arl 0.1.4

Performance

Reduce CRAN test runtime via uniform test thinning: on CRAN, run every 3rd test block per file (controlled by make_cran_thinner(keep_every = 3L)), cutting ~67% of per-test work while keeping full coverage off CRAN.
Reduce TCO test recursion depth on CRAN (6000 vs 10000) to further cut runtime while still validating tail-call optimization well beyond R’s default stack limit.

Remove “for R” from package title per CRAN policy.
Replace bare cat() calls with S3 print methods (arl_cli_output, arl_help_doc, arl_coverage_report) so console output follows CRAN guidelines.
Require explicit output_file argument in report_html() and report_json() — no longer write to the working directory by default.
Convert example vignette code chunks from {r} read-and-cat wrappers to {arl} chunks that execute via the knitr engine.

Pre-build vignettes in the tarball using the .Rmd.orig + stub pattern, cutting ~110s from R CMD check vignette rebuild time.
Speed up test suite by avoiding stdlib prelude loading in tests that don’t need it.
Reduce TCO test recursion depth to reduce runtime while still checking deeper recursion than R’s defaults allow.

Fix multiple temporary file and connection leaks uncovered by stricter CRAN-like checking.
make install now ships pre-built vignettes (previously installed without them).

make check now mirrors CRAN settings more closely and prints wall-clock time for the R CMD check step.
CI fetches gh-pages branch so benchmark data is available during vignette rendering.

Fixed stale module cache poisoning when ambient macros differ between sessions. The module cache now computes a fingerprint of all macros visible in a module’s parent environment chain (not just the prelude) and invalidates entries when macros change.
Fixed missing dependency declaration in looping.arl.
Fixed Windows backslash path handling in tests.
Fixed FileDeps tests to work during R CMD check.
Fixed bad URLs in README and vignettes.
Included vignette files in built tarballs (.Rbuildignore fix).

Initial release of arl, a Scheme-inspired Lisp dialect embedded in R.

Special forms: quote, if, define, set!, lambda, begin, defmacro, quasiquote/unquote/unquote-splicing, and, or, while, delay, import, and module.
Truthiness: #f, #nil, and 0 are falsy; everything else is truthy.
Cons cells: First-class dotted pairs, with both cons-cell and R-native list representations.
Destructuring: Pattern destructuring in define, set!, let, and lambda, including nested patterns, rest parameters, and defaults.
Multiple values: values and call-with-values.
Continuations: call-cc via R’s native callCC.

Compile-time code transformation with quasiquotation templates.
Hygiene via gensym and capture.
Introspection with macro?, macroexpand, macroexpand-1, and macroexpand-all.
Documentation attachment via ;;' comments.

Single-pass compilation from Arl to R expressions.
Self-tail-call optimization: rewrites self-recursive functions in tail position as while loops.
Constant folding, dead code elimination, strength reduction, identity lambda elimination, boolean flattening, and arithmetic infix compilation.

22 modules (~800 definitions) auto-loaded via a prelude:

Core and logic: when, unless, cond, error handling, identity.
Types and equality: Full suite of type predicates; equal?, eqv?, eq?.
List and sequences: append, reverse, take, drop, nth, range, repeat, cycle, zip, and more.
Functional: map, filter, reduce, fold, compose, partial, juxt, complement, constantly.
Control flow: try-catch, finally, condition signaling and handling.
Binding forms: let, let*, letrec with destructuring.
Threading macros: ->, ->>, as->, some->, some->>, cond->, cond->>.
Looping: until, do-list, Clojure-style loop/recur.
Math: inc, dec, abs, sqrt, trigonometric functions, and more.
Strings: str, string-concat, string-split, and friends.
I/O: File reading/writing, display, and formatting.
Data structures: Hash-backed dict and set types with full operation suites; defstruct macro for S3-backed record types.
Sorting: list-sort, sort-by, merge-sorted, stable-sort.
Assertions: assert-equal, assert-true, assert-false, and others.

Direct access to all R base and default-package functions.
Keyword arguments (:name value), formulas, $, [, [[, and @.
r-eval and r-call for dynamic R evaluation.
Bidirectional data exchange: define R objects into Arl, extract results back to R.
Configurable R package visibility.

REPL: Interactive prompt with multi-line input, readline history, bracketed paste, and error recovery.
CLI: Script execution, --eval expressions, quiet mode, bare engine mode, and stdin support.
Help system: (help topic) dispatches to special forms, builtins, stdlib docstrings, or R help. doc and doc! for programmatic access.
Coverage tracking: Line-level and branch-level instrumentation with console, HTML, and JSON reports.
Knitr engine: {arl} code chunks in R Markdown with persistent engine state and syntax highlighting.

Vignettes: getting started, standard library reference, compiler and internals, and troubleshooting.