NEWS
greenAlgoR 0.2.0
ga_autometric_lastrun() reconstructs the last-run resource impact from an autometric log, keeping only the single longest run of each target still present in the targets store and drawing one duration bar per target colored by mean CPU usage.
ga_autometric_plot() renders the CPU and memory usage recorded by the autometric package as a faceted ggplot2 line plot over time, with one facet per metric and lines colored by logging phase, process or name.
ga_targets_meta_plot() draws one bar per target from targets::tar_meta(), mapping runtime to bar length and output size to fill and flagging targets that emitted a warning, summarizing the resource impact of the last pipeline run from its metadata.
ga_targets_network() visualizes a targets pipeline as a resource-annotated dependency network via ggnetwork, mapping per-target runtime and output size to node size and color and the downstream-target runtime to edge width.
greenAlgoR 0.1.2
ga_footprint() gains flexible carbon_intensity input: in addition to the existing data.frame format, the parameter now accepts a single numeric value (used directly as gCO2e/kWh regardless of location_code) or a named numeric vector (overrides/adds entries to the bundled database). This makes it easy to plug in custom carbon-intensity values from sources like the Electricity Maps API.
- Improve documentation
- Create two vignettes: "Introduction to greenAlgoR" and "Targets integration"
- Correct url to v3 of greenalgorithm and update internal data in sysdata.rda
greenAlgoR 0.1.1
- Remove the useless argument
fields from ga_targets()
greenAlgoR 0.1
- Initial github submission.
- Creation of function
ga_footprint() and ga_targets()