[djbower]

Stage 2 of the species-naming work, following #168. This adds a way to select condensed species by phase component (in the Holland and Powell sense: the gas, liquid or mineral that a set of data refers to). The old interface (_s species, PurePhase.from_species, create_condensed(..., state=...)) is unchanged. Migrating usages (stage 3) and removing the old interface (stage 4) will follow in separate PRs.

New interface

components_of("SiO2")  # ("quartz-alpha", "quartz-beta", "cristobalite-beta", "liquid")

PurePhase.from_formula("SiO2")      # O2Si_cd: all condensed phase components, stitched in order of temperature
PurePhase.from_components(("quartz-alpha", "quartz-beta"), name="quartz")   # O2Si_quartz
PurePhase.from_component("cristobalite-beta")       # O2Si_cristobalite-beta
PurePhase.from_component("liquid", formula="H2O")   # H2O_l ("liquid" is shared, so the formula is needed)
PurePhase.from_component("diamond")                 # C_diamond

Each method takes an optional database=; without it, the default database is used. The logic lives on DataBase (condensed_from_formula, condensed_from_components, condensed_from_component and components_of), and the PurePhase methods call it.

Rules

Stitching is done on demand with stitch_glenn_properties from #168.

• Only condensed phase components can be stitched. The gas raises an error.
• All phase components must be of one formula. A mineral fixes the formula, so formula= is needed only for a set of shared names such as liquid.
• Temperature ranges may not overlap or leave a gap. With the current data this affects only Fe2O3 (haematite overlaps the liquid), which waits on the source of the data.
• Several phase components can be stitched only if they all come from a data file (packaged, or GlennDataBase.from_file) and have no volume model. A single phase component can come from anywhere, including add_species, and keeps its volume model. So from_formula("C") gives graphite with its volume, and diamond can only be selected on its own.

Implementation

• The reader builds a table keyed by Hill formula and then phase component: read_glenn_phase_components and glenn_phase_components. The name-keyed data are made from the same table, so existing lookups are unchanged.
• DataBase keeps this table, plus a table of the phase components added with add_species.
• The default database now gives C_graphite the graphite volume, as default_properties already does.
• A custom DataBase subclass must call super().__init__().

Tests

test_phase_components (tests/test_database.py) checks:

• O2Si_cd matches O2Si_s at 1000 K and O2Si_l at 2500 K.
• C_cd matches C_s at 10⁴ bar, including the volume.
• The database without the graphite volume gives a different result.
• The names that each method gives.
• Each rule raises.

Full suite: 243 passed. Pyright and the docs build (-W) pass.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 95.83333% with 5 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
atmodeller/database.py	94.44%	4 Missing ⚠️
atmodeller/thermodata/janaf.py	96.96%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!