The pre-built metacheck Docker image

ghcr.io/scienceverse/metacheck_r_large has the ~750 most common packages found across a corpus scan of real papers' R code already installed, plus JAGS, a JDK, and the GDAL/GEOS/PROJ geospatial stack (see https://github.com/scienceverse/metacheck_docker_reproducibility for the Dockerfile/package list/build instructions – that repo also publishes a smaller metacheck_r_small variant, ~390 packages, seen in at least 3 corpus files, also with JAGS/a JDK but without the geospatial stack), so a run against it needs no install for most papers. (With install_missing = TRUE every declared package is still installed.) This is the DEFAULT image (see .repro_docker_image_for()'s use_declared_version argument) – pinned to whatever R version it was built against, NOT the paper's own declared version, which is the tradeoff docker_use_declared_version on reproducibility_check() exists to let a caller opt out of.

Usage

.repro_docker_default_image

Details

History: this constant pointed at a single, un-split ghcr.io/scienceverse/metacheck_r image until 2026-10, when the _large/_small split described in the build repo's own README/ Dockerfiles was ACTUALLY built and pushed under those names for the first time (confirmed live via docker pull, not assumed from the README) – a prior attempt to point this constant at metacheck_r_large earlier that same session had to be reverted because the name did not yet exist on the registry at that time; it does now. If this constant is ever repointed again, confirm with docker pull first, not from this file's own README alone – see that README's own naming-status note for why that verification step matters here specifically.