News

Project news, release announcements, and design deep-dives from the AsciiChem team.

2026-07-21 — AsciiChem v0.18: Strict body validation — TODO.complete FINISHED

v0.18.0 completes the final item in TODO.complete (all 17 phases done). MechanismBuilder, SpectrumBuilder, and CalculationBuilder now validate body lines at parse time — malformed entries missing the : separator raise ParseError with line number and content, instead of being silently dropped.

This wraps up the architectural improvement arc that started with chemicalml 0.3.0 adoption and native CML wire migration. Every TODO.complete workstream is now done.

2026-07-21 — AsciiChem v0.15-v0.17: parser errors, Crystal SVG, Point3

Three releases finishing the TODO.complete workstream (16 of 17 phases done; Phase 16 strict-grammar deferred as poor ROI):

v0.17.0 — Atom::Point3 value object

Atom::Point3 (Struct with x/y/z, to_a, magnitude). New Atom#cartesian and Atom#fractional accessors bundle the flat coordinate fields. Backwards compatible — flat accessors remain.

v0.16.0 — Crystal 2D SVG projection

Node#to_structural_svg shortcut exposes the StructuralSvg formatter. For Crystal nodes, projects fractional coordinates onto the ab-plane with a unit cell outline and atoms placed at their projected positions.

v0.15.0 — Parser line/col error positions

Parser error messages now report line N, col M for multi-line inputs instead of the global char N offset. The caret pointer is positioned on the relevant line.

2026-07-21 — AsciiChem v0.12-v0.14: Convenience methods, perf, CLI JSON

Three releases continuing the architecture-quality push:

v0.14.0 — CLI JSON lint output

asciichem lint -f json now emits diagnostics as a JSON array for tooling integration (CI, editors, downstream scripts). Each entry has severity, message, and node fields.

v0.13.0 — PeriodicTable atomic masses + formula weight

PeriodicTable.atomic_mass(symbol) exposes IUPAC 2021 standard atomic weights for ~50 common elements. Molecule#formula_weight uses it to compute molecular weight directly — parse("C_6H_12O_6") .nodes.first.formula_weight returns 180.156.

v0.12.0 — Hill formula, linter sorting, parser perf

• Molecule#hill_formula — Hill-system canonical formula (C first, H second, others alphabetical)
• Linter.run sorts by severity then deduplicates
• Parser caches Grammar/Transform instances (~15% throughput gain)

2026-07-21 — AsciiChem v0.11: Native CML wire complete

v0.11.0 completes the native-wire migration arc for every beyond-formulas construct. ZMatrix, Calculation, Mechanism, and reaction conditions now join Crystal (v0.9) and Spectrum (v0.10) in using native CML wire representations.

Construct	v0.10 wire	v0.11 wire
ZMatrix	<aci:zmatrix> text	<zMatrix> inside <molecule>
Calculation	<aci:calculation> text	<propertyList> of <property>
Mechanism	<aci:mechanism> text	<mechanism> inside <reaction>
Conditions	aci:conditionsAbove attrs	<conditionList> of <scalar>

Other CML tools — Avogadro, Open Babel, Jmol, Mnova, ACD/Labs — can now read all AsciiChem output natively without needing the AsciiChem-specific aci: extension namespace. Backwards compat preserved: legacy aci: text-carrier XML still parses correctly.

The native-wire migration arc (Phases 01–06 of TODO.complete) is now complete. Future work focuses on structural representation (currently several constructs use text-form via title attribute; structural length/angle/torsion children for ZMatrix and step-by-step representation for Mechanism are future enhancements).

2026-07-21 — AsciiChem v0.10: Native CML wire for Spectrum

v0.10.0 continues the native-wire migration. Spectrum nodes now emit as a <molecule> containing a native <spectrum> child (with format and condition attributes for nucleus/solvent) plus a <peakList> of <peak> elements. Each peak carries the standard CML attributes: xValue (position), yValue (intensity), yMultiplicity (s/d/t/q/m), and title (assignment).

Spectrometry tools that consume CML (Mnova, ACD/Labs, etc.) can now read AsciiChem spectrum output directly. Legacy aci:spectrum text-carrier XML still parses for backwards compat.

2026-07-21 — AsciiChem v0.9: Native CML wire for Crystal

v0.9.0 begins the migration from aci: text-carrier representation to native CML wire for the beyond-formulas constructs. Crystal is first: crystal nodes now emit as a <molecule> containing a native <crystal> child (with <scalar> cells and <symmetry spaceGroup=...>) plus an <atomArray> with fractional coordinates.

Other CML tools — Avogadro, Jmol, Open Babel, any CML-aware reader — can now consume AsciiChem crystal output directly, without needing the AsciiChem-specific aci: extension namespace.

Backwards compat preserved: legacy aci:crystal text-carrier XML still parses correctly. The next releases will migrate Spectrum, ZMatrix, Mechanism, Calculation, and reaction conditions to native wire following the same pattern.

2026-07-21 — AsciiChem v0.8: chemicalml 0.3.0 + Hydrogen implicit subscripts

Two releases today, both addressing long-standing feature requests.

v0.8.0 — chemicalml 0.3.0 dependency

chemicalml 0.3.0 fixes the wire serialization gaps that had blocked native CML round-trip for Crystal / Spectrum / ZMatrix / Mechanism / Calculation / reaction conditions. Molecule wire now serializes <crystal>, <spectrum>, <zMatrix>, <propertyList> children; Reaction wire now serializes <conditionList>, <spectatorList>, <mechanism> children.

This unblocks the native-wire migration for the beyond-formulas constructs. The current aci: namespace carriers remain as a fallback; future releases will migrate each construct to its native CML form.

v0.8.0 also adds a manual-dispatch GHA release workflow (.github/workflows/release.yml) so gem pushes no longer require maintainer laptops.

v0.7.0 — Implicit subscripts on Hydrogen

Casual chemistry notation uses H2O, CH4, NH3 without explicit underscores. AsciiChem now accepts both forms.

The feature is Hydrogen-only because bare digits after other elements are already used for SMILES-style ring closures (C1-C-C-C-C-C1 is cyclohexane). Hydrogen cannot form ring closures (only one bond), so bare digits after H are unambiguously subscripts. This covers ~99% of the actual user need (formula notation like H2O, H2SO4, C6H12O6) without breaking structural notation.

For other elements, use the explicit form: O_3, S_2, N_2.

2026-07-14 — AsciiChem v0.2: the semantic model

v0.2 is the release that gives AsciiChem a real chemistry model. Every parse now produces a typed tree — Atom, Molecule, Bond, Reaction, ReactionCascade, ElectronConfiguration, EmbeddedMath — that formatters walk via the visitor pattern.

What shipped

• Semantic model. Eleven model classes with accept(visitor) and children() hooks. The model is the hub; formatters are the spokes.
• Six output formats. MathML, HTML, Text (canonicaliser), LaTeX (\ce{} via mhchem), SVG (inline), and Structural SVG (via elkrb).
• Round-trip guarantee. parse(s).to_text == s for any canonical input. The Text formatter is the canonicaliser.
• Prefix isotopes done right. ^14C binds 14 to the atom — not to an empty phantom base like AsciiMath.
• Bonds. Single, double, triple, wedge (>-), hash (-<), dative (~>), and aromatic (~~). All ASCII-spelled.
• Reactions. Four arrow types (->, <-, <=>, <->), with conditions above and below the arrow.
• Electron configurations. [Ar] 3d^6 4s^2 parses to a real model that can be queried for valence shells.
• Embedded math. Backtick escapes hand content to Plurimath.
• Linter. Balance check, valence check, bracket-balance check, isotope sanity check — all registry-driven (OCP).
• CLI. asciichem convert, parse-cml, roundtrip, lint, version. Package name is asciichem.

What’s deferred

• CML round-trip — the chemicalml gem has been refactored to a Schema3 namespace; the translator needs realignment. Specs are pending, not failing.
• Structural SVG — elkrb layout is integrated but bond connectivity needs a fix; flat molecules work, branched structures do not.
• Greek words in conditions — alpha, beta, gamma parse in stereo prefixes but not yet in reaction conditions.

Read the full breakdown on the What’s new in v0.2 page.

Planned posts

• Why we picked ~> for the dative bond — ASCII has no dative-bond convention. ~> was chosen to avoid collision with the reaction arrow (->) while evoking the donation direction.
• The semantic fix: prefix isotopes bind to atoms — a deep-dive into why ^14C and {}^14C are semantically different.
• Comparing mhchem and AsciiChem — where the two syntaxes agree, where they disagree, and why we pick our spots.
• The canonical model architecture — three concerns kept MECE: AsciiChem::Model (syntax), Chemicalml::Model (canonical), Chemicalml::Cml (wire format). The translator is the only thing that bridges them.
• Linter as first-class concern — the registry-driven linter design and how BalanceCheck, ValenceCheck, and IsotopeSanityCheck self-register at load time.