The Analyzer: point at a flat file, review the inferred schema plan, then emit DDL and load scripts — or execute straight into PostgreSQL.
Flat files in. A real relational database out. Every run on the record.
SchemaShift is a desktop application for Windows and Linux that reverse-engineers legacy flat files — CSV, Excel, pipe-delimited, and fixed-width extracts — into fully relational PostgreSQL schemas. It infers tables, keys, and foreign-key relationships from the data itself, shows you an editable plan before a single line of DDL is written, and records every run and every emitted script to a tamper-evident, hash-chained audit log. Deterministic and rule-based — runs entirely on your hardware, no cloud, no telemetry.
The Analyzer: point at a flat file, review the inferred schema plan, then emit DDL and load scripts — or execute straight into PostgreSQL.
Point SchemaShift at a flat file and it detects entities, dimensions, and relationships — a single wide extract becomes a set of properly normalized tables with primary keys assigned.
Every inference lands in a reviewable schema plan first. Rename tables, correct entities, and approve the design before any DDL is generated — the operator catches bad guesses, not production.
DDL, load scripts, and direct execution are independent steps you combine as needed — from a reviewed script handoff to fully unattended, zero-prompt migrations. GUI and CLI in one install.
Cross-file relationships are auto-detected by matching columns to existing keys, then verified with a zero-orphan check before any constraint is enforced. Explicit manual links fail loudly on bad data.
Malformed rows are never silently skipped — or worse, silently mutated. They're preserved verbatim in a quarantine sidecar with a count that must balance: rows in equals rows staged plus rows quarantined, every time.
Every run and every emitted script is logged to a hash-chained audit database. A broken or missing chain writes a permanent discontinuity record, and archived logs are GPG-signed and re-verified.
Streams million-row files in constant memory and has been proven on multi-file bank core extracts — hundreds of thousands of rows landed into an enforced relational schema.
No content filtering, no redaction, no guessing. Empty means NULL but "NA" is data, account numbers are text rather than math, inferred constraints are suggestions you approve — and anything ambiguous stops instead of assuming.
Same signing key, same audit discipline, same house. DataShift makes legacy data examiner-ready; SchemaShift gives that data a real relational home to live in.
Leaving a legacy core? Land your extracts in a modern, queryable SQL database without shipping data to anyone's cloud.
Repeatable, scripted migrations with an audit trail — the same input and plan produce the same schema, every run.
Turn a mainframe dump into a browsable relational schema in minutes — then query it with any standard SQL tool.
Request a demo and watch your own legacy extract become a fully relational database — keys enforced, rows reconciled, on the record.
Request a Demo Download the One-Pager (PDF) Print Edition (PDF) info@cobblestone-digital.com