---
title: "Configuration as data"
author: "R. Heller"
output: rmarkdown::html_vignette
vignette: >
%\VignetteIndexEntry{Configuration as data}
%\VignetteEngine{knitr::rmarkdown}
%\VignetteEncoding{UTF-8}
resource_files:
- figures/logo.svg
- figures/logo-dark.svg
---
```{r setup, include=FALSE}
library(gatekeepR)
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
```
Configuration is explicit and serializable. The example configuration is useful
for smoke tests and for learning the object shape.
```{r config}
cfg <- gk_example_config()
gk_validate_config(cfg)
gk_config_hash(cfg)
```
For a project configuration, declare the panel, signal policies, phenotype
hierarchy, and minimum support rules with `gk_panel()`, `gk_signal_policy()`,
`gk_hierarchy()`, and `gk_min_support()`. Validate it before preparing data:
```{r config-template, eval=FALSE}
cfg <- gk_config(
panel = gk_panel(c("marker_a", "marker_b")),
hierarchy = gk_hierarchy(...),
signal_policy = gk_signal_policy(...),
min_support = gk_min_support(...)
)
gk_validate_config(cfg)
gk_write_config(cfg, "config.json")
```
The configuration hash is recorded with prepared objects, reviews, and exports.
Changing a rule therefore creates a new computational identity instead of
silently changing an existing review.
