Fix: Add missing delete_time index to FileTrash table

[yaoge123]

Fix: Add missing delete_time index to FileTrash table

Problem

The FileTrash table is missing the delete_time index that is defined in the Python model but never added to mysql.sql.

Python Model Definition

File: seafevents/events/models.py

class FileTrash(Base):
    __tablename__ = 'FileTrash'
    
    id = mapped_column(Integer, primary_key=True, autoincrement=True)
    delete_time = mapped_column(DateTime, nullable=False, index=True)  # <-- index=True
    # ...

Current mysql.sql (Missing Index)

CREATE TABLE `FileTrash` (
  -- ... columns ...
  PRIMARY KEY (`id`),
  KEY `ix_FileTrash_repo_id` (`repo_id`)
  -- Missing: KEY `ix_FileTrash_delete_time` (`delete_time`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci;

Code Usage

Global Cleanup Query

File: seafevents/events/db.py

def clean_up_file_trash(session, repo_id=None):
    _timestamp = datetime.datetime.now() - timedelta(days=FILE_TRASH_CLEANING_DAYS)
    
    if repo_id is not None:
        # Uses repo_id index
        stmt = delete(FileTrash).where(FileTrash.repo_id == repo_id, FileTrash.delete_time < _timestamp)
    else:
        # GLOBAL CLEANUP - No repo_id filter!
        # Line ~304
        stmt = delete(FileTrash).where(FileTrash.delete_time < _timestamp)  # <-- Needs delete_time index

Query Pattern: WHERE delete_time < ?
Execution Context: Scheduled cleanup task (runs periodically)

Repository-level Cleanup

File: seafevents/events/handlers.py

# Line ~616
if trash_item and trash_item.delete_time > _timestamp:
    # ...

Why This Index is Critical

Scenario	Query	Without Index	With Index
Global cleanup	WHERE delete_time < ?	Full table scan	Index range scan
Repository cleanup	WHERE repo_id = ? AND delete_time < ?	Uses repo_id index	Uses repo_id index (compound covers)

The Global Cleanup Problem

The scheduled cleanup task runs without a repo_id filter:

DELETE FROM FileTrash WHERE delete_time < '2024-01-01'

Without the delete_time index:

• Full table scan on potentially millions of rows
• High CPU and I/O usage during cleanup
• Risk of locking the table for extended periods

Changes

Add the missing index to sql/mysql.sql:

CREATE TABLE `FileTrash` (
  -- ... existing columns ...
  PRIMARY KEY (`id`),
  KEY `ix_FileTrash_repo_id` (`repo_id`),
  KEY `ix_FileTrash_delete_time` (`delete_time`)  -- Added
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci;

Note on Compound Index

PR #8680 (seahub) / PR #606 (seafevents) also adds a compound index (repo_id, delete_time) for the repository-specific query pattern:

WHERE repo_id = ? AND delete_time > ? ORDER BY delete_time DESC

However, the single-column delete_time index is still needed for:

1. Global cleanup (WHERE delete_time < ? without repo_id filter)
2. The compound index (repo_id, delete_time) cannot satisfy queries that don't include repo_id

Both indexes should coexist:

• ix_FileTrash_delete_time (delete_time) - For global cleanup
• idx_filetrash_repo_delete_time (repo_id, delete_time) - For repo-specific queries (added in PR Remove redundant ix_FileTrash_repo_id index (covered by composite index in seafevents#606) #8680)

Verification

This index is already defined in the Python model (seafevents/events/models.py) but was simply forgotten in mysql.sql.