Skip to content

CHANGELOG.md

Latest commit

 

History

History
164 lines (132 loc) · 10.7 KB

CHANGELOG.md

File metadata and controls

164 lines (132 loc) · 10.7 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

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

[0.0.7] - 2026-03-19

  • Transfer repository to Jameel Institute @jameel-institute oragnisation
  • Added small script docs/sync_readme.jl to update package version in website index and Readme.md

[0.0.6] - 2026-03-16

Changed (Breaking)

  • daedalus() function signature refactored: second positional argument now takes infection parameters (infection) instead of scalar/vector r0. Users must pass parameters via InfectionData object rather than individual keyword arguments
    • Old interface: daedalus(country, r0::Float64; sigma=..., epsilon=..., eta=..., ...)
    • New interface: daedalus(country, infection; npi=..., log_rt=..., time_end=..., increment=..., n_threads=...)
  • Removed all infection-parameter keyword arguments: sigma, p_sigma, epsilon, rho, eta, omega, gamma_Ia, gamma_Is, gamma_H, nu
    • All epidemiological parameters now encapsulated in InfectionData object
    • Users customize parameters by fetching InfectionData and modifying fields before calling daedalus()

Added

  • New dispatch methods for daedalus():
    1. String pathogen name: daedalus(country, "sars-cov-2 delta"; ...)
    2. Single InfectionData: daedalus(country, infection_obj; ...)
    3. Vector InfectionData: daedalus(country, [inf1, inf2, ...]; ...)
  • extract_infection_params() helper function to extract and expand epidemiological parameters from InfectionData
  • InfectionData is mutable, allowing users to customize parameters post-fetch: inf = get_pathogen("sars-cov-2 delta"); inf.r0 = 2.5
  • Infection names are normalized to lowercase: all pathogen names are stored and looked up as lowercase strings (e.g., "sars-cov-2 delta", "influenza 2009")

Migration Guide

# Old (no longer works):
result = daedalus("Australia", 2.5, sigma=0.217, epsilon=0.58, time_end=200.0)

# New (string pathogen, lowercase names):
result = daedalus("Australia", "sars-cov-2 delta", time_end=200.0)

# New (custom infection):
infection = Daedalus.DataLoader.get_pathogen("sars-cov-2 delta")
infection.r0 = 2.5
result = daedalus("Australia", infection, time_end=200.0)

# New (vector of infections):
infections = [
    Daedalus.DataLoader.get_pathogen("sars-cov-2 delta"),
    Daedalus.DataLoader.get_pathogen("influenza 2009")
]
results = daedalus("Australia", infections, time_end=200.0)

[0.0.5] - 2026-03-09

Changed

  • daedalus() function signature: country is now the first positional argument and r0 is the second positional argument, enabling method dispatch on r0 type (scalar Float64 vs vector Vector{Float64}) across two separate implementations in src/Model.jl and src/Ensemble.jl
  • All documentation examples and benchmarks updated to reflect the new calling convention: daedalus(country, r0, ...) instead of previous keywords-first approach
  • All test files updated to use new positional argument signature for daedalus calls
  • Function calls now use positional arguments: daedalus("Australia", 2.5, time_end=200.0) instead of previous keyword-based calling conventions
  • Function get_ngm() requires transmission rate beta and not the $R_0$; function get_beta() is now vectorised over multiple values of $R_0$. The use case is generating multiple NGMs for ensemble runs without running get_beta() an equal number of times. get_ngm() has a method for a vector of beta().

Added

  • Multiple dispatch implementation for daedalus(): scalar and vector R0 inputs are now handled via distinct function methods
  • Vector R0 dispatch in file src/Ensemble.jl: daedalus(country, r0::Vector{Float64}; ...) runs multiple R0 values in a single call
  • Implementation of SciMLBase.EnsembleProblem in src/Model.jl (daedalus_internal function): uses EnsembleThreads() solver with prob_func wrapper to efficiently orchestrate multi-run ODE solving. The ensemble approach reuses a base ODE problem and remakes it for each trajectory with its corresponding parameters, enabling automatic thread-safe parallel execution across multiple r0 values without explicit locking
  • Helper functions prepare_shared_data() and daedalus_internal() exported from Model.jl for use by ensemble dispatch
  • Comprehensive function documentation improvements: added or enhanced docstrings for all exported functions in Helpers.jl, Data.jl, Events.jl, Model.jl, and Ode.jl with argument lists and return type annotations

[0.0.4] - 2026-03-02

Added

  • Documentation page docs/src/settings.md explaining the multiple-contact-settings feature: how to assign a Vector{Matrix{Float64}} to CountryData.contact_matrix, how contacts3d stacks them into a 3D array, and how beta calibration uses total_contacts (sum of all matrices)
  • Tests for multiple contact settings: get_settings count, contacts3d shape, total_contacts element-wise sum, model execution with two settings, and calibration-equivalence check (two equal settings produces the same epidemic as one setting for the same R0)

Fixed

  • Lowered Statistics and LinearAlgebra compat bounds to 1.10.0 (matching Julia 1.10 LTS stdlib versions) so the package resolves correctly on Julia 1.10 LTS

Changed

  • CountryData struct now accepts a Vector of contact matrices as Matrix{Float64} for multiple contact settings. Helper functions process this list, or a single Matrix, to give total contacts where needed including Helpers.get_beta and Helpers.get_ngm.
  • Moved away from using StaticArrays for contact matrices as operating on them was slower than using regular arrays.

Added

  • Docstring for Helpers.weighted_slice_sum! explaining the tensor contraction algorithm, arguments, and performance notes
  • Docstring for Data.total_contacts explaining the dispatch on single vs. vector-of-matrices input
  • Expanded docstring for Data.contacts3d explaining the 3D stacking, the K=1 reshape fallback, and the role of the third dimension in the ODE
  • Tests for Helpers.weighted_slice_sum! covering unit weights, slice selection, scalar scaling, zero weights, in-place overwrite, and agreement with a reference loop

[0.0.3] - 2026-02-27

Changed

  • daedalus now accepts country as either a String or a DataLoader.CountryData struct; a String is resolved to CountryData via DataLoader.get_country at the start of the function, making both call styles equivalent
  • Updated test/test_basic.jl and test/test_eigenvalue.jl to replace removed Data.australia_contacts() calls with DataLoader.get_country("Australia").contact_matrix; replaced zero-arg Data.prepare_contacts() call with Data.prepare_contacts("Australia")

[0.0.2] - 2026-02-25

Fixed

  • prepare_demog(cd::CountryData) now clamps worker counts to a minimum of 1, preventing division-by-zero when the result is used as a scaling denominator. 30 countries (Australia, Belgium, Brunei, Cambodia, Chile, China, Costa Rica, Cyprus, Estonia, Finland, Hong Kong, Iceland, Japan, Kazakhstan, Laos, Latvia, Luxembourg, Malaysia, Malta, Mexico, Morocco, Myanmar, New Zealand, Portugal, Romania, Rwanda, Singapore, Slovenia, Switzerland, Tunisia) had at least one sector with zero workers in the data, causing Inf in the scaled contact matrix, which propagated to NaN in the ODE step-size calculation and immediate solver exit with dt_NaN warnings.

Changed

  • daedalus now requires a country string directly instead of separate initial_state, contacts, and cw arguments; all tests, examples, and documentation updated accordingly
  • Simplified UK example in docs/src/index.md to use daedalus(country="United Kingdom", ...)
  • Updated docs/src/country_data.md to reflect that daedalus accepts country directly

Added

  • Documentation page for country and pathogen data (docs/src/country_data.md)
  • DataLoader module added to function reference autodocs
  • Basic model and helper tests extracted into test/test_basic.jl

Fixed

  • Added explicit du[end] = 0.0 in daedalus_ode! to prevent undefined Rt derivative between callback updates
  • Changed Ia * p.epsilon to Ia .* p.epsilon in ODE for consistency with broadcasting conventions
  • Removed cm_scaling .* p.contacts StaticArrays broadcasting failure; replaced with sum(p.contacts, dims=3)[:,:,1]

[0.0.1] - 2026-02-23

Added

  • Core epidemiological model mirroring the R package {daedalus}
  • ODE-based compartmental disease transmission model with demographic groups
  • Effective reproduction number (Rt) calculation and logging at specified timesteps
  • Next Generation Matrix (NGM) method for Rt calculation
  • Power iteration for faster Rt calculation
  • Flexible event handling system with both timed and reactive (state-dependent) events
  • Non-pharmaceutical intervention (NPI) modeling with timed and reactive triggers
  • Time dependent NPIs as a struct TimedNpi
  • Contact matrix support for modeling population mixing patterns
  • Option to toggle contact matrix scaling
  • Functions for dynamic parameter modification and reset during simulations
  • Support for both increasing and decreasing threshold detection in event callbacks
  • Worker contact modeling with static vector optimization
  • Documentation with basic usage examples and plots
  • Documentation workflow via GitHub Actions
  • Data layer (DataLoader) with lazy-loaded country, pathogen, economic contacts, closure strategy, and vaccination scenario data mirroring the R {daedalus} data package
  • Bundled CSV data files: country demographics, hospital capacity, sector GVA, sector contacts, economic closure strategies, and seven pathogen parameter sets
  • Output timeseries access function for structured post-simulation analysis

Changed

  • Simplified NPI handling and data structures
  • Improved event logic for better state-dependent triggering
  • Unified contact calculation using single contact matrix approach
  • Optimized dependency management for lighter package footprint
  • Updated ODE system to account for worker-specific transmission dynamics
  • Refined beta (transmission rate) calculation for arbitrary contact matrix sizes
  • Enhanced model interface to return NPIs and remove unused inputs
  • Applied JuliaFormatter across codebase for consistent style

Fixed

  • Corrected Rt calculation method
  • Fixed ODE formulation for accurate disease dynamics
  • Resolved issues with SavedValues type handling in callbacks
  • Fixed error when output timebin is an exact factor of tmax

Technical Details

  • Package version: 0.0.1
  • Julia compatibility: 1.6.7+
  • Key dependencies: OrdinaryDiffEq.jl, DiffEqCallbacks.jl, StaticArrays.jl, CSV.jl, DataFrames.jl