Experimental R foundations for a planned QuPath interface. This first source
milestone contains classified conditions and bilingual messages, protocol JSON
helpers, a JSON Schema subset validator, synthetic contract fixtures and
qp_run_external(), a pinned external-method execution boundary.
The live Java bridge, interactive application, project editing, queued jobs and shareable workflow engine are not included in this milestone. The protocol schemas describe future interfaces; their presence does not implement them. No release or CRAN acceptance is claimed.
R >= 4.5.0 is declared. The current candidate was checked locally on Linux with R 4.6.1; the complete cross-platform CI matrix is a separate qualification. Java and QuPath are not needed to install or load these foundations. Optional packages listed in Suggests support tests or planned features and are loaded only where used. No runtime or model is downloaded automatically.
Install the development package and its vignette from GitHub:
install.packages("remotes")
remotes::install_github("CTTIR/qupflowR", build_vignettes = TRUE)From a checked-out source tree, use R CMD build ., then
R CMD INSTALL qupflowR_0.0.0.9000.tar.gz.
This example uses only the implemented R foundations and synthetic values:
library(qupflowR)
payload <- list(image_id = qp_scalar("synthetic-image"), object_ids = c("1", "2"))
json <- qp_json_encode(payload)
qp_json_decode(json)$object_ids
#> [1] "1" "2"
qp_value_state(c(1.5, NA_real_, NaN, Inf, -Inf))
#> [1] "finite" "missing" "nan" "pos_inf" "neg_inf"Mark scalar JSON fields with qp_scalar(); unmarked vectors remain arrays,
including vectors of length one. IDs remain character values. These helpers
encode protocol data; they do not connect to QuPath or modify a project.
The getting-started vignette covers conditions, messages,
JSON, schema validation and a runnable synthetic external-method example. After
installing with vignettes, open vignette("qupflowR", package = "qupflowR").
Use help(package = "qupflowR") for complete argument and return-value contracts,
and citation("qupflowR") for the installed package's citation.
Report reproducible problems with synthetic inputs at
https://github.com/CTTIR/qupflowR/issues. Contributor conventions are in the
CTTIR contributing guide.
The local documentation site uses themakR and builds into site/; a local
build does not establish that the public website is deployed.
qp_run_external() requires explicit trust, an executable/argument vector,
SHA-256 pins for declared executable/script/runtime/model/input files and a new
workspace. It uses a private child environment, checks inputs again after the
process, retains logs and a receipt, and hashes required outputs. Receipts may
contain private paths or caller environment values and belong with private run
artifacts, not in public repositories.
The external-execution profile is Linux-only. It is a hygiene boundary for
trusted code, not a sandbox. Callers must declare transitive dependencies and
validate analytical outputs themselves. execution_complete means process
success and declared output presence, not biological validity or equivalence.
Scripts, model versions, detector settings and measurement definitions are not
rewritten. An older QuPath command-line method can therefore be executed
independently of the planned QuPath 0.7/Java 25 live bridge.
Run testthat::test_local() from R. Generated native-image tests require a
separate explicitly configured fixture tree; their skips do not qualify native
QuPath execution. The synthetic schema/geometry oracles are included. The Java bridge scaffold and synthetic fixture generator are included as
development sources under tools/bridge. The 518 bridge unit tests pass; this
does not qualify native integration or live QuPath execution.
The retained quality gate supports lint, API and coverage checks:
QUPFLOWR_CTTIR_GITHUB= Rscript tools/quality.R --mode all --reportAn empty QUPFLOWR_CTTIR_GITHUB selects the hash-pinned vendored house linters;
no local sibling checkout is required. The committed API history and coverage
baseline are retained unchanged. Shared CI workflow callers are pinned to their
existing source revision. Hosted CI success, browser behavior, native runtime
qualification and second-computer reproduction are separate evidence states.
GPL (>= 3), with the full text in LICENSE.md. Package code is attributed to
the maintainer declared in DESCRIPTION. External engines and dependencies are
not bundled and retain their own licenses. MIT consumers should use the agreed
external process/file interface rather than silently importing GPL code.