AutomataLab v4.0.0

Download the installer for your platform from the Assets below. Desktop auto-updates are supported from v1.0.2 onward.

Added

• Conversion & transformation utilities. A new CONVERT toolbar button opens a dedicated workspace that runs the classic constructions and plays them back step by step on a live diagram preview, then lets you open the result in a new tab:
  • NFA → DFA and ε-NFA → DFA — subset (powerset) construction.
  • ε-NFA → NFA — ε-transition elimination via epsilon-closures.
  • DFA minimization — partition refinement (Moore/Hopcroft), after completing the automaton and dropping unreachable states.
  • Regex → NFA — Thompson's construction from a regular expression (|, *, +, ?, grouping, ε).
  • CFG → PDA — the standard one-state nondeterministic pushdown construction from a context-free grammar.
• Step-by-step construction player. Each conversion is broken into ordered steps with a title and explanation; play/pause, scrub, or click any step to reveal exactly the states and transitions it adds (the new states/edges are highlighted), with a Source ⇄ Result toggle to compare against the input.
• Readable converted states with provenance. Subset construction and minimization no longer produce illegible nested-set labels like {{q3,q5},{q6}}. Converted states are named compactly (q0, q1, …), and the set / equivalence class they stand for is shown on hover, via a short ⇄ full labels toggle in the preview, and in the step text — so the diagram stays clean without losing the construction's meaning (ε-elimination states also show their ε-closure).
• Diagram image export (PNG / SVG). The Export dialog now exports the state diagram itself as a crisp vector SVG or a high-resolution PNG, theme-aware (light/dark), in addition to the existing data exports. Powered by a new dependency-free machine→SVG renderer.
• Clickable ε (epsilon) inserter. A small ε / λ button now appears wherever you type symbols — the transition editor, the δ-table (for ε-NFAs), and the Regex / CFG conversion inputs — so you can enter epsilon without an epsilon key on your keyboard.
• Classic desktop menu bar. A familiar top File / Edit / View / Simulate / Convert / Help menu bar now sits above a refreshed toolbar, so every action (new/open/save, copy/cut/paste, run/step/reset, conversions, batch testing, export, help) is reachable from a predictable place — not just from scattered toolbar icons.
• Keyboard parity for state roles & visible selection. Press I to mark the selected state as the start state and F to toggle it as an accept state; drag with Shift held to rubber-band-select several elements at once (Ctrl/Cmd-click still adds to a selection).

Changed

• The Export dialog gained a Diagram section (SVG / PNG) alongside the transition table, trace, tree, and machine-definition exports.
• Help now documents the conversion utilities, readable converted labels, the ε inserter, and diagram image export.
• Refreshed “classic desktop” look. The toolbar and menus adopt a native-application chrome (subtle bevels, compact icons, a menu bar) for a more familiar, less cluttered workspace; the simulation transport (play / step / step-back / reset / speed) and the canvas tool palette were restyled to match.
• Accessibility & UX pass. Modal dialogs (Help, Export, Convert, Batch) now trap keyboard focus, restore focus when they close, close on Esc, and announce themselves to screen readers. The right side panel remembers which tab you were on (defaulting to the δ transition table instead of an empty History tab) and can be collapsed to reclaim space. A blocked run now explains why — it surfaces the first validation problem and opens the Validate tab (with a count badge) instead of doing nothing. A halted run highlights the witness path on the diagram, and “stuck” is now clearly distinguished from “reject” in tooltips and styling. Epsilon typed as eps / λ / lambda (or left blank) normalizes to ε. The δ-table and nondeterministic history now render set-valued state collections as {q0, q1}.

Performance & robustness

• Nondeterministic runs can no longer exhaust memory. An NPDA distinguishes computation branches by their stack, so a grammar/machine with ε-moves that push different symbols could double the live branch set every step and run the app out of memory in seconds. The engine now caps the live frontier width: when a single step would blow past the limit it stops expanding and decides on what it has (accepting if a valid accepting branch exists, otherwise halting as stuck) — the verdict stays correct, the memory stays bounded.
• Malformed machine files fail gracefully. Opening a corrupt, truncated, or hand-edited .autolab.json (or a file whose name / fields have the wrong type) now reports a clean error instead of crashing; the loader only ever reads known fields (prototype-pollution safe).
• A bad step limit can’t brick a run. A zero, negative, or non-numeric step limit for a TM / PDA now falls back to the default instead of halting the machine on its first step.
• Property-based fuzzing. A new randomized test suite throws thousands of arbitrary machines of every type at the engines (must never crash and must always halt), cross-checks every conversion against its source machine for language equivalence, and verifies Regex → NFA against a reference regular-expression engine.

Fixed

• Transitions now follow a state while you drag it. Previously the connected edges stayed pinned to a state's old position until the drag ended; they now track the node continuously as it moves.

Notes

• Fully backward compatible: conversions never mutate the source machine — results open in a separate tab and use the existing .autolab.json shape, so nothing about saved files or existing machine behavior changes. Converted states carry an optional, additive description (their provenance); older files without it load unchanged.

---

Full changelog: https://github.com/reeshavsinha/AutomataLab/blob/main/CHANGELOG.md