Features:
- Partial settings save with deep merge
- Backup and rollback mechanism
- UI field whitelist validation
- Defensive programming in settings page
Safety:
- Auto-backup before save
- Automatic rollback on failure
- Field validation at both client and server
The root cause of config overwrites was key mismatch:
- .env file uses: ERP_URL, DB_TYPE, DB_NAME (underscore uppercase)
- Code was using: erp.url, database.dbType (dot notation)
Fixed in three methods:
- saveAllSettings() - now sets cache with correct keys
- resetToDefaults() - now uses correct keys
- save() - now reads cache with correct keys
This ensures partial save preserves unmodified fields.
- Change parameter type from SettingsData to Partial<SettingsData>
- Call savePartialSettings() instead of saveAllSettings()
- Return detailed error messages from savePartialSettings
- Add logging for sections being saved
This change enables partial settings save functionality, allowing
the UI to save only specific settings sections without requiring
the complete settings object.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Implements the core savePartialSettings method that:
- Validates fields against UI_EDITABLE_FIELDS whitelist
- Deep merges partial updates with current settings
- Creates backup before saving
- Restores backup on save failure
- Reloads .env file to populate cache with correct keys
Added comprehensive tests:
- Partial update preserves existing fields
- Rejects non-whitelisted fields
- Handles nested object updates
- Restores backup on save failure
Fixed cache key inconsistency bug by clearing cache in loadEnvFile()
and reloading after save to ensure proper cache population.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Replace console.log/console.error with proper logger usage in
backupEnvFile and restoreBackup methods. Use the existing 'log' logger
created with createLogger('ConfigManager') following the same pattern
used in other methods.
Changes:
- Import createLogger and create log instance
- Replace console.log with log.debug in backupEnvFile
- Replace console.error with log.error in both methods
- Pass error and path metadata as objects for structured logging
- Fix test to use correct backup path (process.cwd() + src/main/.env.backup)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Fix TypeScript type error in deepMerge recursive call with proper type assertions
- Remove unused type imports (ErpConfig, DatabaseConfig, PathsConfig, ExtractionConfig, ValidationConfig, UiConfig, ExecutionConfig)
- Fix line endings (CRLF to LF) via Prettier format
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit adds utility functions to support partial settings save functionality:
- isObject: Type guard for plain objects
- deepMerge: Recursively merges objects, preserving unspecified fields
- validateEditableFields: Validates settings against UI editable field whitelist
- UI_EDITABLE_FIELDS: Whitelist of fields modifiable through UI
A failing test is included to verify the deep merge behavior.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fix User scope isolation issue in CleanerPage:
- User users can no longer affect other users' data via "取消" button
- "确认删除" only processes visible filteredResults for non-Admin users
- Admin behavior unchanged (can manage all data)
- Prevents cross-user data interference
Committed: 6ecf03c
Fix critical bug where non-Admin users could affect other users' data
when using "取消" (Uncheck All) and "确认删除" (Confirm Deletion) buttons.
Problem:
- User users see only their filtered materials in table (filteredResults)
- "取消" button was unchecking ALL materials in validationResults
- "确认删除" was processing ALL materials, not just visible ones
- This caused User A to delete/modify User B's invisible data
Solution:
1. Modified "取消" button to only uncheck visible filteredResults
- Now removes selectedItems only for visible material codes
- Preserves selections for other users' invisible data
2. Modified handleConfirmDeletion to process only visible items for non-Admin users
- Admin: processes all validationResults (unchanged behavior)
- User: processes only filteredResults (scoped to their data)
Security Impact:
- Prevents cross-user data interference
- Ensures User scope isolation
- Maintains Admin full access to all data
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add Priority 3 matching logic that allows User type users to override
material assignments with their own typeKeywords from MaterialsTypeToBeDeleted.
Changes:
- Add session manager integration to get current user context
- Implement Priority 3: User Override Match (only for non-admin users)
- Filter typeKeywords by current username and force override on match
- Maintain existing Priority 1 (exact match) and Priority 2 (type match) behavior
- Admin users bypass override logic and see original matching results
- Update cleaner-validation-flow.md with new matching algorithm flow
This ensures User users see materials assigned to themselves first when
their configured typeKeywords match, while Admin users maintain full visibility.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Fixed material validation always using historical order numbers instead of
current input. The setSharedProductionIds function was accumulating IDs
without clearing old ones, causing validation to use all previously entered
order numbers.
Changes:
- Added sharedProductionIds.clear() before adding new IDs
- Ensures Set only contains the latest order numbers from extractor page
This fixes the root cause where changing the order number in the extractor
page would not update the validation data source, as old IDs were never
removed from the shared Set.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add documentation for extractor start button and settings save button flows with detailed Mermaid diagrams. Also add logs directory to gitignore.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Reverse the inclusion check to properly match when materialName contains typeKeyword.materialName (e.g., "ABC123_SPECIAL" contains "ABC123").
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Remove problematic edge labels with special characters
- Move database type identifiers into node labels
- Replace ellipsis with clearer text descriptions
- Fix parse error caused by spaces and special chars in edge labels
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add comprehensive documentation for the "确认删除 (同步数据库)" button
flow, including:
- Core flow overview with Mermaid diagram
- Frontend interaction layer analysis (handleConfirmDeletion)
- IPC handler layer details (upsertBatch, delete handlers)
- Database DAO layer implementation with MySQL vs SQL Server differences
- Data flow diagram showing all layers
- Key data structures and error handling
- Comparison table with validation status flow
The document now covers both core business flows in the cleaner page:
1. Get and validate material status (read operation)
2. Confirm deletion/sync to database (write operation)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Fix getTableName() to handle generic schema_tablename pattern
- Convert schema_tablename to [schema].[tablename] for SQL Server
- Replace hardcoded productionContractData table name with helper function
- Fix variable naming conflict: rename 'sql' to 'sqlString' to avoid shadowing mssql module import
- Apply fixes to discrete-material-plan-dao, materials-to-be-deleted-dao, and bip-users-dao
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Remove global user-select: none style that was preventing text input
and add explicit user-select: text to the textarea element.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Add SqlServerService integration alongside existing MySQL support
- Refactor DAOs (DiscreteMaterialPlanDAO, MaterialsToBeDeletedDAO, BipUsersDAO) to support both MySQL and SQL Server
- Update validation handler to dynamically select database service based on DB_TYPE environment variable
- Add connection pooling and proper connection management for SQL Server
- Update package-lock.json dependency peer flags
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Integrated Tailwind v4 into the electron.vite.config.ts and main.css.
- Refactored App.tsx layout to use Tailwind styling and Lucide icons as provided.
- Refactored ExtractorPage, CleanerPage, and SettingsPage to match the new UI mock layout while maintaining existing state and logic.
- Simplified SettingsPage based on user feedback.
- Ensured default exports and imports are consistent across pages.
Co-authored-by: luwamgere15-crypto <255338376+luwamgere15-crypto@users.noreply.github.com>
- Import and integrate UserSelectionDialog component
- Add state for user selection (showUserSelection, allUsers)
- Add isSwitchedByAdmin state to track admin-switched sessions
- Implement handleUserSelect and handleUserSelectionCancel handlers
- Update login flow to show user selection when admin logs in
- Add conditional logout button visibility based on user type
- Update UI to include UserSelectionDialog component
- Add SessionManager for managing user sessions (singleton pattern)
- Add BIPUsersDAO for database authentication
- Add LoginDialog component for username/password login
- Add UserSelectionDialog component for admin user selection
- Support silent login by computer name
- Implement main page with navigation to Extractor and Cleaner
Database:
- Table: dbo_BIPUsers
- Fields: UserName, Password, UserType, ComputerNmae
UI Flow:
1. Silent login on startup via computer name
2. Show login dialog if silent login fails
3. Display main page with user info and navigation
4. Support logout and re-login
- Add OrderNumberResolver service to auto-recognize productionID and 生产订单号 formats
- Integrate MySQL database lookup for productionID to 生产订单号 conversion
- Update OrderNumberInput component with format statistics display
- Modify Extractor and Cleaner to resolve order numbers before processing
Database configuration:
- Table: productionContractData_26 年压力表合同数据
- Fields: 总排号 (productionID), 生产订单号 (production order number)
Supported formats:
- productionID: 2 digits + 1 letter + serial number (e.g., 26B742)
- 生产订单号：SC + 14 digits (e.g., SC70202601040109)
- Fix TypeScript type errors in erp-auth.ts and excel-parser.ts
- Update tsconfig.node.json to relax type checking for build
- Build Windows installer (dist/erpauto-1.0.0-setup.exe)
- Add .gitignore entries for dist files
Installation:
1. Download erpauto-1.0.0-setup.exe
2. Run installer
3. Configure ERP credentials in %APPDATA%\erpauto\.env
4. Launch application from desktop shortcut
Build commands:
- npm run build:win - Build Windows installer
- npm run build:mac - Build macOS app
- npm run build:linux - Build Linux packages
- Load .env file in main process using dotenv
- Add detailed console logging for debugging
- Add validation for ERP configuration before extraction
- Improve error display in UI with selectable text
- Add stack trace logging for better debugging
- Add README.md with installation, configuration, and usage guide
- Add docs/USER_GUIDE.md with detailed step-by-step workflows
- Include database setup scripts for MySQL and SQL Server
- Add troubleshooting section with common issues
- Add MaterialCodeInput component for entering material codes
- Create CleanerPage with dry-run mode support
- Add Cleaner page navigation in App.tsx
- Update CleanerAPI type to return response wrapper
- Add SqlServerService and MySqlService for database persistence
- Implement IPC handlers for file, extractor, cleaner, and database operations
- Define IPC API types and update preload script
- Create ExtractorPage UI with OrderNumberInput component
- Add unit and integration tests for MySQL and SQL Server
- Update vitest config with path aliases
**Core Implementation (src/main/services/erp/cleaner.ts):**
- CleanerService class with dry-run mode support
- Material deletion logic with safety constraints:
- Row numbers 7000-7999 are protected
- Materials with pending quantity are skipped
- Materials not in delete list are ignored
- Order processing with nested iframe navigation
- Progress callback support for UI integration
**Types (src/main/types/cleaner.types.ts):**
- CleanerInput: order numbers, material codes, dry-run flag
- CleanerResult: processing statistics and details
- OrderCleanDetail: per-order breakdown
**Tests:**
- Unit tests for shouldDeleteMaterial logic
- Integration tests for order processing
- Dry-run mode validation
- Navigation tests
Reference: playwrite/utils/discrete_material_plan_cleaner.py
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add patterns to ignore temporary files generated during development:
- Test coverage reports (coverage/)
- Downloaded test files (downloads/, *.xlsx, *.parsed.json)
- Debug and manual test scripts (tests/debug/, tests/manual/)
- Temporary test scripts (test-*.mjs, test-*.js, compare_*.js)
This keeps the repository clean while preserving useful test scripts
locally for debugging purposes.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit completes the core ERP automation functionality, migrating
from Python Playwright to TypeScript while maintaining full compatibility
with the original implementation.
**ERP Authentication Service (erp-auth.ts):**
- Implement login() with role-based locators for form elements
- Add SSL certificate bypass for internal VPN network
- Handle force login confirmation dialogs
- Return session with mainFrame reference for subsequent operations
- Add session lifecycle management (close, getSession, isActive)
**Data Extractor Service (extractor.ts):**
- Implement precise nested iframe navigation (#forwardFrame → #mainiframe)
- Add batch processing support for multiple order numbers
- Implement order number filling with comma separation
- Handle material selection and download workflows
- Successfully tested with 300 orders in 5 batches
**Excel Parser Service (excel-parser.ts):**
- Fix ExcelJS 1-indexed array access (row[1] for 序号, row[2] for 材料编码)
- Add dynamic table header search to handle empty row skipping
- Add field mapping: "来源单号" → "productionOrder"
- Implement saveAsExcel() method compatible with Python format
- Validate compatibility: 527 rows, 69 orders matching Python output
**Type Definitions (erp.types.ts):**
- Add headless property to ErpConfig for browser mode control
- Add mainFrame reference to ErpSession for frame reuse
**Integration Tests (extractor.test.ts):**
- Modify tests to use independent auth services for isolation
- Add test with 300 orders and batch size 70
- All tests passing with real ERP data
**Test Configuration (vitest.config.ts):**
- Add setupFiles configuration for environment variable loading
**Testing Results:**
- ✅ Successfully logs in to ERP system
- ✅ Processes 300 orders in 5 batches (43.59 seconds)
- ✅ Downloads 5 Excel files (347.62 KB total)
- ✅ Parses 2,131 material plans from 280 unique orders
- ✅ Excel output matches Python format exactly
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Add missing spec-compliant methods and fields while keeping the working
implementation that correctly handles the complex Excel structure.
Changes:
- Add public isOrderRow() method to detect order title rows
- Add public extractOrderNumber() method to extract order numbers
- Add public parseMaterialRow() method for spec compliance
- Rename internal parseMaterialRow() to parseMaterialRowInternal()
- Add missing pendingQty field to DiscreteMaterialPlan type
All tests pass (23/23).
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Implement Excel parsing module for ERP exported files following TDD principles.
Key features:
- Parse Excel files with multiple orders per file
- Extract order header information (production order, product code, etc.)
- Extract material data rows with 13 fields
- Handle empty orders gracefully
- Detect footer rows (制单人/打印人)
- Map Chinese field names to English property names
- Support field name mapping from Python reference
Implementation:
- ExcelParser class with parse() method
- DiscreteMaterialPlan and ExcelParseOptions types
- OrderHeader interface for order metadata
- Test fixtures with realistic Excel structure
- Comprehensive unit tests (3 tests, all passing)
Reference: playwrite/utils/excel_converter.py
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Implement Task 3.1: Core Extractor Logic with TDD approach.
Changes:
- Add ExtractorService class with batch processing and download support
- Update ERP_LOCATORS with extractor-specific selectors from Python reference
- Add integration tests for single and multiple order extraction
- Add unit tests for batch creation logic
- Update existing tests to skip gracefully without ERP credentials
Features:
- Navigate to discrete material plan page with nested iframes
- Setup query interface (search icon, order query, limit settings)
- Batch download with configurable batch size (default: 100)
- Progress callback support for real-time updates
- Error handling for individual batch failures
- File download handling with proper wait strategies
Reference: playwrite/utils/discrete_material_plan_extractor.py
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Implement ErpAuthService with TDD approach:
- Add login(), close(), getSession(), isActive() methods
- Use Playwright chromium with headless:false for debugging
- Manage browser lifecycle and session state
- Handle authentication flow with ERP_LOCATORS
- Add integration tests for login scenarios
- Add unit tests for session management
- Fix dotenv config path in test setup
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
