# v0.1.3 — SQL Coverage Completion and Hardening

> **Full technical details:** [v0.1.x.md-full.md](v0.1.x.md-full.md)

**Status: ✅ Released** | **Scope: Patch** (~3 weeks)

> Closes nearly all known SQL compatibility gaps, hardens the WAL decoder,
> and validates all 22 TPC-H benchmark queries.

---

## What problem does this solve?

v0.1.0 and its patches covered the most common SQL patterns well, but
systematic testing against the full TPC-H benchmark suite revealed gaps in
handling specific SQL constructs. v0.1.3 is a comprehensive hardening release
that systematically closes those gaps.

---

## 50 of 51 SQL Gaps Closed

A systematic audit of the differential engine against a broad SQL test
corpus identified 51 missing or incorrect behaviours. This release resolves
50 of them — edge cases in HAVING clauses, subquery handling, column
aliasing, and JOIN combinations that could produce incorrect differential
results.

*In plain terms:* pg_trickle now handles a significantly wider range of SQL
queries correctly in differential mode, matching what you would get from a
full recomputation.

---

## Adaptive Fallback Improvements

When a query cannot be processed differentially (because it uses a SQL
pattern the engine does not yet support), pg_trickle falls back to a full
refresh. The fallback logic was made smarter: it now detects more patterns
early, gives clearer warning messages about *why* a fallback occurred, and
avoids unnecessary fallbacks for patterns that are actually supported.

---

## WAL Decoder Hardening

The WAL-based change decoder received robustness improvements for edge cases:
tables with unusual OID distributions, schema reloads mid-transaction, and
replication slot restarts after a PostgreSQL restart.

---

## TPC-H 22/22 Passing

The industry-standard **TPC-H benchmark** consists of 22 complex analytical
queries representing realistic business reporting workloads. As of v0.1.3,
all 22 queries now pass correctness validation — their differential results
match the expected output of a full recomputation.

---

## Test Suite Growth: 460 End-to-End Tests

The end-to-end test suite grew to 460 tests covering the breadth of SQL
patterns and operational scenarios. These run against a real PostgreSQL
instance using Docker containers.

---

## Scope

v0.1.3 closes the SQL compatibility gaps found during systematic testing and
hardens the operational robustness of the WAL decoder. The TPC-H 22/22
milestone marks the first time the engine is validated against an industry
benchmark suite. Upgrade from any v0.1.x is straightforward.