How to use: 1 Select a problem → 2 Write or load a student query → 3 Click ▶ Analyze Query → 4 Explore tabs 2–6 for detailed results
Base Query (Reference)
Read-only
Student Query
Editable
📋
Run analysis on the Editor tab to see detailed feedback.
Abstract Syntax Tree Comparison
🌳
Run analysis first.
Provenance Trace
🔍
Run analysis first.
Edge Case Analysis
⚡
Run analysis first.
Database Schema
Sample Data
Misconception Analysis
🧠
Run analysis first.
EviSQL
A research prototype for automated SQL query evaluation & feedback generation.
A structural detector proposes a diagnosis; an evidence gate releases it only when execution shows the effect that diagnosis itself predicts.
System Overview Combines real SQL execution (SQLite), AST-based structural comparison, provenance tracing (I-REX / RATest), edge-case testing (5 specialized databases), and misconception detection to deliver rich, actionable feedback — with special focus on relational division.
Pipeline Architecture
① Parse
SQL → AST
Validate
→
② Diff
AST compare
Structure
→
③ Execute
Real SQLite
Results
→
④ Provenance
Trace lineage
Divergence
→
⑤ Edge Cases
5 test DBs
Coverage
→
⑥ Feedback
Grade + Report
Suggest
vs. Related Work
FeatureRATest '19I-REX '20CaJaDE '22This System
Real SQL execution✓✓✓✓
Small counterexamples✓✓—✓
Division query focus———✓
Edge case library———✓
Graded feedback———✓
Misconception detection———✓
Context-augmented prov.——✓Planned
📘 EviSQL — research prototype Structural detector proposes  →  evidence gate releases only what execution supports