Changelog

All notable changes to utf8proj are documented here.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.17.0] - 2026-02-09

Added

Duplicate task ID validation (E004) — Sibling tasks sharing the same ID now emit an E004 error diagnostic, preventing undefined scheduling behavior
Gantt header date range label — Header now shows the actual date range (e.g. "Feb – Mar 2026") instead of only the start month; applies to both HTML and SVG renderers

[0.16.1] - 2026-02-09

Added

Critical path toggle — New highlight_critical option on HtmlGanttRenderer allows disabling red critical path highlighting (all tasks use normal color when off)
- Builder method: renderer.hide_critical_path()
- WASM binding: playground.set_highlight_critical(enabled)
- Playground UI: "Critical Path" checkbox in header toolbar (persisted in share URLs)

Fixed

Gantt chart scale shows empty space beyond tasks — The x-axis now computes a tight date range from actually rendered task bars instead of using project.start / schedule.project_end, eliminating wasted space when the project start date precedes the first task or the last task finishes before project_end

[0.12.0] - 2026-01-23

Added

Hybrid BDD Leveling (RFC-0014 Phase 1) — 4-5x faster resource leveling for large projects
- New --leveling-strategy=hybrid CLI flag enables BDD-based conflict cluster analysis
- Uses BDD to identify independent conflict clusters, then applies heuristic leveling within clusters
- Achieves O(n + sum(k²)) complexity instead of O(n²), where k is cluster size
- Produces identical results to standard leveling (deterministic, same project end dates)
- New types: ConflictCluster, ClusterAnalysis for BDD-based conflict analysis
- Diagnostic notes now show cluster information (e.g., "Cluster: 5 tasks competing for 2 resources")

Performance

Hybrid leveling provides 4-5x speedup compared to standard leveling:
- 500 tasks: 0.23s → 0.05s (4.6x faster)
- 1000 tasks: 0.82s → 0.18s (4.5x faster)
- 2000 tasks: 5.35s → 1.07s (5x faster)
Standard leveling unchanged (default for backwards compatibility)

[0.11.2] - 2026-01-22

Changed

Resource leveling optimization (RFC-0014 Phase 0) — Improved slot-finding performance
- Changed ResourceTimeline.usage from HashMap to BTreeMap for sorted iteration
- Added skip-blocked-runs algorithm: when a slot is blocked, skip directly to the end of the blocked period instead of checking day-by-day
- Added SlotCheckResult enum and find_blocked_run_end() helper for efficient gap detection

Performance

Leveling performance improved for projects with heavy resource conflicts:
- 100 tasks (88 leveled): 0.16s
- 500 tasks (492 leveled): 0.12s
- 1000 tasks (986 leveled): 0.7s
- 2000 tasks (1983 leveled): 5.4s
Larger projects (5000+ tasks) still benefit from the 2000-day search limit but require Phase 1 (Hybrid BDD leveling) for sub-second performance

[0.11.1] - 2026-01-21

Fixed

Resource leveling infinite loop — find_available_slot() had an unbounded loop that would hang forever when no slot could be found (e.g., when task units exceed resource capacity). Real project files with resource assignments would hang while synthetic benchmarks worked.
- Added 2000 working day search limit to prevent infinite loops
- Function now returns Option<NaiveDate> to handle "no slot found" gracefully
- Early exit when units > capacity (impossible to schedule)
- Emits L002 diagnostic when slot cannot be found within search horizon

Added

Stress test examples for performance validation:
- examples/acso_tutorial.proj — TaskJuggler tutorial equivalent (15 tasks, 6 resources)
- examples/enterprise_10k.proj — Enterprise-scale project (10k tasks, 1k resources)
- tools/generate_large_project.py — Python generator for custom stress tests

Performance

Leveling now completes for real projects that previously hung:
- acso_tutorial.proj (15 tasks): hung → 0.3s
- crm_simple.proj (28 tasks): hung → 0.1s
- 1000 tasks with multi-assignment: hung → 0.4s
Scheduling (without leveling) scales to 1M+ tasks

[0.11.0] - 2026-01-20

Added

RFC-0013: Baseline Management — Schedule snapshots for variance analysis
- Capture frozen schedule snapshots to answer "Compared to what?"
- Two-file architecture: project.proj + project.proj.baselines (sidecar)
- Only leaf tasks are baselined (containers excluded to prevent double-counting)
- Immutable baselines (no --force flag — delete and recreate instead)
- Fully-qualified task IDs (e.g., phase1.design) prevent collisions
- New CLI commands:
  - utf8proj baseline save --name <name> — Save a baseline snapshot
  - utf8proj baseline list — List all baselines for a project
  - utf8proj baseline show --name <name> — Show baseline details
  - utf8proj baseline remove --name <name> — Remove a baseline (with confirmation)
  - utf8proj compare --baseline <name> — Compare current schedule vs baseline
- Output formats: text (default), CSV (--format csv), JSON (--format json)
- Filtering options: --show-unchanged, --threshold <days>
- New diagnostics: B001-B009 for baseline operations
- Full RFC in docs/rfc/RFC-0013-baseline-management.md

New Diagnostics

Code	Severity	Meaning
B001	Info	Baseline saved successfully
B002	Warning	Task lacks explicit ID (using inferred)
B003	Error	Baseline already exists
B004	Error	Baseline not found
B005	Info	Task removed since baseline
B006	Info	Task added since baseline
B007	Warning	No baselines file found
B008	Warning	Container excluded from baseline
B009	Error	Cannot baseline: tasks have no ID

Example Usage

# Save initial baseline
utf8proj baseline save --name original --description "Initial plan" project.proj

# Make changes to project, then compare
utf8proj compare --baseline original project.proj

# Output:
# Schedule Variance vs "original" (saved 2026-01-15)
#
# Task                 Baseline Finish   Current Finish   Variance
# design               2026-01-10        2026-01-12       +2d !!
# build                2026-02-15        2026-02-20       +5d !!
# [+] security_audit   -                 2026-02-28       (added)
#
# Summary:
#   Compared: 2 tasks
#   Delayed: 2
#   Added: 1
#   Project slip: +7 days

# JSON output for CI integration
utf8proj compare --baseline original --format json project.proj

Baseline File Format

# project.proj.baselines
baseline original {
    saved: 2026-01-15T10:30:00Z
    description: "Initial approved plan"

    design: 2026-01-01 -> 2026-01-10
    build: 2026-01-11 -> 2026-02-15
}

baseline change_order_1 {
    saved: 2026-02-01T14:20:00Z
    parent: original

    design: 2026-01-01 -> 2026-01-12
    build: 2026-01-13 -> 2026-02-20
}

[0.10.0] - 2026-01-18

Added

RFC-0012: Temporal Regimes — Explicit time semantics for tasks
- Three regimes: Work (effort-bearing), Event (point-in-time), Deadline (contractual)
- New regime: task attribute: regime: work, regime: event, regime: deadline
- Task.effective_regime() method derives regime from explicit setting or milestone flag
- New diagnostics: R001-R005 for regime validation
- --explain CLI flag shows detailed explanations for all diagnostic codes
- Full RFC in docs/rfc/RFC-0012-TEMPORAL-REGIMES.md

Changed

Milestones are now treated as events with exact dates (Event regime)
Solver no longer special-cases milestones — uses effective_regime() instead
Constraint rounding is regime-driven, not calendar-driven

Migration Notes (0.9.x → 0.10.0)

Backward Compatible: Existing projects work without changes.

The new Temporal Regimes feature is opt-in. If you don't specify regime:, the system uses:

Event for milestones (zero-duration tasks)
Work for all other tasks

Key behavioral improvements:

Milestones constrained to weekends/holidays now stay on those dates (Event regime)
Previously, milestones would round to the nearest working day

New regime: syntax (optional):

# Explicit Work regime (default for non-milestones)
task dev "Development" {
    effort: 5d
    regime: work
}

# Explicit Event regime (default for milestones)
milestone release "Release v2.0" {
    regime: event
    start_no_earlier_than: 2025-01-12  # Sunday - stays on Sunday
}

# Deadline regime (external constraints)
task contract "Contract Deadline" {
    duration: 1d
    regime: deadline
    finish_no_later_than: 2025-01-31
}

New diagnostics:

Code	Severity	Meaning
R001	Info	Event regime task has non-zero duration
R002	Info	Work regime constraint falls on non-working day
R003	Warning	Deadline regime without finish constraint
R004	Info	Implicit Event regime applied to milestone
R005	Info	Mixed regime dependency (informational)

Use --explain for detailed guidance:

utf8proj check --explain project.proj

[0.9.1] - 2026-01-16

Fixed

Excel Auto-Fit Bug Fixes
- Fixed add_schedule_sheet to use get_effective_weeks() instead of self.schedule_weeks
- Fixed calculate_auto_fit_weeks to use actual max task finish date (not just schedule.project_end)
- Both fixes ensure all tasks are covered in week columns for long projects
- Affected methods: write_schedule_row_simple, write_schedule_row_with_deps, write_week_columns, write_schedule_totals

[0.9.0] - 2026-01-15

Added

RFC-0006: Focus View for Gantt Charts
- focus() and context_depth() builder methods on HtmlGanttRenderer
- Pattern matching: prefix, contains, glob patterns for task filtering
- Context depth control: show ancestors/siblings of focused tasks
- CLI options: --focus="pattern" and --context-depth=N
- 12 focus view tests in crates/utf8proj-render/src/gantt.rs
- Full RFC in docs/rfc/RFC-0006-FOCUS-VIEW.md
RFC-0008: Progress-Aware CPM Scheduling
- Status date resolution chain: CLI --as-of > project.status_date > today()
- Task state classification: Complete/InProgress/NotStarted via ProgressState enum
- Progress-aware forward pass respects completion status
- Task.explicit_remaining field for user override of remaining work
- P005 diagnostic: remaining vs complete% conflicts
- P006 diagnostic: container progress mismatch (>10% threshold)
- 12 progress-aware CPM tests
- Full RFC in docs/rfc/RFC-0008-PROGRESS-AWARE-CPM.md

[0.2.0] - 2026-01-14

Added

GitHub Actions Release Workflow
- .github/workflows/release.yml for cross-platform binary distribution
- Tag-triggered releases on vX.Y.Z semantic version tags
- Builds for: Windows, Linux, macOS Intel, macOS ARM
- Produces utf8proj CLI and utf8proj-lsp binaries
- SHA256 checksums for integrity verification
- Unsigned binaries (matches ripgrep/bat/fd convention)

Fixed

Cargo.lock now tracked (was previously gitignored, required for --locked builds)

[Unreleased - 2026-01-10]

Fixed

mpp_to_proj Dependency Type Detection
- SS/FF/SF dependencies from MPP files now correctly converted (was always FS)
- MPXJ returns "SS", "FF", "SF" but converter checked for "START_START" etc.
- Fix: Check for both formats using .upper() in tools/mpp_to_proj/mpp_to_proj.py
Grammar Support for Dependency Type + Lag
- Grammar now allows type AND lag together (e.g., depends: task SS +5d)
- Changed grammar from dep_modifier? to dep_type? ~ dep_lag?
- Location: crates/utf8proj-parser/src/native/grammar.pest

[Unreleased - 2026-01-09]

Added

v1 Editor Support
- TextMate grammar for .proj syntax highlighting (syntax/utf8proj.tmLanguage.json)
- Vim syntax file for Neovim/Vim (syntax/proj.vim)
- LSP navigation: go-to-definition and find-references
- Supported symbols: tasks, resources, calendars, profiles, traits
- Setup instructions in docs/EDITOR_SETUP.md
- 7 navigation tests
RFC-0005 Status & LSP Leveling Hover
- docs/rfc/RFC-0005-RESOURCE-LEVELING-STATUS.md: Phase 1 complete, Phase 2 deferred
- L001-L004 diagnostics in LSP hover with icons
- Leveling coverage improved: 85.7% → 93.2%
MS Project Companion Tool: Effort Extraction
- Work field extraction in tools/mpp_to_proj/mpp_to_proj.py
- Maps MS Project "Work" → utf8proj effort: property
- 8 unit tests covering effort extraction

Fixed

MPXJ Unit Fix: Units returned as lowercase 'h' not 'HOURS' - hours now correctly converted to days (÷8)
fix container-deps Preserves Effort Values
- serialize_task() used else if for effort, so effort was only written if duration was absent
- Changed to separate if statements so both duration and effort are preserved
Mermaid Renderer Hierarchical Task Names
- Nested tasks showed full path IDs instead of names
- Fix: Extract leaf ID using rsplit('.') before lookup

[Unreleased - 2026-01-08]

Added

RFC-0003: Deterministic Resource Leveling
- Phase 1 implementation in crates/utf8proj-solver/src/leveling.rs
- LevelingOptions, LevelingStrategy, LevelingReason, LevelingMetrics types
- Deterministic conflict sorting: (resource_id, start_date), task_id tie-breaker
- LevelingResult preserves original_schedule for audit trail
- L001-L004 diagnostic codes
- CLI --max-delay-factor option
- 20 leveling tests
README Rewrite
- Updated tagline to "Explainable Project Scheduling"
- Fixed broken badges and documentation links
- Added "Why utf8proj?" comparison table
- Linked to EXPLAINABILITY.md manifesto

[Unreleased - 2026-01-07]

Added

Calendar Diagnostics (C001-C023)
- CalendarImpact struct tracking working days vs calendar days per task
- 23 calendar-specific diagnostic codes
- filter_task_diagnostics() for diagnostic→task linking
- Excel export with Calendar Analysis and Diagnostics sheets
- LSP hover with CalendarImpact display

[Unreleased - Earlier]

Added

RFC-0004: Progressive Resource Refinement Grammar
- resource_profile declaration with specializes, skills, traits, rate ranges
- trait declaration with description and rate_multiplier
- Rate range block syntax (min/max/currency)
- Quantified assignment syntax: assign: developer * 2
- Extended resource_ref: percentage (@50%) and quantity (*N)
PMI-Compliant Effort Scheduling
- Fixed effort-to-duration calculation: Duration = Effort / Total_Resource_Units
- Task::assign_with_units() for partial allocations
- See docs/SCHEDULING_ANALYSIS.md
BDD Conflict Analysis (crates/utf8proj-solver/src/bdd.rs)
- BddConflictAnalyzer using Biodivine library
- Encodes resource conflicts as Boolean satisfiability
- 5 tests
CLI Enhancements
- -l/--leveling flag for schedule command
- bdd-benchmark subcommand
WASM Playground (crates/utf8proj-wasm/, playground/)
- Playground struct with WASM bindings
- Monaco editor with custom syntax highlighting
- Real-time validation, live Gantt preview
- Share functionality, theme toggle
Interactive Gantt Chart (crates/utf8proj-render/src/gantt.rs)
- Standalone HTML with embedded SVG
- Dependency arrows, tooltips, zoom controls
- Light and dark themes
Resource Leveling (crates/utf8proj-solver/src/leveling.rs)
- ResourceTimeline, detect_overallocations, level_resources
- 12 integration tests
MermaidJS Renderer (crates/utf8proj-render/src/mermaid.rs)
- Critical path markers, milestone detection, dependency syntax
- 12 tests
PlantUML Renderer (crates/utf8proj-render/src/plantuml.rs)
- Critical path coloring, dependency syntax, milestone markers
- 17 tests
Excel Costing Report (crates/utf8proj-render/src/excel.rs)
- XLSX files with rust_xlsxwriter
- Formula-driven scheduling with VLOOKUP dependencies
- 13 tests
Extended Native DSL Grammar
- Project: timezone: attribute
- Resources: email:, role:, leave: attributes
- Tasks: note:, tag:, cost:, payment: attributes
- Milestones: Dedicated milestone id "name" { } syntax
- Reports, constraints, holidays (single-date support)
Tutorial & Benchmark Documentation (docs/)
- tutorial.md - CRM migration example
- benchmark-report.md - TaskJuggler comparison

Historical Bug Fixes

TJP Parser: Choice Rule Unwrapping (2026-01-04)

Location: crates/utf8proj-parser/src/tjp/mod.rs

Problem: The project_attr and resource_attr choice rules in the pest grammar were not being unwrapped before matching. Attributes like currency and efficiency were silently ignored.

Root Cause: Pest returns choice rules as their container type (e.g., project_attr) rather than the matched alternative (e.g., currency_attr).

Fix: Added unwrapping logic:

let actual_attr = if attr.as_rule() == Rule::project_attr {
    attr.into_inner().next().unwrap()
} else {
    attr
};

Tests Added: parse_project_currency, parse_resource_efficiency, parse_dependency_onstart, parse_dependency_onend

CPM Backward Pass for Non-FS Dependencies (2026-01-05)

Location: crates/utf8proj-solver/src/lib.rs

Problem: The backward pass treated all dependencies as FS, computing LF(pred) = min(LS(succ)) regardless of dependency type. This caused incorrect slack calculations for SS/FF/SF dependencies.

Fix: Enhanced backward pass to look up the actual dependency type and apply the correct formula:

FS: LF(pred) <= LS(succ) - lag
SS: LF(pred) <= LS(succ) - lag + duration(pred)
FF: LF(pred) <= LF(succ) - lag
SF: LF(pred) <= LF(succ) - lag + duration(pred)

Tests Added: 5 tests in crates/utf8proj-solver/tests/cpm_correctness.rs

Impact: Correct slack calculations for all dependency types.

FilesExpand file tree

CHANGELOG.md

Latest commit

History

CHANGELOG.md

File metadata and controls

Changelog

[0.17.0] - 2026-02-09

Added

[0.16.1] - 2026-02-09

Added

Fixed

[0.12.0] - 2026-01-23

Added

Performance

[0.11.2] - 2026-01-22

Changed

Performance

[0.11.1] - 2026-01-21

Fixed

Added

Performance

[0.11.0] - 2026-01-20

Added

New Diagnostics

Example Usage

Baseline File Format

[0.10.0] - 2026-01-18

Added

Changed

Migration Notes (0.9.x → 0.10.0)

[0.9.1] - 2026-01-16

Fixed

[0.9.0] - 2026-01-15

Added

[0.2.0] - 2026-01-14

Added

Fixed

[Unreleased - 2026-01-10]

Fixed

[Unreleased - 2026-01-09]

Added

Fixed

[Unreleased - 2026-01-08]

Added

[Unreleased - 2026-01-07]

Added

[Unreleased - Earlier]

Added

Historical Bug Fixes

TJP Parser: Choice Rule Unwrapping (2026-01-04)

CPM Backward Pass for Non-FS Dependencies (2026-01-05)