Required Coefficients File Not Backwards Compatible or Documented

[danielsclint]

Pull Request #325 introduced a new requirement that each model define a "coefficients" file. This makes the structure of the config files cleaner in the long run. However, the documentation has not been updated (as far as I can tell) to reflect this new requirement, and it breaks prior setups (e.g., ARC implementation) that does not abide by this standard.

activitysim/activitysim/core/simulate.py

Lines 138 to 140 in 05a25f6

	assert 'COEFFICIENTS' in model_settings, \
	"'COEFFICIENTS' tag not in model_settings in %s" % model_settings.get('source_file_paths')
	file_name = model_settings['COEFFICIENTS']

We will update the ARC code base appropriately to follow the rules.

Might be worth discussing whether this new feature should truly be a requirement or whether it should just be strongly recommended.

[danielsclint]

The COEFFICIENTS and COEFFICIENTS_TEMPLATE seem prone to user input error with all of the mappings between COEFFICIENT names, COEFFICIENT_TEMPLATE aliases, and actual usage in the the model specification or YAML file. Could the same logic in the COEFFECIENTS and COEFFICIENTS_TEMPLATE table be handled in a slightly more verbose single table?

PROPOSED COEFFICIENT TABLE SCHEMA (used to map specific coefficient names to generic names)

coefficient_name	segment_name	value	constrain
coef_ivt	default	-0.0175	F
coef_ivt	school_univ	-0.0224	F
coef_ivt	work	-0.0134	F
coef_ivt	at_work	-0.0188	F

In the table above, the segment name would be read and applied to the specific segment when available. The default segment would be used if no other segment matches and is assumed to be universal.

This seems easier than the current implementation which requires keeping multiple files in sync.

EXISTING COEFFICIENT TEMPLATE TABLE (used to map specific coefficient names to generic names)

coefficient_name	eatout	escort	othdiscr	othmaint	school	shopping	social	univ	work	atwork
coef_ivt	coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	coef_ivt_school_univ	coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	coef_ivt_school_univ	coef_ivt_work	coef_ivt_atwork

EXISTING COEFFICIENT TABLE

coefficient_name	value	constrain
coef_ivt_eatout_escort_othdiscr_othmaint_shopping_social	-0.0175	F
coef_ivt_school_univ	-0.0224	F
coef_ivt_work	-0.0134	F
coef_ivt_atwork	-0.0188	F

(Admittedly, I've been out of the loop for a while, so I could clearly be missing the importance of this extra step too)

[danielsclint]

This was implemented to support Larch, and the implementation is more clear. Closing this issue.