def test_find_matching_scenarios_get_precise_values():
# We should get out explicit numbers if we ask for them
all_data = find_matching_scenarios(
df_to_test,
simple_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
return_all_info=True,
)
assert all_data[0][1] == all_data[1][1]
assert all_data[0][1] == 0
assert all_data[2][1] > 0
unsplit_model_c = all_data[2][1]
# And if we separate out by model, we no longer combine the high and low values
# to produce a good match
all_data = find_matching_scenarios(
df_to_test,
simple_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
classify_models=["high_model", _mc],
return_all_info=True,
)
assert all_data[-1][0][0] == "high_model"
assert all_data[0][0][0] == _mc
assert all_data[0][1] == unsplit_model_c
def test_find_matching_scenarios_no_data_for_time():
time_col = simple_df.time_col
half_simple_df = simple_df.filter(scenario="scen_a")
half_simple_df.data[time_col].loc[0] = 0
with pytest.raises(ValueError):
find_matching_scenarios(
simple_df,
half_simple_df,
variable_follower,
variable_leaders,
["scen_a", "scen_b"],
)
def test_find_matching_scenarios_dual_region():
multiregion_df = simple_df.data.append(
pd.DataFrame(
[[_mc, _sa, "Country", _eco2, _gtc, 2010, 2]],
columns=_msrvu + [simple_df.time_col, "value"],
))
multiregion_df = pyam.IamDataFrame(multiregion_df)
with pytest.raises(AssertionError):
find_matching_scenarios(
df_to_test,
multiregion_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
return_all_info=True,
)
def test_find_matching_scenarios_complicated(options, expected):
# This is similar to the above case except with multiple models involved and
# requiring specific interpolation. Tests:
# 1) Closest option chosen
# 2) Invalid options ignored, if tied the earlier option is selected instead
# 3) This reverses as expected
scenario = find_matching_scenarios(df_to_test, simple_df,
variable_follower, variable_leaders,
options)
assert scenario == ("*", expected)
def test_find_matching_scenarios_differential():
# If we use a differential measurement, they should all be the same in this case
all_data = find_matching_scenarios(
df_to_test,
simple_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
classify_models=["high_model", _mc],
return_all_info=True,
use_change_not_abs=True,
)
assert all_data[0][0] == ("high_model", "right_scenario")
assert all_data[0][1] == all_data[5][1]
# But if we add a small amount to only one point in the differential, it will
# be downgraded
df_to_test["value"].iloc[0] = df_to_test["value"].iloc[0] + 0.1
all_data = find_matching_scenarios(
df_to_test,
simple_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
classify_models=["high_model", _mc],
return_all_info=True,
use_change_not_abs=True,
)
assert all_data[0][0] == ("high_model", "right_scenario")
assert all_data[0][1] != all_data[1][1]
df_to_test["value"].iloc[0] = df_to_test["value"].iloc[0] - 0.6
all_data = find_matching_scenarios(
df_to_test,
simple_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
classify_models=["high_model", _mc],
return_all_info=True,
use_change_not_abs=True,
)
assert all_data[0][0] == (_mc, "right_scenario")
assert all_data[0][1] == all_data[1][1]
def test_find_matching_scenarios_empty():
noregion_df = simple_df.filter(scenario="impossible")
nothing = find_matching_scenarios(
df_to_test,
noregion_df,
variable_follower,
variable_leaders,
["right_scenario", "wrong_scenario", "scen_a", "scen_b"],
return_all_info=True,
)
assert nothing is None
def test_find_matching_scenarios_use_change_instead_of_absolute():
# In this case, we will ignore any offset
half_simple_df = simple_df.filter(scenario="scen_a")
half_simple_df.data["value"] = half_simple_df.data["value"] + 10000
scenarios = find_matching_scenarios(
simple_df,
half_simple_df,
variable_follower,
variable_leaders,
["scen_a", "scen_b"],
use_change_not_abs=True,
)
assert scenarios == ("*", "scen_a")
def test_find_matching_scenarios_matched(half_val, expected):
# Tests
# 1) that 1st option is used in the case of equality
# 2) and if it's closer,
# 3) But not if it's further away
half_simple_df = simple_df.filter(scenario="scen_a")
half_simple_df.data["value"].loc[0] = half_val
scenarios = find_matching_scenarios(
simple_df,
half_simple_df,
variable_follower,
variable_leaders,
["scen_a", "scen_b"],
)
assert scenarios == ("*", expected)