Parse and Emit 'YAML' 1.2

Converts between 'YAML' 1.2 (< https://yaml.org/spec/1.2.2/>) and ordinary R objects using a bundled copy of the 'cyaml' C11 parser and emitter (< https://github.com/andrewmd5/cyaml>), so there is no system dependency and no runtime dependency beyond R itself. Ambiguous 'YAML' features are handled strictly and predictably: duplicate keys are refused by default, the 'YAML' 1.2 core schema is followed so that yes and no resolve as strings, and integers beyond double precision are preserved rather than silently rounded. A stream of documents and a sequence are different things, and the interface keeps them apart. Input size, nesting depth and the number of values materialised are all bounded, which makes the parser usable on untrusted input.

zuyaml

zuyaml converts between YAML 1.2 and ordinary R objects, using a bundled copy of the cyaml C11 parser and emitter, so there is no system dependency and no runtime dependency beyond R itself. Ambiguous YAML is handled strictly: the 1.2 core schema (so yes is a string), duplicate keys refused, large integers preserved, and a document stream kept apart from a sequence.

Installation

install.packages("zuyaml")

Or the development version from GitHub:

# install.packages("pak")
pak::pak("pedrobtz/zuyaml")

Usage

yaml_parse() turns YAML text into R objects:

library(zuyaml)

yaml_parse("host: localhost\nport: 8080\ntls: true\n")
#> $host
#> [1] "localhost"
#> 
#> $port
#> [1] 8080
#> 
#> $tls
#> [1] TRUE

yaml_emit() goes the other way. Note the quoting: "42" is a string in R, so it stays a string in YAML and on the way back. The same holds for the cases that are easy to get wrong — line breaks, leading and trailing whitespace, mapping keys, whole-numbered doubles — so what you emit parses back to what you had.

cat(yaml_emit(list(version = "42", ratio = 0.5)))
#> version: "42"
#> ratio: 0.5

Both have an _all variant for a YAML stream — several documents separated by ---, which is a different thing from a sequence — and yaml_read() / yaml_write() do the same for files:

yaml_parse_all("a: 1\n---\na: 2\n")
#> [[1]]
#> [[1]]$a
#> [1] 1
#> 
#> 
#> [[2]]
#> [[2]]$a
#> [1] 2

The getting started article documents the conversion rules in full, including the cases that do not round-trip.

Testing

• tests/testthat/ runs a curated 84-case yaml-test-suite subset in every R CMD check; ZUYAML_TEST_SUITE points it at a full 402-case checkout instead.
• tools/conformance.R walks the same corpus but reports rather than asserts — counts per boundary, failures named, against a baseline pinned in the script.
• The split is deliberate: a test cannot hold a number like 278/279 that is meant to change deliberately rather than break a build.

R-CMD-check.yaml, native-checks.yaml and coverage.yaml call the reusable workflows in r-actions rather than keeping a second copy here. hardening.yaml does not: it holds the two checks that are specific to this package.