v0.2.0

[0.2.0] - 2026-09-24

Added

• Homebrew formula (Formula/excalc.rb), installable via brew tap dblock/excalc-rs https://github.com/dblock/excalc-rs && brew install excalc. Builds from source via cargo install. CI audits/installs/tests the formula on every push/PR via .github/workflows/homebrew.yml; RELEASING.md documents updating its url/sha256 on every release.

• Windows MSI installer, built with cargo-wix from wix/main.wxs (bundles excalc, calc, and excalc-mcp, with an optional PATH entry). .github/workflows/msi.yml builds it on every push/PR; .github/workflows/release.yml rebuilds it and attaches it to each published GitHub release.

• calc binary alias for excalc (same CLI, shorter name).

• docs/ reference documentation, per function category, marked ported (implemented) vs. planned (extracted from the original Pascal source as a porting reference).

• README ### Examples section covering every ported operator and function, verified against the evaluator by tests/readme_examples.rs so it can't silently drift out of sync.

• CONTRIBUTING.md with build/test/PR instructions, replacing the README's ## Development section.

• CI: test coverage reported to Coveralls via cargo llvm-cov, plus a CI status and coverage badge in README.

• Test suite brought to ~99% line coverage: unit tests for every reciprocal/hyperbolic trig function and error path, statistics error paths, evaluator error paths (unknown variable, factorial overflow, log/root domain errors), lexer edge cases (decimals, scientific notation), parser error paths, and a new tests/cli.rs exercising the compiled binary end-to-end. The 3 remaining uncovered lines are tan/sec/sech division-by-zero guards that are mathematically unreachable for any real input.

• excalc-mcp binary: an MCP server exposing a single evaluate tool over stdio (via rmcp), so AI agents that speak MCP can call the evaluator directly instead of shelling out to the CLI. Built and installed by default via the mcp Cargo feature (opt out with cargo install excalc --no-default-features for a CLI-only install). Covered by tests/mcp.rs, including a test that keeps the README's example expression in sync with the real evaluator. README documents one-liners to register it with GitHub Copilot CLI (copilot mcp add) and Claude Code (claude mcp add), plus a worked example combining several functions in one expression.

• General / rounding functions: abs, frac, intg (alias of trunc), round (round-half-up), trunc, ceil, floor, random(x) (uniform random number in [0, x], via the rand crate).

• Number theory functions: gcd, lcm, fib, prime?, moebius, mersenne, perfect, fermat, safeprime, primec, primen, mersennegen, mersgen, genmers, sigma, tau, phi/eind, backed by a deterministic Miller-Rabin primality test (valid for the full u64 range).

• Comparison and logical/bitwise operators: =, >, < (comparison, returning 1/0), xor, xnor, and (& synonym), nand, or, nor (bitwise, operating on operands truncated to i64), plus not(x), shl(x, y), shr(x, y). Precedence: comparison loosest, then logical-or-family, then logical-and-family, then the existing arithmetic chain. Deviates from the original: = now tests equality instead of assigning a variable, and ? (the original's equality test) was dropped since it's no longer needed.

• Advanced / special functions: gamma, beta, pochhammer, bth, bman, ellipticE, ellipticF (alias ellipticK), ellipticCE, ellipticCK, dilog, dawson, erf, erfc, si, ssi, ci, chi, fresnelC, fresnelS, fresnelF, fresnelG. Most of these are thin wrappers around a generic numeric-integration primitive in the original engine; this port routes them through the same shared adaptive-quadrature engine used by the general-purpose numeric integration functions below (gamma keeps its own fixed-step Riemann sum instead, to match its pre-existing worked example) — see docs/functions/advanced.md.

• Financial functions: pv, fv, pmt, npv, rate, cterm, term, sln, syd, ddb, db, and the extended payment-timing-aware variants irate, nper, paymt, fval, pval, ipaymt, ppaymt. Unlike the advanced/special functions, these are all closed-form (or, for irate, a literal secant-method root find) and don't depend on numeric integration.

• General-purpose numeric integration: named composite quadrature rules trapezoid (aliases trapez/trapezoide), simpson, newton, boole, ordersix (alias ordresix), weddle (each over a whole-number sub-interval count n), plus adaptive quadrature int (alias gauss, refining to a requested tolerance). These are the only functions in the port that take an unevaluated expression and a bare variable as arguments instead of plain numbers, evaluating the expression repeatedly at sample points; see docs/functions/numeric-integration.md. This closes out every function category originally planned for the port.

• Variables: name := expr assigns to a variable, visible to later statements in the same input, which can now be a ;- or newline-separated sequence of statements (assignments and/or plain expressions); the value of the last statement is the result. Assignment is a statement, not an expression, so it can't be nested or chained (x := y := 5), keeping it unambiguous with ='s equality meaning. pi/e remain reserved and can't be reassigned. Variables don't persist across separate calc invocations or MCP tool calls.

• Degree/radian support: deg(x)/rad(x) convert between radians and degrees, and sind/cosd/tand/asind/acosd/atand are degree-native variants of the six basic circular trig functions. Deliberately stateless/explicit rather than reviving the original's global CalcMode toggle — see docs/functions/standard-math.md and DESIGN.md#angle-units.

Fixed

• Removed dead/unreachable Token::Minus arm in parser.rs's parse_primary (unary minus is always handled earlier by parse_unary, so this branch could never execute).
• The original's Ci/Chi formulas reference an undefined variable G (presumably meant to be the Euler-Mascheroni constant), which the generic variable-lookup mechanism silently defaults to 0, making the original results meaningless. This port hardcodes the actual Euler-Mascheroni constant instead.
• docs/functions/advanced.md's placeholder example ellipticCK(1) → 5.99158934050168 was wrong; corrected to 1.57079632679490 (pi/2), matching ellipticCE(1) since both reduce to the same integral at that boundary.
• docs/functions/financial.md's draft named the fixed-declining-balance depreciation function fdb; corrected to db, matching the name actually dispatched in the original engine.