fz 2.0.0

Aligned with the funz-fz 2.0 release (Funz/fz, 2026-11-08). The R functions keep their signatures; the changes below come from the Python side and are visible from R. See the funz-fz NEWS.md (“Upgrading from 1.2”) for the full list.

Behaviour changes in funz-fz 2.0 that affect R code

• Python >= 3.9 is required (SystemRequirements); Python 3.8 is no longer supported by funz-fz.
• fzr(timeout = 0) (and FZ_RUN_TIMEOUT=0) now means “no timeout”; it used to time every case out immediately. A negative timeout is an error. ssh:// and slurm:// calculators have no default timeout unless FZ_RUN_TIMEOUT is set; sh:// and funz:// keep the 1 hour default.
• A model without delim (nor var_delim) reads both $(x) and ${x} as variables. Models that contain other ${...} text (shell snippets) should set delim; all examples in this package already set delim = "{}".
• cache:// uses a new SHA-256 cache key: caches written by funz-fz <= 1.2 are ignored unless FZ_CACHE_ACCEPT_LEGACY=1.
• sh:// commands resolve a file name to the launch directory only if the file is absent from the case directory. With 1.2, a command such as sh://cat in.txt > out.txt could read the un-substituted template: results obtained that way should be re-checked.
• Case directory names (case_naming = "path") percent-encode unsafe characters (a/b becomes a%2Fb); fzo() decodes them.
• fzr() now writes a manifest.json (campaign traceability) in results_dir, recording among others the funz-fz version.

Python version synchronisation

• The package now expects the funz-fz release with its own major and minor version (funz-fz>=2.0,<2.1 for this release). It declares it with reticulate::py_require(), so reticulate (now >= 1.41) provisions it on first use when no Python is configured. This downloads Python and funz-fz.
• fz_install() installs that version by default.
• fz_available() detects another funz-fz version in an existing Python environment: it offers to install the matching one in an interactive session, and warns otherwise. Set options(fz.sync_python = FALSE) to disable this.
• The error raised when Python is missing no longer suggests installing the unrelated PyPI package fz.

Tests

• tests/testthat/test-funz-fz-2.R checks timeout = 0, the rejection of a negative timeout and the manifest.json written by fzr(). It is skipped when the installed funz-fz is older than 2.0.

Documentation

• fzr()’s timeout documents the 2.0 resolution rules.

fz 1.2.0

Aligned with the funz-fz 1.2 release (Funz/fz, 2026-09-04).

The win-builder incoming pre-tests for this version flagged two problems, both fixed here (see “Fixes” below): the --run-donttest AttributeError: module 'fz' has no attribute 'fzc' and the invalid LICENSE.md file URI in README.md.

New arguments tracking funz-fz 1.2

• fzr() gains case_naming ("path" default, "hash", or "index") to name each case’s result/temp subdirectory. "hash"/"index" sidestep filesystem name-length limits when a study has many variables and write a cases.csv manifest at the results root.
• fzr(), fzc(), fzi() and fzd() gain input_static: a character vector of files identical across every case (a shared reference dataset, a weather CSV, …). They are never templated or scanned for variables, never re-hashed per case, and — for relative paths — symlinked into each case directory instead of duplicated (and transferred to ssh:///slurm:/// funz:// calculators).
• fzd()’s output_expression now also accepts a character vector of length > 1 for multi-objective algorithms (e.g. NSGA-II): each case yields one scalar per expression. Vector-valued outputs can be reduced to a scalar objective with mean(), sum(), len(), median(), stdev(), variance(), indexing/slicing, and zip().
• fzr()’s timeout default (NULL) now documents the funz-fz 1.2 resolution order: the model’s own "timeout" entry, then FZ_RUN_TIMEOUT (raised to 1 hour in 1.2). An explicit timeout still wins.

Runnable example

• Added a self-contained external-simulator example under inst/examples/perfectgas/ (perfectgas.txt template + perfectgas.py ideal-gas solver). The README and vignette now show a complete fzr() parametric run driven by a real external program (calculators = "sh://python3 perfectgas.py", input_static = "perfectgas.py"), instead of a placeholder run.sh. tests/testthat/test-perfectgas-example.R runs it end to end.

Documented model features from funz-fz 1.2

• fzo() documents the shell-free output extraction prefixes: python://, jq://, yq://, xpath:// (and the explicit bash://). These are portable on Windows without a bash install and preserve list results as vectors; a plain shell command still simplifies a single-element result to a scalar.
• fzo()/fzr() outputs may now resolve to a vector (time series, spectrum, per-node profile), stored per case unmodified.

Direct R function models in fzd()

• fzd() accepts an R function as model, letting adaptive design-of-experiments algorithms drive an R function directly instead of a file-based model (input_path = NULL, output_expression optional). This is available in funz-fz >= 1.2; earlier releases do not support callable models.
• R closures are bridged in via reticulate, which is only safe to call from the main thread, so the wrapper always forces calculators = 1 (strictly sequential) for function models regardless of the value passed in (Funz/fz#73). A value other than 1 emits a warning.
• fz_install() gains a packages argument (default "funz-fz") so a development build can be installed from GitHub with fz_install(packages = "git+https://github.com/Funz/fz.git").

Fixes

• fz_available() now returns TRUE only when the imported fz module actually exposes the funz-fz API (fzi/fzc/fzo/fzr), instead of merely checking that some module named fz is importable. On a machine where an unrelated PyPI package is importable as fz (as on the CRAN check farm), the if (fz_available()) guards now skip cleanly rather than letting examples fail with AttributeError: module 'fz' has no attribute 'fzc'.
• Fixed an invalid file URI (LICENSE.md) in README.md flagged by R CMD check --as-cran; it now links to the file on GitHub.
• Fixed fzd()’s own algorithm_options example/documentation, which used a "key=val;key2=val2" string that funz-fz has never actually accepted (only a named list, JSON string, or path to a JSON file) — replaced with a named list.
• Restored the LICENSE file (the templated file R’s packaging convention requires alongside License: BSD_3_clause + file LICENSE), which had been mistakenly deleted, and fixed R-CMD-check CI to install funz-fz and point reticulate at it via RETICULATE_PYTHON — without this, reticulate’s automatic environment provisioning was silently resolving to an unrelated PyPI package literally named fz (not funz-fz).

fz 1.1

First release, aligned with funz-fz 1.1 on PyPI.

Core functions

• fzi(input_path, model) — parse variable names and defaults from a template file
• fzc(input_path, input_variables, model, output_dir) — compile template by substituting variable values
• fzo(output_path, model) — read and parse output files
• fzr(input_path, input_variables, model, ...) — run full parametric study
• fzl(models, calculators, check) — list installed models and calculators
• fzd(input_path, input_variables, model, output_expression, algorithm, ...) — iterative algorithm-driven design of experiments

Model and algorithm management

• install_model(source, global) / install_algorithm(source, global) — install from GitHub, URL, or local zip
• uninstall_model(model_name, global) / uninstall_algorithm(algorithm_name, global) — remove installed items
• list_installed_models(global) / list_installed_algorithms(global) — list what is installed
• list_models() — alias for list_installed_models()
• install() / uninstall() — generic aliases for model install/uninstall

Configuration

• get_interpreter() / set_interpreter(interpreter) — get or set the formula interpreter ("python" or "R")
• get_log_level() / set_log_level(level) — control logging verbosity
• get_config() / print_config() / reload_config() — inspect and reload FZ_* environment variable settings

Package helpers

• fz_install() — install the funz-fz Python package via reticulate
• fz_available() — check whether the Python package is importable