from the shared PlannerPartitionSpec / Iceberg schema source of truth (crates/icegate-common/src/schema.rs) by threading the PlannerPartitionSpec (or resolved partition field name) into the sorter (e.g., add it to the function signature or struct) and then call batch.schema().index_of(partition_field_name) and downcast that column instead of using "tenant_id"; ensure you use the canonical Iceberg schema helpers from icegate-common rather than a string literal so the WAL sorter stays consistent with planner partitioning.crates/icegate-ingest/src/shift/executor.rs (1)
60-77:⚠️ Potential issue | 🟠 Major | 🏗️ Heavy liftKeep the shift-task payload backward compatible across deploys.
ShiftInputis part of the persisted task contract, and the shifter restores jobs from durable storage. Movingrow_groupsunderSegmentToReadmeans any queued pre-rollout shift task will deserialize into a missingsegmentsfield and fail after deployment. Please accept both payload shapes during rollout, or make draining existing jobs a hard deployment prerequisite.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/icegate-ingest/src/shift/executor.rs` around lines 60 - 77, ShiftInput is part of a persisted task contract and must accept both the new shape (segments: Vec<SegmentToRead>) and the old shape (row_groups: Vec<PlannedRowGroup> at the top level) to avoid breaking restores; implement a backward-compatible deserializer for ShiftInput (e.g., a custom Deserialize impl or a serde untagged enum) that: if "segments" exists, uses it; otherwise, if top-level "row_groups" exists, wraps that Vec<PlannedRowGroup> into a single SegmentToRead (filling any missing fields like segment_offset with a safe default) and preserves trace_context, then populate ShiftInput.segments accordingly so both payload shapes deserialize successfully.
🧹 Nitpick comments (1)
crates/icegate-ingest/src/shift/plan_runner.rs (1)
154-188: ⚡ Quick winStabilize
ShiftInputsegment ordering before serialization.
by_segment.into_iter()makes identical plans produce differentShiftInputpayloads across runs. Sorting bysegment_offsetkeeps task payloads and logs deterministic without changing behavior.♻️ Suggested change
- let segments: Vec<SegmentToRead> = by_segment - .into_iter() - .map(|(segment_offset, row_groups)| SegmentToRead { - segment_offset, - row_groups, - }) - .collect(); + let mut segments: Vec<SegmentToRead> = by_segment + .into_iter() + .map(|(segment_offset, mut row_groups)| { + row_groups.sort_by_key(|row_group| row_group.row_group_idx); + SegmentToRead { + segment_offset, + row_groups, + } + }) + .collect(); + segments.sort_by_key(|segment| segment.segment_offset);🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@crates/icegate-ingest/src/shift/plan_runner.rs` around lines 154 - 188, The code builds a HashMap by_segment and then collects segments via by_segment.into_iter(), which yields non-deterministic ordering; in schedule_shift_task_for_chunk you should stabilize the order by sorting the produced Vec<SegmentToRead> by segment_offset (e.g., collect into segments and then call sort_by_key(|s| s.segment_offset)) before calling create_shift_task so ShiftInput serialization and logs are deterministic; target symbols: schedule_shift_task_for_chunk, by_segment, SegmentToRead, segment_offset, create_shift_task.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@config/docker/grafana/dashboards/Icegate-Ingest-1769530183590.json`:
- Around line 349-371: The panels use queries that group by protocol, signal,
status (and sometimes encoding) but their legendFormat only shows p50/p95/p99 or
incomplete labels, causing indistinguishable series; update each offending query
(e.g., refIds like
icegate_ingest_shift_plan_duration_seconds_bucket-p99-histogram_quantile, -p90-,
-p50- and the other refId blocks at the ranges called out) to either remove the
extra labels from the sum by(...) clause (e.g., drop
protocol/signal/status/encoding so aggregation matches simple legend) or include
those labels in legendFormat (e.g., "{{protocol}} - {{signal}} - {{status}} -
{{encoding}} - p95") so the legend uniquely identifies each series; apply the
same change consistently for the other listed panels (lines ~462-484, ~576-598,
~804-826, ~1032-1055, ~1866-1889).
In `@crates/icegate-ingest/src/shift/shift_runner.rs`:
- Around line 1292-1303: The current computation uses ts_col.value(i).into()
which treats null slots as valid values; change the logic to only consider
non-null values (use ts_col.is_valid(i) / !ts_col.is_null(i) and collect
ts_col.value(i) for those indices), compute min/max from that iterator, and only
call extracted.insert(..., TimestampMicrosRange(min, max)) when you actually
have at least one valid timestamp; otherwise do not insert the
PLAN_FIELD_TIMESTAMP_RANGE entry. Target symbols: ts_col
(TimestampMicrosecondArray), the min_ts/max_ts computation, and the
extracted.insert(...) using PLAN_FIELD_TIMESTAMP_RANGE.
In `@crates/icegate-queue/src/extract.rs`:
- Around line 53-61: Reject duplicate ExtractField names up front by validating
ExtractField::name before populating the resulting RowGroupPlanEntry::extracted
map: iterate the incoming Vec<ExtractField>, track names in a HashSet, and if
insert returns false return an error (or panic if codebase convention uses
unwraps) describing the duplicate name; ensure the validation occurs where the
Vec<ExtractField> is converted into the extracted map so no duplicate keys can
collapse silently.
- Around line 240-246: The code currently treats any empty extracted value as a
metadata error (the value.is_empty() branch that returns
Err(QueueError::Metadata(...))) which incorrectly rejects valid singleton empty
values (min == max == b""). Remove that empty-check branch (or stop returning
QueueError::Metadata for empty strings) so that the function returns
Ok(value.to_string()) for empty strings as well; reference the value.is_empty()
check and the Err(QueueError::Metadata(...)) that mentions column_name,
row_group_idx, and wal_offset to locate and modify the logic in extract.rs.
---
Outside diff comments:
In `@crates/icegate-ingest/src/shift/config.rs`:
- Around line 164-188: ShiftReadConfig currently ignores a legacy
max_input_bytes_per_task key when deserializing, causing silent behavior
changes; update deserialization or Default handling so legacy overrides are
honored or rejected: detect the presence of max_input_bytes_per_task during
config load and either (1) map it to set both lower_bound_input_bytes_per_task
and upper_bound_input_bytes_per_task (or compute appropriate bounds) or (2)
return a clear error refusing the legacy key; implement this mapping/rejection
in the config deserialization path that constructs ShiftReadConfig (or in
Default::default() wrapper used during deserialization) to ensure
max_input_bytes_per_task is not silently dropped.
In `@crates/icegate-ingest/src/shift/executor.rs`:
- Around line 60-77: ShiftInput is part of a persisted task contract and must
accept both the new shape (segments: Vec<SegmentToRead>) and the old shape
(row_groups: Vec<PlannedRowGroup> at the top level) to avoid breaking restores;
implement a backward-compatible deserializer for ShiftInput (e.g., a custom
Deserialize impl or a serde untagged enum) that: if "segments" exists, uses it;
otherwise, if top-level "row_groups" exists, wraps that Vec<PlannedRowGroup>
into a single SegmentToRead (filling any missing fields like segment_offset with
a safe default) and preserves trace_context, then populate ShiftInput.segments
accordingly so both payload shapes deserialize successfully.
In `@crates/icegate-ingest/src/wal/sorter.rs`:
- Around line 47-56: The sorter currently hardcodes "tenant_id" when locating
the partition column (see usage around tenant_idx, batch.schema(), and
tenant_id_column); replace that literal with the partition key from the shared
PlannerPartitionSpec / Iceberg schema source of truth
(crates/icegate-common/src/schema.rs) by threading the PlannerPartitionSpec (or
resolved partition field name) into the sorter (e.g., add it to the function
signature or struct) and then call batch.schema().index_of(partition_field_name)
and downcast that column instead of using "tenant_id"; ensure you use the
canonical Iceberg schema helpers from icegate-common rather than a string
literal so the WAL sorter stays consistent with planner partitioning.
---
Nitpick comments:
In `@crates/icegate-ingest/src/shift/plan_runner.rs`:
- Around line 154-188: The code builds a HashMap by_segment and then collects
segments via by_segment.into_iter(), which yields non-deterministic ordering; in
schedule_shift_task_for_chunk you should stabilize the order by sorting the
produced Vec<SegmentToRead> by segment_offset (e.g., collect into segments and
then call sort_by_key(|s| s.segment_offset)) before calling create_shift_task so
ShiftInput serialization and logs are deterministic; target symbols:
schedule_shift_task_for_chunk, by_segment, SegmentToRead, segment_offset,
create_shift_task.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: f4e0e0f8-fbb0-4328-ae32-d94986ac79ff
⛔ Files ignored due to path filters (1)
Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (32)
AGENTS.mdconfig/docker/grafana/dashboards/Icegate-Ingest-1769530183590.jsonconfig/docker/grafana/dashboards/Icegate-Query-1773746395045923.jsonconfig/helm/icegate/templates/configmap-ingest.yamlconfig/helm/icegate/values.yamlcrates/icegate-common/src/catalog/config.rscrates/icegate-ingest/src/cli/commands/run.rscrates/icegate-ingest/src/infra/metrics.rscrates/icegate-ingest/src/shift/config.rscrates/icegate-ingest/src/shift/executor.rscrates/icegate-ingest/src/shift/iceberg_storage.rscrates/icegate-ingest/src/shift/instrumentation.rscrates/icegate-ingest/src/shift/mod.rscrates/icegate-ingest/src/shift/plan_runner.rscrates/icegate-ingest/src/shift/planner.rscrates/icegate-ingest/src/shift/planner_partitioning.rscrates/icegate-ingest/src/shift/row_groups_merger.rscrates/icegate-ingest/src/shift/shift_runner.rscrates/icegate-ingest/src/wal/boundary.rscrates/icegate-ingest/src/wal/columns.rscrates/icegate-ingest/src/wal/metadata.rscrates/icegate-ingest/src/wal/mod.rscrates/icegate-ingest/src/wal/sorter.rscrates/icegate-ingest/src/wal/test_utils.rscrates/icegate-queue/src/config.rscrates/icegate-queue/src/extract.rscrates/icegate-queue/src/lib.rscrates/icegate-queue/src/reader.rscrates/icegate-queue/src/writer.rscrates/icegate-queue/tests/common/mod.rscrates/icegate-queue/tests/reader_integration.rscrates/icegate-queue/tests/writer_integration.rs
💤 Files with no reviewable changes (1)
- crates/icegate-queue/src/config.rs
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
config/docker/grafana/dashboards/Icegate-Ingest-1769530183590.json (1)
1563-1582:⚠️ Potential issue | 🟠 Major | ⚡ Quick winMetric name mismatch will cause this panel to show no data.
The queries reference
icegate_ingest_wal_segment_bytes_bucket, but the metric defined incrates/icegate-ingest/src/infra/metrics.rsis namedicegate_ingest_wal_segment(which becomesicegate_ingest_wal_segment_bucketin Prometheus):segment_bytes: meter.f64_histogram("icegate_ingest_wal_segment").build(),Fix: Remove `_bytes` from the metric names in all three queries
- "expr": "histogram_quantile(0.95, sum by (le, topic) (rate(icegate_ingest_wal_segment_bytes_bucket{job=\"ingest\"}[1m])))\n", + "expr": "histogram_quantile(0.95, sum by (le, topic) (rate(icegate_ingest_wal_segment_bucket{job=\"ingest\"}[1m])))\n",- "expr": "histogram_quantile(0.5, sum by (le, topic) (rate(icegate_ingest_wal_segment_bytes_bucket{job=\"ingest\"}[1m])))", + "expr": "histogram_quantile(0.5, sum by (le, topic) (rate(icegate_ingest_wal_segment_bucket{job=\"ingest\"}[1m])))",- "expr": "histogram_quantile(0.99, sum by (le, topic) (rate(icegate_ingest_wal_segment_bytes_bucket{job=\"ingest\"}[1m])))", + "expr": "histogram_quantile(0.99, sum by (le, topic) (rate(icegate_ingest_wal_segment_bucket{job=\"ingest\"}[1m])))",🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@config/docker/grafana/dashboards/Icegate-Ingest-1769530183590.json` around lines 1563 - 1582, The panel queries reference the wrong metric name suffix: update the three Prometheus expressions that use icegate_ingest_wal_segment_bytes_bucket to use icegate_ingest_wal_segment_bucket (the histogram created by segment_bytes in metrics.rs), i.e. replace occurrences of icegate_ingest_wal_segment_bytes_bucket with icegate_ingest_wal_segment_bucket in the expressions (those blocks with refId values like icegate_ingest_shift_plan_duration_seconds_bucket-p99-histogram_quantile and A).
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Outside diff comments:
In `@config/docker/grafana/dashboards/Icegate-Ingest-1769530183590.json`:
- Around line 1563-1582: The panel queries reference the wrong metric name
suffix: update the three Prometheus expressions that use
icegate_ingest_wal_segment_bytes_bucket to use icegate_ingest_wal_segment_bucket
(the histogram created by segment_bytes in metrics.rs), i.e. replace occurrences
of icegate_ingest_wal_segment_bytes_bucket with
icegate_ingest_wal_segment_bucket in the expressions (those blocks with refId
values like
icegate_ingest_shift_plan_duration_seconds_bucket-p99-histogram_quantile and A).
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 2ae27301-7956-4277-8cfd-9579261a21e6
📒 Files selected for processing (4)
config/docker/grafana/dashboards/Icegate-Ingest-1769530183590.jsoncrates/icegate-ingest/src/shift/shift_runner.rscrates/icegate-queue/src/extract.rscrates/icegate-queue/src/reader.rs
🚧 Files skipped from review as they are similar to previous changes (1)
- crates/icegate-queue/src/extract.rs
frisbeeman
reviewed
May 4, 2026
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
crates/icegate-ingest/src/shift/config.rs (1)
164-181:⚠️ Potential issue | 🟠 Major | ⚡ Quick winFix Helm/Rust config key and unit mismatch for input bounds.
The Helm template (
configmap-ingest.yaml) defineslower_bound_input_bytes_per_taskandupper_bound_input_bytes_per_task, but the Rust struct exposeslower_bound_input_mb_per_taskandupper_bound_input_mb_per_task. With#[serde(default)]onShiftReadConfig, mismatched keys will silently use defaults (64 MB and 128 MB) instead of the operator-supplied values, with no deserialization error. Additionally, the Helm keys imply byte units while the Rust struct fields are in megabytes—even if keys were renamed, the values would be off by approximately 1000× without unit conversion. This silently breaks production deployment configurations.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@crates/icegate-ingest/src/shift/config.rs` around lines 164 - 181, ShiftReadConfig’s fields lower_bound_input_mb_per_task and upper_bound_input_mb_per_task don’t match the Helm keys lower_bound_input_bytes_per_task/upper_bound_input_bytes_per_task and so operator values are ignored; to fix, update the struct to accept the bytes keys and convert units: either add serde attributes to those fields (e.g. #[serde(alias = "lower_bound_input_bytes_per_task", deserialize_with = "...")] and the corresponding alias for upper) and implement a small deserialize_with function that accepts bytes (u64) and converts to MB (divide by 1_048_576) or change the field types to bytes (u64) and convert to MB where consumed; ensure the serde default behavior and any default values remain correct and reference ShiftReadConfig, lower_bound_input_mb_per_task, upper_bound_input_mb_per_task, and the Helm keys in configmap-ingest.yaml when making the change.
🧹 Nitpick comments (1)
crates/icegate-ingest/src/shift/mod.rs (1)
113-127: 💤 Low valueConsider a helper on
ShiftReadConfigfor MB→bytes conversion.
upper_bound_input_mb_per_task * 1024 * 1024is computed here, inplan_runner.rs(for both lower and upper bounds), and in tests. A single inherent method likeShiftReadConfig::upper_bound_input_bytes(&self) -> u64(and a sibling for the lower bound) would centralize the unit conversion, make call sites read more clearly at the storage/planner boundary, and give one place to add saturating arithmetic should the per-task budget ever grow towardu64limits.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@crates/icegate-ingest/src/shift/mod.rs` around lines 113 - 127, Add inherent byte-conversion helpers on ShiftReadConfig (e.g., upper_bound_input_bytes(&self) -> u64 and lower_bound_input_bytes(&self) -> u64) and replace direct multiplications of upper_bound_input_mb_per_task * 1024 * 1024 (and the lower-bound equivalents) with calls to these methods in shift/mod.rs (where writer_max_parquet_bytes is called), in plan_runner.rs, and in tests; implement the helpers using saturating multiplication (saturating_mul or checked_mul with fallback) to avoid overflow if the MB value grows.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@crates/icegate-ingest/src/shift/config.rs`:
- Around line 164-181: ShiftReadConfig’s fields lower_bound_input_mb_per_task
and upper_bound_input_mb_per_task don’t match the Helm keys
lower_bound_input_bytes_per_task/upper_bound_input_bytes_per_task and so
operator values are ignored; to fix, update the struct to accept the bytes keys
and convert units: either add serde attributes to those fields (e.g.
#[serde(alias = "lower_bound_input_bytes_per_task", deserialize_with = "...")]
and the corresponding alias for upper) and implement a small deserialize_with
function that accepts bytes (u64) and converts to MB (divide by 1_048_576) or
change the field types to bytes (u64) and convert to MB where consumed; ensure
the serde default behavior and any default values remain correct and reference
ShiftReadConfig, lower_bound_input_mb_per_task, upper_bound_input_mb_per_task,
and the Helm keys in configmap-ingest.yaml when making the change.
---
Nitpick comments:
In `@crates/icegate-ingest/src/shift/mod.rs`:
- Around line 113-127: Add inherent byte-conversion helpers on ShiftReadConfig
(e.g., upper_bound_input_bytes(&self) -> u64 and lower_bound_input_bytes(&self)
-> u64) and replace direct multiplications of upper_bound_input_mb_per_task *
1024 * 1024 (and the lower-bound equivalents) with calls to these methods in
shift/mod.rs (where writer_max_parquet_bytes is called), in plan_runner.rs, and
in tests; implement the helpers using saturating multiplication (saturating_mul
or checked_mul with fallback) to avoid overflow if the MB value grows.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: b9d907dc-9959-4750-ae32-773b8fb41f24
📒 Files selected for processing (5)
crates/icegate-ingest/src/shift/config.rscrates/icegate-ingest/src/shift/iceberg_storage.rscrates/icegate-ingest/src/shift/mod.rscrates/icegate-ingest/src/shift/plan_runner.rscrates/icegate-ingest/src/shift/shift_runner.rs
frisbeeman
approved these changes
May 6, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary by CodeRabbit
New Features
Improvements
Configuration Changes
Documentation