def test_ensemblecombination_sparse():
"""Test ensemble combinations where the ensembles are not so similiar,
something missing in some ensembles etc.
"""
if "__file__" in globals():
# Easen up copying test code into interactive sessions
testdir = os.path.dirname(os.path.abspath(__file__))
else:
testdir = os.path.abspath(".")
ref = ensemble.ScratchEnsemble(
"reektest", testdir + "/data/testensemble-reek001/" + "realization-*/iter-0"
)
ref.load_smry(time_index="yearly", column_keys=["F*"])
# Initialize over again to get two different objects
ior = ensemble.ScratchEnsemble(
"reektest", testdir + "/data/testensemble-reek001/" + "realization-*/iter-0"
)
ior.load_smry(time_index="yearly", column_keys=["F*"])
ior.remove_realizations(3)
assert len(ref) == 5
assert len(ior) == 4
assert len(ior - ref) == 4
assert 3 not in (ior - ref)["parameters"].REAL.unique()
assert 3 not in (ior - ref)["unsmry--yearly"].REAL.unique()
# Delete a specific date in the ior ensemble
dframe = ior[4].data["share/results/tables/unsmry--yearly.csv"]
print(dframe.DATE.dtype)
dframe.drop(2, inplace=True) # index 2 is for date 2002-01-1
# Inject into ensemble again:
ior[4].data["share/results/tables/unsmry--yearly.csv"] = dframe
assert "2002-01-01" not in list((ior - ref)["unsmry--yearly.csv"].DATE.unique())
# Convert ref case to virtual:
vref = ref.to_virtual()
# Do the same checks over again:
assert 3 not in (ior - vref)["parameters"].REAL.unique()
assert 3 not in (ior - vref)["unsmry--yearly"].REAL.unique()
assert "2002-01-01" not in list((ior - vref)["unsmry--yearly.csv"].DATE.unique())
assert len((ior - vref)["unsmry--yearly"]) == 19
unsmry = vref.data["share/results/tables/unsmry--yearly.csv"]
del unsmry["FWIR"]
vref.data["share/results/tables/unsmry--yearly.csv"] = unsmry
assert "FWIR" in ior.get_df("unsmry--yearly").columns
assert "FWIR" not in vref.get_df("unsmry--yearly").columns
assert "FWIR" not in (ior - vref)["unsmry--yearly"].columns
def build_ensemble_df_list(ensemble_paths: Optional[List[str]],
vectors: List[str]) -> List[pd.DataFrame]:
"""Helper function to read and prepare ensemble data.
Args:
ensemble_paths: The ensemble paths to retrieve data from
vectors: List of vector to extract
Returns:
List of ensemble dataframe with required data to create plots.
"""
data: list = []
if ensemble_paths is not None:
for prior in ensemble_paths:
df_data = ensemble.ScratchEnsemble(
"flownet_ensemble",
paths=prior.replace("%d", "*"),
).get_smry(column_keys=vectors)
df_data_sorted = df_data.sort_values("DATE")
df_realizations = df_data_sorted[
df_data_sorted["DATE"] == df_data_sorted.values[-1][0]]["REAL"]
data.append(df_data.merge(df_realizations, how="inner"))
return data
def gatherresults(config):
"""Gathers .csv files from different realisations into one"""
ensemblepaths = config["results"]["ensemblepaths"]
print("Gathering csv files from:")
print(ensemblepaths)
singleresultfile = config["results"]["singleresultfile"]
print(singleresultfile)
doe_ensemble = ensemble.ScratchEnsemble("doe_ensemble", ensemblepaths)
results = doe_ensemble.load_csv(singleresultfile)
if config["results"]["writeresultfile"]:
outdir = config["results"]["exportdir"]
outfilename = config["results"]["exportfilename"]
if not os.path.exists(outdir):
os.mkdir(outdir)
fullname = os.path.join(outdir, outfilename)
results.to_csv(fullname, index=False)
return results
def test_ensemblecomb_observations():
"""Test a combination of EnsembleCombinations and mismatch calculations"""
if "__file__" in globals():
# Easen up copying test code into interactive sessions
testdir = os.path.dirname(os.path.abspath(__file__))
else:
testdir = os.path.abspath(".")
reekensemble = ensemble.ScratchEnsemble(
"reektest", testdir + "/data/testensemble-reek001/" + "realization-*/iter-0"
)
reekensemble.load_smry(time_index="yearly", column_keys=["F*"])
delta_ens = reekensemble - 0.5 * reekensemble
# Now find the realization in delta_ens that is closest to
# the mean delta-FOPT profile:
mean = delta_ens.agg("mean") # a virtual realization
obs = ensemble.Observations({})
obs.load_smry(mean, "FOPT", time_index="yearly")
mis = obs.mismatch(delta_ens)
# Realization 4 is best representing the mean delta FOPT:
assert mis.groupby("REAL").sum()["L2"].sort_values().index.values[0] == 4
def test_ensemblecombination_basic():
if "__file__" in globals():
# Easen up copying test code into interactive sessions
testdir = os.path.dirname(os.path.abspath(__file__))
else:
testdir = os.path.abspath(".")
reekensemble = ensemble.ScratchEnsemble(
"reektest",
testdir + "/data/testensemble-reek001/" + "realization-*/iter-0")
reekensemble.load_smry(time_index="yearly", column_keys=["F*"])
reekensemble.load_scalar("npv.txt", convert_numeric=True)
# Difference with itself should give zero..
diff = ensemble.EnsembleCombination(ref=reekensemble, sub=reekensemble)
assert diff["parameters"]["KRW1"].sum() == 0
assert diff["unsmry--yearly"]["FOPT"].sum() == 0
assert diff["npv.txt"]["npv.txt"].sum() == 0
foptsum = reekensemble.get_df("unsmry--yearly")["FOPT"].sum()
half = 0.5 * reekensemble
assert half["unsmry--yearly"]["FOPT"].sum() == 0.5 * foptsum
assert half["npv.txt"].sort_values("REAL")["npv.txt"].iloc[0] == 1722
# This is only true since we only juggle one ensemble here:
assert len(half.get_smry_dates(freq="monthly")) == len(
reekensemble.get_smry_dates(freq="monthly"))
# Test something long:
zero = (reekensemble + 4 * reekensemble - reekensemble * 2 -
(-1) * reekensemble - reekensemble * 4)
assert zero["parameters"]["KRW1"].sum() == 0
assert len(
diff.get_smry(column_keys=["FOPR", "FGPR", "FWCT"]).columns) == 5
# eclipse summary vector statistics for a given ensemble
df_stats = diff.get_smry_stats(column_keys=["FOPR", "FGPR"],
time_index="monthly")
assert isinstance(df_stats, pd.DataFrame)
assert len(df_stats.columns) == 2
assert isinstance(df_stats["FOPR"]["mean"], pd.Series)
assert len(df_stats["FOPR"]["mean"].index) == 38
# check if wild cards also work for get_smry_stats
df_stats = reekensemble.get_smry_stats(column_keys=["FOP*", "FGP*"],
time_index="monthly")
assert len(df_stats.columns) == len(
reekensemble.get_smrykeys(["FOP*", "FGP*"]))
# Check webviz requirements for dataframe
stats = df_stats.index.levels[0]
assert "minimum" in stats
assert "maximum" in stats
assert "p10" in stats
assert "p90" in stats
assert "mean" in stats
assert df_stats["FOPR"]["minimum"].iloc[-2] < df_stats["FOPR"][
"maximum"].iloc[-2]
# Virtualization of ensemble combinations
# (equals comutation of everything)
vzero = zero.to_virtual()
assert len(vzero.keys()) == len(zero.keys())
def test_ensemblecombination_basic():
"""Basic tests for linear combinations of ensemble objects"""
if "__file__" in globals():
# Easen up copying test code into interactive sessions
testdir = os.path.dirname(os.path.abspath(__file__))
else:
testdir = os.path.abspath(".")
reekensemble = ensemble.ScratchEnsemble(
"reektest", testdir + "/data/testensemble-reek001/" + "realization-*/iter-0"
)
reekensemble.load_smry(time_index="yearly", column_keys=["F*"])
reekensemble.load_scalar("npv.txt", convert_numeric=True)
# Difference with itself should give zero..
diff = ensemble.EnsembleCombination(ref=reekensemble, sub=reekensemble)
assert len(diff) == 5
# Implicit virtualization and aggregation
assert diff.agg("mean")["parameters"]["KRW1"] == 0
assert diff.parameters["KRW1"].sum() == 0
assert diff["parameters"]["KRW1"].sum() == 0
assert diff["unsmry--yearly"]["FOPT"].sum() == 0
assert diff["npv.txt"]["npv.txt"].sum() == 0
assert (
diff.get_volumetric_rates(column_keys="FOPT", time_index="yearly")["FOPR"].sum()
== 0
)
foptsum = reekensemble.get_df("unsmry--yearly")["FOPT"].sum()
half = 0.5 * reekensemble
assert half["unsmry--yearly"]["FOPT"].sum() == 0.5 * foptsum
assert half["npv.txt"].sort_values("REAL")["npv.txt"].iloc[0] == 1722
assert (
half.get_smry(column_keys="FOPT", time_index="last")["FOPT"].sum()
== half.get_volumetric_rates(column_keys="FOPT", time_index="yearly")[
"FOPR"
].sum()
)
smrymeta = diff.get_smry_meta(["FO*"])
assert "FOPT" in smrymeta
# This is only true since we only juggle one ensemble here:
assert len(half.get_smry_dates(freq="monthly")) == len(
reekensemble.get_smry_dates(freq="monthly")
)
# Test comb of virtualized ensembles.
vhalf = 0.5 * reekensemble.to_virtual()
assert vhalf["unsmry--yearly"]["FOPT"].sum() == 0.5 * foptsum
vhalf_filtered = (0.5 * reekensemble).to_virtual(keyfilter="parameters")
# This means that unsmry--yearly is not found:
assert vhalf_filtered.shortcut2path("unsmry--yearly") == "unsmry--yearly"
# Ask to include summary data:
vhalf_filtered2 = (0.5 * reekensemble).to_virtual(keyfilter="unsmry")
assert not vhalf_filtered2.get_df("unsmry--yearly").empty
with pytest.raises((KeyError, ValueError)):
# pylint: disable=pointless-statement
vhalf_filtered2.parameters
# Get summary data with parameters:
smry_params = vhalf.get_df("unsmry--yearly", merge="parameters.txt")
assert "SORG1" in smry_params
assert "FWCT" in smry_params
# Data is merged before ensemble computations:
assert smry_params["T_1OG"].min() == reekensemble.parameters["T_1OG"].min() * 0.5
print(smry_params)
# Test something long:
# zero = (
# reekensemble
# + 4 * reekensemble
# - reekensemble * 2
# - (-1) * reekensemble
# - reekensemble * 4
# )
# The above takes too long time to compute until #10 is solved.
# We can test something cheaper:
zero = reekensemble + reekensemble - 2 * reekensemble
assert zero["parameters"]["KRW1"].sum() == 0
smrymeta = zero.get_smry_meta(["FO*"])
assert "FOPT" in smrymeta
vzero = (
reekensemble.to_virtual()
+ reekensemble.to_virtual()
- 2 * reekensemble.to_virtual()
)
assert vzero["parameters"]["KRW1"].sum() == 0
assert len(diff.get_smry(column_keys=["FOPR", "FGPR", "FWCT"]).columns) == 5
# eclipse summary vector statistics for a given ensemble
df_stats = diff.get_smry_stats(column_keys=["FOPR", "FGPR"], time_index="monthly")
assert isinstance(df_stats, pd.DataFrame)
assert len(df_stats.columns) == 2
assert isinstance(df_stats["FOPR"]["mean"], pd.Series)
assert len(df_stats["FOPR"]["mean"].index) == 38
# check if wild cards also work for get_smry_stats
df_stats = reekensemble.get_smry_stats(
column_keys=["FOP*", "FGP*"], time_index="monthly"
)
assert len(df_stats.columns) == len(reekensemble.get_smrykeys(["FOP*", "FGP*"]))
# Check webviz requirements for dataframe
stats = df_stats.index.levels[0]
assert "minimum" in stats
assert "maximum" in stats
assert "p10" in stats
assert "p90" in stats
assert "mean" in stats
assert df_stats["FOPR"]["minimum"].iloc[-2] < df_stats["FOPR"]["maximum"].iloc[-2]
# Virtualization of ensemble combinations
# (equals comutation of everything)
vzero = zero.to_virtual()
assert len(vzero.keys()) == len(zero.keys())
