When defining a unit programmatically, how does Pint decide what to use for formatting?

[nealkruis]

I have a few common aliases and units that I use frequently:

# Add new aliases
unit_registry.define("@alias inch_H2O_39F = in_H2O")
unit_registry.define("@alias ton_of_refrigeration = ton_ref")
unit_registry.define("@alias pound = lb_m")
unit_registry.define("@alias force_pound = lb_f")

# Add new units
unit_registry.define("cubic_feet_per_minute = cu_ft / min = cfm")
unit_registry.define("gallons_per_minute = gallon / min = gpm")
unit_registry.define("thermal_resistance_SI = m**2*K/W = R_value_SI = R_SI")
unit_registry.define("thermal_resistance_IP = ft**2*degR*h/Btu = R_value_IP = R_IP")
unit_registry.define("thermal_conductance_SI = W/(m**2*K) = U_factor_SI = U_SI")
unit_registry.define("thermal_conductance_IP = Btu/(ft**2*degR*h) = U_factor_IP = U_IP")

It's not clear to me how Pint decides what to use when formatting these units with ~P, or how/if I can change the default behavior. For example, how can I specify:

• thermal_resistance_SI prints as m²·K/W (and not R_value_SI)
• ton_ref prints as ton_ref (and not refrigeration_ton)