README.md
In zuyaml: Parse and Emit 'YAML' 1.2

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.

install.packages("zuyaml")

Or the development version from GitHub:

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

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.

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.

| Workflow | Jobs, and what they do | When | Runs | |---|---|---|---| | R-CMD-check.yaml | R CMD check --as-cran on macOS, Windows and Ubuntu at R release and oldrel-1, and in the CRAN-like containers clang23, ubuntu-clang and ubuntu-gcc16 — the package vendors C11, so the compiler CRAN uses is the one that has to see it | push, PR | view | | hardening.yaml | full-test-suite — all 402 yaml-test-suite cases, then tools/conformance.R against its pinned baseline. fuzz — the wrapper fuzzer at 20,000 iterations | push, PR, weekly | view | | native-checks.yaml | sanitizers — UBSan on a runner, ASan in the clang-asan and gcc-asan images. valgrind, lto, rchk (failing on findings), and gctorture at step 500 on every push with the sensitive step-20 run weekly | push, PR, weekly | view | | coverage.yaml | covr, with a per-file breakdown in the job summary | push, PR | view |

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.