Resolve requested cleaned var names to actual draws-matrix columns

Maps each name in vars to itself, then to b_<name>, in the set of draws_colnames, mirroring the both-directions lookup .extract_truth() already performs. This fixes the F2 mismatch where generators strip the b_ prefix (so vars_of_interest holds cleaned names like c("x", "Intercept")) but brms draws matrices keep it (c("b_x","b_Intercept", "sigma")).

Usage

resolve_draw_columns(vars, draws_colnames)

Arguments

vars

Character vector of cleaned names (e.g. c("x","Intercept", "sigma")).

draws_colnames

Character vector of available draws-matrix column names.

Value

A named character vector: names are the cleaned vars (use these for output field naming), values are the actual draws column to read. Empty input returns character(0) (names preserved).

Details

Errors with a bayesim_config_error (same condition class as .extract_truth()) when a requested var is genuinely absent. A silent NA would corrupt SBC ranks and credible intervals without diagnostics.