Execute the cleanup: delete S3 objects first, then DB rows in one transaction. S3 before DB so a partial S3 failure leaves only orphaned (logged) obje

Process CSV file and insert records into PostgreSQL database Stores only metadata (reference fields) to minimize storage @param csvPath Path to the C

Get all versions of a record by ID Fetches full record data from S3 CSV files

Executes the given operation with retry logic and exponential backoff. @param operation The operation to execute @param maxAttempts Maximum n

Read one CSV line from a stream, handling newlines inside quoted fields. Returns null at EOF.

Check if a value looks like a Salesforce ID (15 or 18 alphanumeric characters)

Classify the status of a record from its version history. A record can have multiple statuses simultaneously.

Compute the cutoff period (YYMM) for a retention window. Backups with period > 0 AND period < cutoff are eligible for cleanup. period is YYMM-encoded

Format a timestamp for SOQL WHERE clause

Generates a comprehensive summary report of the backup job. @return Formatted summary report string

Determines if an exception is retryable (transient error) or not. Retryable errors: - IOException (network timeouts, connection failures) - WebClient

Complete a backup run with final statistics

Find the index of a column in the header array (case-insensitive)

Format a period integer into human-readable form. 2601 -> "Jan 2026", -2602 -> "Archive Feb 2026"

Calculate current period as YYMM (e.g., 2511 for November 2025)

Get object statistics

Get restore history for a record

Simple CSV line parser that handles quoted fields

Calculate diff between two versions

Find column index by name (case-insensitive)

Log a restore operation

Gate: cleanup runs only when enabled AND today is the configured day-of-month. Since the batch job runs daily, this keeps the cleanup to once a month.

Find the best column to use as display name. Tries Name first, then common alternatives for objects that don't have a Name field.

Get current batch size for an object. Returns default if not yet adjusted.

Get the latest checkpoint for an object to resume from. Only looks within the CURRENT period - a new period means a fresh full backup. Within a period

Create a new backup run in the database Returns the generated run_id

Get deleted records for a specific object type

Get related child records that reference this record ID in their metadata

Update checkpoint for a running backup This is called frequently (every 10-20 files) so it must be fast

Get the latest archive checkpoint for an object. Looks across ALL archive periods (period < 0) so the checkpoint is never reset by month changes.

Get list of objects that have deleted records

Get related files for a record via ContentDocumentLink -> ContentVersion. Split into two queries to avoid self-join on the large objects table.

Create the necessary database tables if they don't exist Uses the schema from "One Backup to Rule Them All"

Mark a file as failed (after all retries exhausted) Failed files block the sliding window - checkpoint won't advance past them @param recordId Conten

Resets all statistics. Useful for testing or starting a new job.