fzc/fzr: pre-evaluate formulas from inline variable defaults

[yannrichet-asnr]

Problem

Given an input file that declares a variable default and a formula that uses it:

x = $(x~3)
y = @{x * 2}

fzi already pre-evaluated both ({'x': 3, 'x * 2': 6}), but fzc / fzr only
substituted the variable and left the formula uncompiled:

x = 3
y = @{x * 2}          # should have been "y = 6"

Cause

compile_to_result_directories (used by both fzc and fzr) passed only the
caller-supplied input_variables to replace_variables_in_content and
evaluate_formulas, and never looked at the $(var~default) defaults embedded
in the file. replace_variables_in_content recovers defaults on its own (so the
bare variable was substituted), but evaluate_formulas failed with
name 'x' is not defined and left the formula verbatim.

Fix

• Factor fzi's inline-default extraction into a shared helper
  fz.interpreter.parse_variable_defaults_from_content(content, varprefix, delim).
• fzi now uses it (behaviour unchanged; ~34 inline lines removed).
• compile_to_result_directories seeds the compile context with those defaults
  ({**inline_defaults, **var_combo}), so formulas depending only on defaulted
  variables are evaluated during compilation, matching fzi.
• Explicitly passed input_variables still win; list/bounds defaults that don't
  resolve to a usable scalar are skipped.

Tests

New tests/test_fzc_defaults.py (11 tests):

• reference guard that fzi pre-evaluates default + formula;
• fzc: pre-eval from default, parity with fzi, explicit override, formula
  left unevaluated when a needed variable has no default, default still used with
  partial vars, float default;
• fzr: end-to-end pre-eval (checks the compiled file in results/ and the
  parsed output), explicit override over a 2-point grid;
• unit tests for parse_variable_defaults_from_content (int/float/bare & quoted
  string/scientific notation/valid list literal/truncated [ -> None/;
  metadata ignored/no default/empty delim).

NEWS.md updated under ## Unreleased.

The unrelated tests/test_demos.py::TestDisplayResultsTmp::test_get_analysis_tmp_is_called
failure seen locally is a broken environment (NumPy 2.4.6 vs matplotlib ABI,
_ARRAY_API not found) and fails identically on a clean tree.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UCiJZSajaqyKYJ9sNoE6fn

[Copilot]

🟡 Changes recommended

Inline defaults that represent bounds/discrete-values metadata (e.g. [0,1], {0,0.1}) can currently be seeded into formula evaluation as real values, risking incorrect pre-evaluation instead of leaving formulas unevaluated.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Aligns compilation (fzc/fzr) with inspection (fzi) by pre-evaluating formulas that depend only on inline variable defaults ($(x~3)), so compiled inputs no longer retain unevaluated formulas like @{x * 2} when all required values are available from defaults.

Changes:

• Added shared helper parse_variable_defaults_from_content() and refactored fzi() to use it.
• Updated compile_to_result_directories() (used by fzc/fzr) to seed the formula/replace context with inline defaults so compilation can pre-evaluate those formulas.
• Added regression/unit tests and documented the behavior change in NEWS.md.

File summaries

File	Description
fz/interpreter.py	Adds parse_variable_defaults_from_content() to extract inline defaults for reuse across code paths.
fz/core.py	Refactors fzi() to use the new shared inline-defaults helper.
fz/helpers.py	Seeds compilation variable context with inline defaults to enable formula pre-evaluation during fzc/fzr.
tests/test_fzc_defaults.py	Adds regression/unit coverage for fzc/fzr pre-evaluation and the new helper.
NEWS.md	Documents the new fzc/fzr behavior under Unreleased.

Review details

• Files reviewed: 5/5 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.