Comparing changes

- Added 'employed persons', 'employed', 'employment level', 'total employment', 'labour force' to STOCK_PATTERNS - Added comprehensive E2E test for Employed Persons indicator - Test verifies stock indicators do NOT get time conversion (periodicity is just release cadence) - Test covers quarterly and monthly data from Angola, Albania, Armenia - Validates that values stay unchanged (no division by time period) - Validates normalized units are 'Thousands' not 'thousands per month' - Documents that auto-targeting should skip time dimension for stock indicators All 409 tests passing ✅

…ate indicators BREAKING CHANGE: Auto-targeting now intelligently skips time dimension for stock and rate indicators ## What Changed Auto-targeting now uses indicator classification to determine which dimensions to target: - **Stock indicators** (population, debt, reserves, employed persons) → Skip time dimension - **Rate indicators** (CPI, unemployment rate, inflation) → Skip time dimension - **Flow indicators** (GDP, exports, sales, revenue) → Include time dimension - **Non-monetary indicators** → Can participate in magnitude/time targeting when currency not in dimensions ## Why This Matters Previously, auto-targeting would apply time conversion to ALL indicators, including stocks: - ❌ "Employed Persons: 12,814 thousand" → divided by 3 → "4,271 thousand per month" - ❌ "Debt: 500 billion" → "500 billion per month" (nonsensical) Now it correctly handles each type: - ✅ "Employed Persons: 12,814 thousand" → stays "12,814 thousand" (snapshot) - ✅ "GDP: 300 million per quarter" → "100 million per month" (flow, converted) - ✅ "CPI: 105.2 points" → stays "105.2 points" (index, no time) ## Implementation 1. **Auto-targeting classification**: Calls `classifyIndicator()` for each indicator group 2. **Smart dimension selection**: Skips time dimension when `classification.type === 'stock' || 'rate'` 3. **Non-monetary inclusion**: Only filters by `isMonetary` when currency is in `autoTargetDimensions` 4. **Clear reasoning**: Explain metadata shows `time=skipped(stock indicator, no time dimension)` ## Test Updates - Added comprehensive E2E test for Employed Persons (stock indicator) - Updated auto-targeting tests to use flow indicators (Exports) instead of stock (Debt) - Added test for stock indicators automatically skipping time dimension - Updated tests to handle non-monetary indicators in magnitude/time targeting ## Files Changed - `src/normalization/auto_targets.ts` - Smart classification-based dimension selection - `src/workflows/e2e_comprehensive_test.ts` - E2E test for Employed Persons - `src/normalization/auto_targets_test.ts` - Updated tests for new behavior - `src/api/pipeline_api_test.ts` - Fixed tests to use flow indicators All 410 tests passing ✅

## Documentation Updates Added detailed documentation for the new smart auto-targeting feature that intelligently skips time dimension for stock/rate indicators. ### Changes 1. **README.md** - Added Smart Auto-Targeting to Core Capabilities - Highlighted the prevention of incorrect conversions (e.g., employed persons ÷ 3) 2. **docs/guides/per-indicator-normalization.md** - Added comprehensive "Smart Auto-Targeting: Indicator Classification" section - Documented all three indicator types (Stock, Flow, Rate) with examples - Showed before/after comparisons demonstrating the fix - Explained global configuration with smart targeting - Updated summary to include smart targeting benefits ### Key Documentation Points - **Stock Indicators**: Population, Debt, Employed Persons → Time dimension skipped - **Flow Indicators**: GDP, Exports, Revenue → Time dimension included - **Rate Indicators**: CPI, Inflation, Unemployment Rate → Time dimension skipped - **Real-world examples** showing correct vs incorrect behavior - **Configuration guidance** for using global autoTargetDimensions ### Examples Included - Employed Persons: 12,814 thousand stays 12,814 (not divided by 3) - GDP: 300M/quarter → 100M/month (correctly converted) - CPI: 105.2 points stays 105.2 (no conversion) This documentation helps users understand: 1. Why their stock indicator data was being incorrectly converted 2. How the new smart targeting fixes it automatically 3. How to configure their pipeline to benefit from it

## Version 1.2.0 Release Major feature release introducing Smart Auto-Targeting that fixes critical stock indicator conversion bugs. ### Highlights 🎯 **Smart Auto-Targeting** - Automatically detects stock/flow/rate indicators - Skips time dimension for stock indicators (Population, Debt, Employed Persons) - Prevents incorrect conversions: "12,814 employed persons" no longer ÷ 3 → "4,271 per month" 🐛 **Critical Bug Fix** - Stock indicators were incorrectly time-converted - Now correctly treats them as snapshots, not rates over time - Affects: Population, Debt, Reserves, Money Supply, Assets, Employed Persons, etc. 📚 **Comprehensive Documentation** - Added Smart Auto-Targeting guide with real-world examples - Before/after comparisons showing the fix - Configuration guidance for global autoTargetDimensions ✅ **All 410 Tests Passing** ### Breaking Changes Auto-targeting behavior changed for stock/rate indicators - they now correctly skip time dimension normalization. This is a correctness fix for mathematically incorrect behavior. ### Files Changed - CHANGELOG.md - Added v1.2.0 release notes - deno.json - Bumped version to 1.2.0

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Comparing changes

Open a pull request

Commits on Sep 30, 2025

This comparison is taking too long to generate.

Uh oh!