fix: correct multi-liquid selection flag in pure-component flash - #182

What's wrong

In thermo/flash/flash_utils.py, six pure-component flash helpers set a selection flag from a count instead of a boolean:

one_liquid = len(liquids)
...
if one_liquid:
    lowest_phase = liquids[0].to_TP_zs(T, P, zs)
else:
    # pick the phase with the lowest Gibbs energy

len(liquids) is truthy for any count greater than zero, so the if one_liquid: branch runs whenever there is at least one liquid phase, not just when there is exactly one. With two or more liquid phases, this skips the Gibbs-energy comparison entirely and always returns liquids[0], regardless of which phase is actually stable. With zero liquids, it falls to the else branch, the phase list comprehension is empty, lowest_phase stays None, and the next line crashes calling .fugacities() on None.

Same pattern for one_other = len(other_phases) and one_solid = len(solids) in TSF_pure_newton and PSF_pure_newton.

Fix

Make the flags actual booleans:

one_liquid = len(liquids) == 1

applied to all six functions: TVF_pure_newton, TVF_pure_secant, PVF_pure_newton, PVF_pure_secant, TSF_pure_newton, PSF_pure_newton. This preserves the fast path for the single-phase case and correctly runs the Gibbs-energy comparison whenever there's more than one candidate phase.

Testing

I wrote a standalone repro using lightweight fake phase objects (mocking to_TP_zs, G, fugacities, dfugacities_dP/dT) to isolate the selection logic from the full EOS/data stack, since these functions don't have dedicated unit tests. With two liquid phases passed in liquids[0] = high Gibbs energy, liquids[1] = low Gibbs energy, and gas fugacity matched to the low-Gibbs phase:

• On current master, TVF_pure_secant raises fluids.numerics.SamePointError because it locks onto liquids[0] (the wrong, high-Gibbs phase) and the solver never converges.
• With this fix, all four TVF_pure_*/PVF_pure_* functions correctly select the low-Gibbs phase and converge.
• A single-liquid case confirms the len == 1 fast path is unaffected.

I also confirmed git blame shows this line unchanged since it was introduced in 2021.