NEWS
climatehealth 1.0.3 (2026-07-05)
- Updated wildfire v5 PM2.5 handling:
- treats joined wildfire PM2.5 values as already in micrograms, without the
legacy kg-to-microgram scaling step
- continues to normalise custom exposure columns to
mean_PM for downstream
wildfire calculations
- refreshes wildfire tests and documentation for the v5 data-source workflow
- Improved temperature-related mortality national-output handling:
- supports configurable country labels for pooled national outputs
- includes national attribution fixes used by the meta-analysis workflow
- updates temperature-mortality documentation for supported spline settings
- Improved CRAN release hygiene:
- removes test warnings from case-insensitive fixed-string matching and
tidyselect column selection
- hardens the CI R package cache setup used by package checks
climatehealth 1.0.2 (2026-06-09)
- Improved accessibility of saved plot outputs across the package:
- added shared helpers for accessible base R, ggplot2, and patchwork outputs
- applies consistent SOSCHI styling, accessible colours, figure titles, logos,
and wrapped alternative-text captions
- improves sizing, layout, and pagination for multi-panel indicator reports
- Integrated descriptive statistics more closely with the main analysis
workflows:
- added optional descriptive-statistics controls to the indicator
do_analysis() functions where relevant
- supports summary statistics, correlation outputs, moving-average plots,
missing-data summaries, and outlier checks alongside indicator outputs
- preserves run-scoped descriptive output folders and forwards the selected
correlation method through the indicator workflows
- Updated indicator workflows and outputs:
- air pollution now supports additional continuous covariates in regional GAMs,
normalises custom column names, ignores unusable covariates, handles
no-excess national power outputs, and returns API-safe results
- wildfire handling was improved for region/date ordering, region-level
outputs, PM2.5 scaling, VIF/QAIC checks, descriptive-statistics column
mapping, and missing population checks
- mental health and temperature-related workflows gained more reproducible
attribution options, clearer model-validation outputs, and refreshed
documentation
- malaria and diarrhea workflows gained optional descriptive statistics,
API-mode safeguards, and accessible output updates
- Improved API/runtime support:
- added an air pollution endpoint to the bundled plumber API
- added support for base64-encoded map zip uploads for malaria and diarrhea
API requests
- prevents API-mode calls from writing plot/CSV outputs or returning
non-JSON-safe fitted model objects
- Refreshed package documentation and presentation:
- added UNSC endorsement and CRAN availability information to the README
- added SOSCHI branding/logo assets for README and report outputs
- added a README notes column to flag the ongoing temperature-related health
effects review work
- updated indicator titles and documentation for consistency
- Improved package maintenance and CI reliability:
- expanded unit and integration coverage for accessibility, descriptive
statistics, API mode, and indicator workflows
- keeps heavy integration tests opt-in for coverage reporting and extends the
Codecov step timeout
- removed unused plotting helpers/dependencies and added
png for logo
rendering
climatehealth 1.0.1 (2026-04-08)
- CRAN resubmission focused on Fedora Linux check reliability and test stability.
- Fixed INLA model execution under check-core constraints:
- limits INLA thread usage when
_R_CHECK_LIMIT_CORES_ is set
- keeps compatibility across INLA versions by conditionally passing
thread-related arguments only when supported
- Hardened CRAN behavior for fragile INLA end-to-end disease tests:
- retained CRAN-safe skips for diarrhea and malaria integration paths
- avoids Fedora-specific parallel worker failures seen in prior checks
- Improved integration-test governance in CI:
- heavy integration tests are opt-in via
RUN_INTEGRATION=true
- standard PR/push CI remains fast and high-signal by default
- Build hygiene improvement:
- excludes
.covrignore from package build input via .Rbuildignore
climatehealth 1.0.0 (2026-03-30)
- Added new descriptive statistics public APIs:
run_descriptive_stats()
run_descriptive_stats_api()
- Added run-scoped output folders for descriptive stats:
<output_path>/descriptive_stats/<run_id>/All/
<output_path>/descriptive_stats/<run_id>/<Region>/
- Kept backward-compatible aliases with runtime deprecation warnings:
common_descriptive_stats_core()
common_descriptive_stats()
common_descriptive_stats_api()
- Improved descriptive stats robustness:
- correlation now validates and filters numeric columns explicitly
- centralized preflight column and parameter validation for enabled features
- typed summary behavior for non-numeric columns
- safer plot device cleanup with managed PDF helper
- Expanded tests for wrapper behavior, run-folder outputs, API payload validation,
endpoint-style contract responses, and deprecated alias compatibility.
- Moved API runtime assets previously held in
climatehealth_pipelines into the package:
- plumber routes and startup scripts under
inst/plumber/
- throttling modules and tests
- package-scoped API and developer config templates under
inst/extdata/config_templates/
- operational helper scripts under
inst/scripts/
- Improved package/API compatibility for runtime usage:
- enabled API mode in plumber runtime
- added direct endpoint bindings for descriptive statistics, malaria, and diarrhea
- enforced API-safe return payloads for malaria and diarrhea in API mode
- extended internals to accept API-style record payloads and map records
- updated throttle estimation to size JSON-record payloads instead of file paths
- Cleaned up the air pollution module:
- standardized argument and variable naming
- refreshed documentation
- resolved R CMD check notes for global variable bindings
- added backward-compatible support for legacy
air_pollution_do_analysis() argument names
- expanded air pollution regression and integration tests