def create_toggle_partial_tree_scorer(model_list,
data,
tree,
num_extra_nodes=0
):
'''Simple constructor for a TogglePartialTreeScorer and LikeCalcEnvironment
to support it.
Infers:
data_type (num_states) and asrv from the models in model_list
num_leaves, num_state_code_arrays, num_partials for len(data)
num_patterns from len(data[0])
Assumes that you will want a enough prob_matrix for every edge in a rooted,
binary tree to have two sets of matrices for each model/rate-category
combination.
Assumes that you want only two eigen solution per model.
Assumes that you want only two partials for internal nodes solution per
model and `num_extra_partials` in addition to this (thus if you want
one "extra" node that can be swapped in and out of the tree you
should use num_extra_partials=1, and then manually load up the
_LCE_xxx attributes for that node.
Assumes that you want one rescaling array for every 6 edges (every 4 leaves)
'''
from pytbeaglehon.like_calc_environ import LikeCalcEnvironment
asrv_list = []
num_model_rate_cats = 0
num_leaves = len(data)
num_patterns = len(data[0])
num_models = len(model_list) #TODO more generic for mixtures!
for model in model_list:
a = model.asrv
if a is None:
num_model_rate_cats += 1
else:
num_model_rate_cats += a.num_categories
num_internals = (num_leaves - 1)
num_nodes = num_internals + num_leaves
LCE = LikeCalcEnvironment(model_list=model_list,
num_patterns=num_patterns,
num_leaves=num_leaves,
num_state_code_arrays=num_leaves,
num_partials=(num_internals + num_extra_nodes)*2*num_model_rate_cats,
num_prob_matrices=(num_nodes + num_extra_nodes)*2*num_model_rate_cats,
num_eigen_storage_structs=2*num_models,
num_rescalings_multipliers= 2*(1 + num_leaves//4))
for n, row in enumerate(data):
LCE.set_state_code_array(n, row)
scorer = TogglePartialTreeScorer(LCE, tree)
return scorer
