risingwave_hummock_sdk/

compact_task.rs

// Copyright 2024 RisingWave Labs
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

use std::collections::{BTreeMap, HashMap};
use std::mem::size_of;

use itertools::Itertools;
use risingwave_common::catalog::TableId;
use risingwave_pb::hummock::compact_task::{PbTaskStatus, PbTaskType, TaskStatus, TaskType};
use risingwave_pb::hummock::subscribe_compaction_event_request::PbReportTask;
use risingwave_pb::hummock::{
    LevelType, PbCompactTask, PbKeyRange, PbTableOption, PbTableSchema, PbTableStats,
    PbValidationTask,
};
use risingwave_pb::id::WorkerId;

use crate::compaction_group::StateTableId;
use crate::key_range::KeyRange;
use crate::level::InputLevel;
use crate::sstable_info::SstableInfo;
use crate::table_watermark::{TableWatermarks, WatermarkSerdeType};
use crate::{CompactionGroupId, HummockSstableObjectId};

#[derive(Clone, PartialEq, Default, Debug)]
pub struct CompactTask {
    /// SSTs selected by meta, which will be removed from LSM after compaction succeeds.
    ///
    /// Each SST's `table_ids` comes from version metadata, where truncate/drop table deltas prune
    /// table ids before later compaction tasks are picked.
    pub input_ssts: Vec<InputLevel>,
    /// In ideal case, the compaction will generate `splits.len()` tables which have key range
    /// corresponding to that in `splits`, respectively
    pub splits: Vec<KeyRange>,
    /// compaction output, which will be added to `target_level` of LSM after compaction
    pub sorted_output_ssts: Vec<SstableInfo>,
    /// task id assigned by hummock storage service
    pub task_id: u64,
    /// compaction output will be added to `target_level` of LSM after compaction
    pub target_level: u32,
    pub gc_delete_keys: bool,
    /// Lbase in LSM
    pub base_level: u32,
    pub task_status: PbTaskStatus,
    /// compaction group the task belongs to.
    pub compaction_group_id: CompactionGroupId,
    /// compaction group id when the compaction task is created
    pub compaction_group_version_id: u64,
    /// Table ids that still exist in the version when the compaction task is picked.
    ///
    /// This may be broader than the table ids touched by this task's input SSTs. Task-local table
    /// ids should be read from `input_ssts[*].table_ids`.
    pub existing_table_ids: Vec<TableId>,
    pub compression_algorithm: u32,
    pub target_file_size: u64,
    pub compaction_filter_mask: u32,
    pub table_options: BTreeMap<TableId, PbTableOption>,
    pub current_epoch_time: u64,
    pub target_sub_level_id: u64,
    /// Identifies whether the task is `space_reclaim`, if the `compact_task_type` increases, it will be refactored to enum
    pub task_type: PbTaskType,
    /// Deprecated. use `table_vnode_partition` instead;
    pub split_by_state_table: bool,
    /// Compaction needs to cut the state table every time 1/weight of vnodes in the table have been processed.
    /// Deprecated. use `table_vnode_partition` instead;
    pub split_weight_by_vnode: u32,
    pub table_vnode_partition: BTreeMap<TableId, u32>,
    /// The table watermark of any table id. In compaction we only use the table watermarks on safe epoch,
    /// so we only need to include the table watermarks on safe epoch to reduce the size of metadata.
    pub pk_prefix_table_watermarks: BTreeMap<TableId, TableWatermarks>,

    pub non_pk_prefix_table_watermarks: BTreeMap<TableId, TableWatermarks>,
    pub value_table_watermarks: BTreeMap<TableId, TableWatermarks>,

    pub table_schemas: BTreeMap<TableId, PbTableSchema>,

    pub max_sub_compaction: u32,

    pub max_kv_count_for_xor16: Option<u64>,

    pub max_vnode_key_range_bytes: Option<u64>,
}

impl CompactTask {
    pub fn estimated_encode_len(&self) -> usize {
        self.input_ssts
            .iter()
            .map(|input_level| input_level.estimated_encode_len())
            .sum::<usize>()
            + self
                .splits
                .iter()
                .map(|split| split.left.len() + split.right.len() + size_of::<bool>())
                .sum::<usize>()
            + size_of::<u64>()
            + self
                .sorted_output_ssts
                .iter()
                .map(|sst| sst.estimated_encode_len())
                .sum::<usize>()
            + size_of::<u64>()
            + size_of::<u32>()
            + size_of::<bool>()
            + size_of::<u32>()
            + size_of::<i32>()
            + size_of::<u64>()
            + self.existing_table_ids.len() * size_of::<u32>()
            + size_of::<u32>()
            + size_of::<u64>()
            + size_of::<u32>()
            + self.table_options.len() * size_of::<u64>()
            + size_of::<u64>()
            + size_of::<u64>()
            + size_of::<i32>()
            + size_of::<bool>()
            + size_of::<u32>()
            + self.table_vnode_partition.len() * size_of::<u64>()
            + self
                .pk_prefix_table_watermarks
                .values()
                .map(|table_watermark| size_of::<u32>() + table_watermark.estimated_encode_len())
                .sum::<usize>()
            + self
                .non_pk_prefix_table_watermarks
                .values()
                .map(|table_watermark| size_of::<u32>() + table_watermark.estimated_encode_len())
                .sum::<usize>()
            + self
                .max_vnode_key_range_bytes
                .map(|_| size_of::<u32>())
                .unwrap_or_default()
            + self
                .value_table_watermarks
                .values()
                .map(|table_watermark| size_of::<u32>() + table_watermark.estimated_encode_len())
                .sum::<usize>()
    }

    pub fn is_trivial_move_task(&self) -> bool {
        if self.task_type != TaskType::Dynamic && self.task_type != TaskType::Emergency {
            return false;
        }

        if self.input_ssts.len() != 2 || self.input_ssts[0].level_type != LevelType::Nonoverlapping
        {
            return false;
        }

        // it may be a manual compaction task
        if self.input_ssts[0].level_idx == self.input_ssts[1].level_idx
            && self.input_ssts[0].level_idx > 0
        {
            return false;
        }

        if self.input_ssts[1].level_idx == self.target_level
            && self.input_ssts[1].table_infos.is_empty()
        {
            return true;
        }

        false
    }

    pub fn task_label(&self) -> &'static str {
        if self.is_trivial_reclaim() {
            return "trivial-space-reclaim";
        }

        if self.is_trivial_move_task() {
            return "trivial-move";
        }

        "normal"
    }

    pub fn is_trivial_reclaim(&self) -> bool {
        // Currently all VnodeWatermark tasks are trivial reclaim.
        if self.task_type == TaskType::VnodeWatermark {
            return true;
        }
        self.input_ssts
            .iter()
            .flat_map(|level| level.table_infos.iter())
            .all(|sst| sst.table_ids.is_empty())
    }
}

impl CompactTask {
    // The compact task may need to reclaim key with TTL
    pub fn contains_ttl(&self) -> bool {
        self.table_options
            .iter()
            .any(|(_, table_option)| table_option.retention_seconds.is_some_and(|ttl| ttl > 0))
    }

    // The compact task may need to reclaim key with range tombstone
    pub fn contains_range_tombstone(&self) -> bool {
        self.read_input_ssts()
            .any(|sst| sst.range_tombstone_count > 0)
    }

    // The compact task may need to reclaim key with split sst
    pub fn contains_split_sst(&self) -> bool {
        self.read_input_ssts()
            .any(|sst| sst.sst_id.as_raw_id() != sst.object_id.as_raw_id())
    }

    /// Returns sorted and deduplicated table ids from this task's normalized input SST metadata.
    pub fn get_table_ids_from_input_ssts(&self) -> impl Iterator<Item = StateTableId> + use<> {
        self.input_ssts
            .iter()
            .flat_map(|level| level.table_infos.iter())
            .flat_map(|sst| sst.table_ids.iter().copied())
            .sorted()
            .dedup()
    }

    pub fn is_expired(&self, compaction_group_version_id_expected: u64) -> bool {
        is_compaction_task_expired(
            self.compaction_group_version_id,
            compaction_group_version_id_expected,
        )
    }

    /// Determines whether to use block-based filter for this compaction task.
    /// Returns true if the total key count exceeds the configured threshold.
    pub fn should_use_block_based_filter(&self) -> bool {
        let kv_count = self
            .read_input_ssts()
            .map(|sst| sst.total_key_count)
            .sum::<u64>();

        crate::filter_utils::is_kv_count_too_large_for_xor16(kv_count, self.max_kv_count_for_xor16)
    }

    /// Returns the effective vnode key-range hint limit (in bytes) for this compaction task.
    ///
    /// The hint is only meaningful when all input SSTs belong to a single table.
    pub fn effective_max_vnode_key_range_bytes(&self) -> Option<usize> {
        let limit = self.max_vnode_key_range_bytes.filter(|&v| v > 0)? as usize;
        (self.get_table_ids_from_input_ssts().count() == 1).then_some(limit)
    }

    /// Returns input SSTs that should be read by the compactor.
    ///
    /// SST entries with empty `table_ids` are ignored defensively and should not be read by the
    /// compactor.
    pub fn read_input_ssts(&self) -> impl Iterator<Item = &SstableInfo> {
        self.input_ssts
            .iter()
            .flat_map(|level| level.read_sstable_infos())
    }
}

fn split_watermark_serde_types(
    pb_compact_task: &PbCompactTask,
) -> (
    BTreeMap<TableId, TableWatermarks>,
    BTreeMap<TableId, TableWatermarks>,
    BTreeMap<TableId, TableWatermarks>,
) {
    let mut pk_prefix_table_watermarks = BTreeMap::default();
    let mut non_pk_prefix_table_watermarks = BTreeMap::default();
    let mut value_table_watermarks = BTreeMap::default();
    for (table_id, pbwatermark) in &pb_compact_task.table_watermarks {
        let watermark = TableWatermarks::from(pbwatermark);
        match watermark.watermark_type {
            WatermarkSerdeType::PkPrefix => {
                pk_prefix_table_watermarks.insert(*table_id, watermark);
            }
            WatermarkSerdeType::NonPkPrefix => {
                non_pk_prefix_table_watermarks.insert(*table_id, watermark);
            }
            WatermarkSerdeType::Value => {
                value_table_watermarks.insert(*table_id, watermark);
            }
        }
    }
    (
        pk_prefix_table_watermarks,
        non_pk_prefix_table_watermarks,
        value_table_watermarks,
    )
}

pub fn is_compaction_task_expired(
    compaction_group_version_id_in_task: u64,
    compaction_group_version_id_expected: u64,
) -> bool {
    compaction_group_version_id_in_task != compaction_group_version_id_expected
}

impl From<PbCompactTask> for CompactTask {
    fn from(pb_compact_task: PbCompactTask) -> Self {
        let (pk_prefix_table_watermarks, non_pk_prefix_table_watermarks, value_table_watermarks) =
            split_watermark_serde_types(&pb_compact_task);
        #[expect(deprecated)]
        Self {
            input_ssts: pb_compact_task
                .input_ssts
                .into_iter()
                .map(InputLevel::from)
                .collect_vec(),
            splits: pb_compact_task
                .splits
                .into_iter()
                .map(|pb_keyrange| KeyRange {
                    left: pb_keyrange.left.into(),
                    right: pb_keyrange.right.into(),
                    right_exclusive: pb_keyrange.right_exclusive,
                })
                .collect_vec(),
            sorted_output_ssts: pb_compact_task
                .sorted_output_ssts
                .into_iter()
                .map(SstableInfo::from)
                .collect_vec(),
            task_id: pb_compact_task.task_id,
            target_level: pb_compact_task.target_level,
            gc_delete_keys: pb_compact_task.gc_delete_keys,
            base_level: pb_compact_task.base_level,
            task_status: TaskStatus::try_from(pb_compact_task.task_status).unwrap(),
            compaction_group_id: pb_compact_task.compaction_group_id,
            existing_table_ids: pb_compact_task.existing_table_ids.clone(),
            compression_algorithm: pb_compact_task.compression_algorithm,
            target_file_size: pb_compact_task.target_file_size,
            compaction_filter_mask: pb_compact_task.compaction_filter_mask,
            table_options: pb_compact_task
                .table_options
                .iter()
                .map(|(table_id, v)| (*table_id, *v))
                .collect(),
            current_epoch_time: pb_compact_task.current_epoch_time,
            target_sub_level_id: pb_compact_task.target_sub_level_id,
            task_type: PbTaskType::try_from(pb_compact_task.task_type).unwrap(),
            split_by_state_table: pb_compact_task.split_by_state_table,
            split_weight_by_vnode: pb_compact_task.split_weight_by_vnode,
            table_vnode_partition: pb_compact_task
                .table_vnode_partition
                .iter()
                .map(|(table_id, v)| (*table_id, *v))
                .collect(),
            pk_prefix_table_watermarks,
            non_pk_prefix_table_watermarks,
            value_table_watermarks,
            table_schemas: pb_compact_task
                .table_schemas
                .iter()
                .map(|(table_id, v)| (*table_id, v.clone()))
                .collect(),
            max_sub_compaction: pb_compact_task.max_sub_compaction,
            compaction_group_version_id: pb_compact_task.compaction_group_version_id,
            max_kv_count_for_xor16: pb_compact_task.max_kv_count_for_xor16,
            max_vnode_key_range_bytes: pb_compact_task.max_vnode_key_range_bytes,
        }
    }
}

impl From<&PbCompactTask> for CompactTask {
    fn from(pb_compact_task: &PbCompactTask) -> Self {
        let (pk_prefix_table_watermarks, non_pk_prefix_table_watermarks, value_table_watermarks) =
            split_watermark_serde_types(pb_compact_task);
        #[expect(deprecated)]
        Self {
            input_ssts: pb_compact_task
                .input_ssts
                .iter()
                .map(InputLevel::from)
                .collect_vec(),
            splits: pb_compact_task
                .splits
                .iter()
                .map(|pb_keyrange| KeyRange {
                    left: pb_keyrange.left.clone().into(),
                    right: pb_keyrange.right.clone().into(),
                    right_exclusive: pb_keyrange.right_exclusive,
                })
                .collect_vec(),
            sorted_output_ssts: pb_compact_task
                .sorted_output_ssts
                .iter()
                .map(SstableInfo::from)
                .collect_vec(),
            task_id: pb_compact_task.task_id,
            target_level: pb_compact_task.target_level,
            gc_delete_keys: pb_compact_task.gc_delete_keys,
            base_level: pb_compact_task.base_level,
            task_status: TaskStatus::try_from(pb_compact_task.task_status).unwrap(),
            compaction_group_id: pb_compact_task.compaction_group_id,
            existing_table_ids: pb_compact_task.existing_table_ids.clone(),
            compression_algorithm: pb_compact_task.compression_algorithm,
            target_file_size: pb_compact_task.target_file_size,
            compaction_filter_mask: pb_compact_task.compaction_filter_mask,
            table_options: pb_compact_task
                .table_options
                .iter()
                .map(|(table_id, v)| (*table_id, *v))
                .collect(),
            current_epoch_time: pb_compact_task.current_epoch_time,
            target_sub_level_id: pb_compact_task.target_sub_level_id,
            task_type: PbTaskType::try_from(pb_compact_task.task_type).unwrap(),
            split_by_state_table: pb_compact_task.split_by_state_table,
            split_weight_by_vnode: pb_compact_task.split_weight_by_vnode,
            table_vnode_partition: pb_compact_task
                .table_vnode_partition
                .iter()
                .map(|(table_id, v)| (*table_id, *v))
                .collect(),
            pk_prefix_table_watermarks,
            non_pk_prefix_table_watermarks,
            value_table_watermarks,
            table_schemas: pb_compact_task
                .table_schemas
                .iter()
                .map(|(table_id, v)| (*table_id, v.clone()))
                .collect(),
            max_sub_compaction: pb_compact_task.max_sub_compaction,
            compaction_group_version_id: pb_compact_task.compaction_group_version_id,
            max_kv_count_for_xor16: pb_compact_task.max_kv_count_for_xor16,
            max_vnode_key_range_bytes: pb_compact_task.max_vnode_key_range_bytes,
        }
    }
}

impl From<CompactTask> for PbCompactTask {
    fn from(compact_task: CompactTask) -> Self {
        #[expect(deprecated)]
        Self {
            input_ssts: compact_task
                .input_ssts
                .into_iter()
                .map(|input_level| input_level.into())
                .collect_vec(),
            splits: compact_task
                .splits
                .into_iter()
                .map(|keyrange| PbKeyRange {
                    left: keyrange.left.into(),
                    right: keyrange.right.into(),
                    right_exclusive: keyrange.right_exclusive,
                })
                .collect_vec(),
            sorted_output_ssts: compact_task
                .sorted_output_ssts
                .into_iter()
                .map(|sst| sst.into())
                .collect_vec(),
            task_id: compact_task.task_id,
            target_level: compact_task.target_level,
            gc_delete_keys: compact_task.gc_delete_keys,
            base_level: compact_task.base_level,
            task_status: compact_task.task_status.into(),
            compaction_group_id: compact_task.compaction_group_id,
            existing_table_ids: compact_task.existing_table_ids.clone(),
            compression_algorithm: compact_task.compression_algorithm,
            target_file_size: compact_task.target_file_size,
            compaction_filter_mask: compact_task.compaction_filter_mask,
            table_options: compact_task.table_options.clone(),
            current_epoch_time: compact_task.current_epoch_time,
            target_sub_level_id: compact_task.target_sub_level_id,
            task_type: compact_task.task_type.into(),
            split_weight_by_vnode: compact_task.split_weight_by_vnode,
            table_vnode_partition: compact_task.table_vnode_partition.clone(),
            table_watermarks: compact_task
                .pk_prefix_table_watermarks
                .into_iter()
                .chain(compact_task.non_pk_prefix_table_watermarks)
                .chain(compact_task.value_table_watermarks)
                .map(|(table_id, table_watermark)| (table_id, table_watermark.into()))
                .collect(),
            split_by_state_table: compact_task.split_by_state_table,
            table_schemas: compact_task.table_schemas.clone(),
            max_sub_compaction: compact_task.max_sub_compaction,
            compaction_group_version_id: compact_task.compaction_group_version_id,
            max_kv_count_for_xor16: compact_task.max_kv_count_for_xor16,
            max_vnode_key_range_bytes: compact_task.max_vnode_key_range_bytes,
        }
    }
}

impl From<&CompactTask> for PbCompactTask {
    fn from(compact_task: &CompactTask) -> Self {
        #[expect(deprecated)]
        Self {
            input_ssts: compact_task
                .input_ssts
                .iter()
                .map(|input_level| input_level.into())
                .collect_vec(),
            splits: compact_task
                .splits
                .iter()
                .map(|keyrange| PbKeyRange {
                    left: keyrange.left.to_vec(),
                    right: keyrange.right.to_vec(),
                    right_exclusive: keyrange.right_exclusive,
                })
                .collect_vec(),
            sorted_output_ssts: compact_task
                .sorted_output_ssts
                .iter()
                .map(|sst| sst.into())
                .collect_vec(),
            task_id: compact_task.task_id,
            target_level: compact_task.target_level,
            gc_delete_keys: compact_task.gc_delete_keys,
            base_level: compact_task.base_level,
            task_status: compact_task.task_status.into(),
            compaction_group_id: compact_task.compaction_group_id,
            existing_table_ids: compact_task.existing_table_ids.clone(),
            compression_algorithm: compact_task.compression_algorithm,
            target_file_size: compact_task.target_file_size,
            compaction_filter_mask: compact_task.compaction_filter_mask,
            table_options: compact_task.table_options.clone(),
            current_epoch_time: compact_task.current_epoch_time,
            target_sub_level_id: compact_task.target_sub_level_id,
            task_type: compact_task.task_type.into(),
            split_weight_by_vnode: compact_task.split_weight_by_vnode,
            table_vnode_partition: compact_task.table_vnode_partition.clone(),
            table_watermarks: compact_task
                .pk_prefix_table_watermarks
                .iter()
                .chain(compact_task.non_pk_prefix_table_watermarks.iter())
                .chain(compact_task.value_table_watermarks.iter())
                .map(|(table_id, table_watermark)| (*table_id, table_watermark.into()))
                .collect(),
            split_by_state_table: compact_task.split_by_state_table,
            table_schemas: compact_task.table_schemas.clone(),
            max_sub_compaction: compact_task.max_sub_compaction,
            compaction_group_version_id: compact_task.compaction_group_version_id,
            max_kv_count_for_xor16: compact_task.max_kv_count_for_xor16,
            max_vnode_key_range_bytes: compact_task.max_vnode_key_range_bytes,
        }
    }
}

#[derive(Clone, PartialEq, Default)]
pub struct ValidationTask {
    pub sst_infos: Vec<SstableInfo>,
    pub sst_id_to_worker_id: HashMap<HummockSstableObjectId, WorkerId>,
}

impl From<PbValidationTask> for ValidationTask {
    fn from(pb_validation_task: PbValidationTask) -> Self {
        Self {
            sst_infos: pb_validation_task
                .sst_infos
                .into_iter()
                .map(SstableInfo::from)
                .collect_vec(),
            sst_id_to_worker_id: pb_validation_task.sst_id_to_worker_id.into_iter().collect(),
        }
    }
}

impl From<ValidationTask> for PbValidationTask {
    fn from(validation_task: ValidationTask) -> Self {
        Self {
            sst_infos: validation_task
                .sst_infos
                .into_iter()
                .map(|sst| sst.into())
                .collect_vec(),
            sst_id_to_worker_id: validation_task.sst_id_to_worker_id,
        }
    }
}

impl ValidationTask {
    pub fn estimated_encode_len(&self) -> usize {
        self.sst_infos
            .iter()
            .map(|sst| sst.estimated_encode_len())
            .sum::<usize>()
            + self.sst_id_to_worker_id.len() * (size_of::<u64>() + size_of::<u32>())
            + size_of::<u64>()
    }
}

#[derive(Clone, PartialEq, Default, Debug)]
pub struct ReportTask {
    pub table_stats_change: HashMap<TableId, PbTableStats>,
    pub task_id: u64,
    pub task_status: TaskStatus,
    pub sorted_output_ssts: Vec<SstableInfo>,
    pub object_timestamps: HashMap<HummockSstableObjectId, u64>,
}

impl From<PbReportTask> for ReportTask {
    fn from(value: PbReportTask) -> Self {
        Self {
            table_stats_change: value.table_stats_change.clone(),
            task_id: value.task_id,
            task_status: PbTaskStatus::try_from(value.task_status).unwrap(),
            sorted_output_ssts: value
                .sorted_output_ssts
                .into_iter()
                .map(SstableInfo::from)
                .collect_vec(),
            object_timestamps: value.object_timestamps,
        }
    }
}

impl From<ReportTask> for PbReportTask {
    fn from(value: ReportTask) -> Self {
        Self {
            table_stats_change: value.table_stats_change.clone(),
            task_id: value.task_id,
            task_status: value.task_status.into(),
            sorted_output_ssts: value
                .sorted_output_ssts
                .into_iter()
                .map(|sst| sst.into())
                .collect_vec(),
            object_timestamps: value.object_timestamps,
        }
    }
}

#[cfg(test)]
mod tests {
    use risingwave_common::catalog::TableId;
    use risingwave_pb::hummock::PbLevelType;
    use risingwave_pb::hummock::compact_task::TaskType;

    use super::CompactTask;
    use crate::level::InputLevel;
    use crate::sstable_info::{SstableInfo, SstableInfoInner};

    fn test_sstable(sst_id: u64, table_ids: Vec<TableId>) -> SstableInfo {
        SstableInfo::from(SstableInfoInner {
            object_id: sst_id.into(),
            sst_id: sst_id.into(),
            table_ids,
            ..Default::default()
        })
    }

    fn test_read_property_sstable(table_ids: Vec<TableId>) -> SstableInfo {
        SstableInfo::from(SstableInfoInner {
            object_id: 1.into(),
            sst_id: 2.into(),
            table_ids,
            range_tombstone_count: 1,
            total_key_count: 100,
            ..Default::default()
        })
    }

    #[test]
    fn test_empty_table_ids_are_reclaim_and_have_no_input_table_ids() {
        let task = CompactTask {
            input_ssts: vec![InputLevel {
                table_infos: vec![test_sstable(1, vec![])],
                ..Default::default()
            }],
            existing_table_ids: vec![TableId::new(1)],
            ..Default::default()
        };

        assert!(task.get_table_ids_from_input_ssts().next().is_none());
        assert!(task.is_trivial_reclaim());
    }

    #[test]
    fn test_read_properties_ignore_empty_table_ids() {
        let task = CompactTask {
            input_ssts: vec![InputLevel {
                table_infos: vec![
                    test_read_property_sstable(vec![]),
                    SstableInfo::from(SstableInfoInner {
                        object_id: 3.into(),
                        sst_id: 3.into(),
                        table_ids: vec![TableId::new(1)],
                        total_key_count: 1,
                        ..Default::default()
                    }),
                ],
                ..Default::default()
            }],
            max_kv_count_for_xor16: Some(10),
            ..Default::default()
        };

        assert!(!task.contains_range_tombstone());
        assert!(!task.contains_split_sst());
        assert!(!task.should_use_block_based_filter());

        let task = CompactTask {
            input_ssts: vec![InputLevel {
                table_infos: vec![test_read_property_sstable(vec![TableId::new(1)])],
                ..Default::default()
            }],
            max_kv_count_for_xor16: Some(10),
            ..Default::default()
        };

        assert!(task.contains_range_tombstone());
        assert!(task.contains_split_sst());
        assert!(task.should_use_block_based_filter());
    }

    #[test]
    fn test_task_label_for_trivial_move_and_reclaim() {
        let trivial_move_task = CompactTask {
            input_ssts: vec![
                InputLevel {
                    level_idx: 1,
                    level_type: PbLevelType::Nonoverlapping,
                    table_infos: vec![
                        test_sstable(10, vec![TableId::new(1)]),
                        test_sstable(11, vec![]),
                    ],
                },
                InputLevel {
                    level_idx: 2,
                    level_type: PbLevelType::Nonoverlapping,
                    table_infos: vec![],
                },
            ],
            target_level: 2,
            task_type: TaskType::Dynamic,
            ..Default::default()
        };

        assert!(!trivial_move_task.is_trivial_reclaim());
        assert_eq!(trivial_move_task.task_label(), "trivial-move");

        let trivial_reclaim_task = CompactTask {
            input_ssts: vec![
                InputLevel {
                    level_idx: 1,
                    level_type: PbLevelType::Nonoverlapping,
                    table_infos: vec![test_sstable(10, vec![])],
                },
                InputLevel {
                    level_idx: 2,
                    level_type: PbLevelType::Nonoverlapping,
                    table_infos: vec![],
                },
            ],
            target_level: 2,
            task_type: TaskType::Dynamic,
            ..Default::default()
        };

        assert!(trivial_reclaim_task.is_trivial_reclaim());
        assert!(trivial_reclaim_task.is_trivial_move_task());
        assert_eq!(trivial_reclaim_task.task_label(), "trivial-space-reclaim");
    }
}