def process_data(cls, reader, stream, metadata):
index = np.arange(metadata["firstyear"], metadata["lastyear"] + 1)
columns = {}
data = []
for i in range(metadata["datacolumns"]):
column_header, column_data = cls.process_variable(stream)
for k in column_header:
if k not in columns:
columns[k] = []
columns[k].append(column_header[k])
data.append(column_data)
df = pd.DataFrame(np.asarray(data).T, index=index)
if isinstance(df.index, pd.core.indexes.numeric.Float64Index):
df.index = df.index.to_series().round(3)
df.index.name = "time"
# Convert the regions to openscm regions
columns["region"] = convert_magicc_to_openscm_regions(columns["region"])
# Convert the variable names to openscm variables
columns["variable"] = [
d[4:] if d.startswith("DAT_") else d for d in columns["variable"]
]
columns["variable"] = convert_magicc6_to_magicc7_variables(columns["variable"])
columns["variable"] = convert_magicc7_to_openscm_variables(columns["variable"])
return df, metadata, columns
def _get_initial_nml_and_data_block(self):
data_block = self.data_block
regions = convert_magicc_to_openscm_regions(
data_block.columns.get_level_values("region").tolist(),
inverse=True)
regions = self._ensure_file_region_type_consistency(regions)
variables = convert_magicc7_to_openscm_variables(
data_block.columns.get_level_values("variable").tolist(),
inverse=True)
# trailing EMIS is incompatible, for now
variables = [v.replace("_EMIS", "") for v in variables]
units = convert_pint_to_fortran_safe_units(
data_block.columns.get_level_values("unit").tolist())
todos = data_block.columns.get_level_values("todo").tolist()
data_block = data_block.rename(columns=str).reset_index()
data_block.columns = [
[self._variable_header_row_name] + variables,
["TODO"] + todos,
["UNITS"] + units,
["YEARS"] + regions,
]
nml = Namelist()
nml["THISFILE_SPECIFICATIONS"] = Namelist()
nml["THISFILE_SPECIFICATIONS"]["THISFILE_DATACOLUMNS"] = (
len(data_block.columns) - 1 # for YEARS column
)
nml["THISFILE_SPECIFICATIONS"]["THISFILE_DATAROWS"] = len(data_block)
nml["THISFILE_SPECIFICATIONS"]["THISFILE_FIRSTYEAR"] = int(
np.floor(data_block.iloc[0, 0]))
nml["THISFILE_SPECIFICATIONS"]["THISFILE_LASTYEAR"] = int(
np.floor(data_block.iloc[-1, 0]))
step_length = data_block.iloc[1:, 0].values - data_block.iloc[:-1,
0].values
try:
np.testing.assert_allclose(step_length, step_length[0], rtol=0.02)
step_length = step_length[0]
annual_steps = np.round(1 / step_length, 1)
if annual_steps < 1:
annual_steps = 0
else:
annual_steps = int(annual_steps)
except AssertionError:
annual_steps = 0 # irregular timesteps
nml["THISFILE_SPECIFICATIONS"]["THISFILE_ANNUALSTEPS"] = annual_steps
units_unique = list(set(self._get_df_header_row("unit")))
nml["THISFILE_SPECIFICATIONS"]["THISFILE_UNITS"] = (
convert_pint_to_fortran_safe_units(units_unique[0])
if len(units_unique) == 1 else "MISC")
nml["THISFILE_SPECIFICATIONS"].update(
self._get_dattype_regionmode(regions))
return nml, data_block
def _get_region_order(self, data_block):
regions = data_block.columns.get_level_values("region").tolist()
region_order_magicc = get_region_order(regions, self._scen_7)
region_order = convert_magicc_to_openscm_regions(region_order_magicc)
unrecognised_regions = set(regions) - set(region_order)
if unrecognised_regions:
error_msg = ("Are all of your regions OpenSCM regions? I don't "
"recognise: {}".format(sorted(unrecognised_regions)))
raise ValueError(error_msg)
return region_order
def _get_region_order(self, data_block):
try:
regions = data_block.columns.get_level_values("region").tolist()
region_order_magicc = get_region_order(regions, self._scen_7)
region_order = convert_magicc_to_openscm_regions(
region_order_magicc)
return region_order
except ValueError:
abbreviations = [
convert_magicc_to_openscm_regions(r, inverse=True)
for r in set(regions)
]
unrecognised_regions = [
a for a in abbreviations
if a in regions or DATA_HIERARCHY_SEPARATOR in a
]
if unrecognised_regions:
warnings.warn(
"Not abbreviating regions, could not find abbreviation for {}"
.format(unrecognised_regions))
return regions
def read_data_block(self):
number_years = int(self.lines[0].strip())
# go through datablocks until there are none left
while True:
ch = {}
pos_block = self._stream.tell()
region = convert_magicc_to_openscm_regions(self._stream.readline().strip())
try:
variables = self._read_data_header_line(
self._stream, ["Years", "Year", "YEARS", "YEAR"]
)
except IndexError: # tried to get variables from empty string
break
except AssertionError: # tried to get variables from a notes line
break
variables = convert_magicc6_to_magicc7_variables(variables)
ch["variable"] = convert_magicc7_to_openscm_variables(
[v + "_EMIS" for v in variables]
)
ch["unit"] = self._read_data_header_line(self._stream, ["Yrs", "YEARS"])
ch = self._read_units(ch)
ch["todo"] = ["SET"] * len(variables)
ch["region"] = [region] * len(variables)
region_block = StringIO()
for i in range(number_years):
region_block.write(self._stream.readline())
region_block.seek(0)
region_df = self._convert_data_block_to_df(region_block)
try:
df = pd.concat([region_df, df], axis="columns")
columns = {key: ch[key] + columns[key] for key in columns}
except NameError:
df = region_df
columns = ch
self._stream.seek(pos_block)
try:
return df, columns
except NameError:
error_msg = (
"This is unexpected, please raise an issue on "
"https://github.com/openscm/pymagicc/issues"
)
raise Exception(error_msg)
def _get_column_headers_and_update_metadata(self, stream, metadata):
if self._magicc7_style_header():
column_headers, metadata = self._read_magicc7_style_header(
stream, metadata)
else:
column_headers, metadata = self._read_magicc6_style_header(
stream, metadata)
column_headers["variable"] = convert_magicc7_to_openscm_variables(
column_headers["variable"])
column_headers["region"] = convert_magicc_to_openscm_regions(
column_headers["region"])
return column_headers, metadata
def _write_header(self, output):
header_lines = []
header_lines.append("{}".format(len(self.data_block)))
variables = self._get_df_header_row("variable")
variables = convert_magicc7_to_openscm_variables(variables, inverse=True)
variables = _strip_emis_variables(variables)
regions = self._get_df_header_row("region")
regions = convert_magicc_to_openscm_regions(regions, inverse=True)
regions = self._ensure_file_region_type_consistency(regions)
special_scen_code = get_special_scen_code(regions=regions, emissions=variables)
header_lines.append("{}".format(special_scen_code))
# for a scen file, the convention is (although all these lines are
# actually ignored by source so could be anything):
# - line 3 is name
# - line 4 is description
# - line 5 is notes (other notes lines go at the end)
# - line 6 is empty
header_lines.append("name: {}".format(self.minput["scenario"].unique()[0]))
header_lines.append(
"description: {}".format(
self.minput.metadata.pop(
"description", "metadata['description'] is written here"
)
)
)
header_lines.append(
"notes: {}".format(
self.minput.metadata.pop("notes", "metadata['notes'] is written here")
)
)
header_lines.append("")
try:
header_lines.append(self.minput.metadata.pop("header"))
except KeyError:
pass
for k, v in self.minput.metadata.items():
header_lines.append("{}: {}".format(k, v))
output.write(self._newline_char.join(header_lines))
output.write(self._newline_char)
return output
def _get_dattype_regionmode_regions_row(regions, scen7=False):
regions_unique = set([
convert_magicc_to_openscm_regions(r, inverse=True)
for r in set(regions)
])
def find_region(x):
return set(x) == regions_unique
region_rows = DATTYPE_REGIONMODE_REGIONS["regions"].apply(find_region)
scen7_rows = DATTYPE_REGIONMODE_REGIONS["thisfile_dattype"] == "SCEN7"
dattype_rows = scen7_rows if scen7 else ~scen7_rows
region_dattype_row = region_rows & dattype_rows
if sum(region_dattype_row) != 1:
error_msg = (
"Unrecognised regions, they must be part of "
"pymagicc.definitions.DATTYPE_REGIONMODE_REGIONS. If that doesn't make "
"sense, please raise an issue at "
"https://github.com/openscm/pymagicc/issues")
raise ValueError(error_msg)
return region_dattype_row
def test_convert_openscm_to_magicc_regions_one_way(magicc7, openscm):
assert convert_magicc_to_openscm_regions(magicc7, inverse=False) == openscm
def test_convert_openscm_to_magicc_regions(magicc7, openscm):
assert convert_magicc_to_openscm_regions(magicc7, inverse=False) == openscm
assert convert_magicc_to_openscm_regions(openscm, inverse=True) == magicc7
def _convert_compact_table_to_df_metadata_column_headers(
self, compact_table):
ts_cols = [c for c in compact_table if "__" in c]
para_cols = [c for c in compact_table if "__" not in c]
ts = compact_table[ts_cols]
ts = ts.T
def sort_ts_ids(inid):
variable, region, year = inid.split("__")
variable = variable.replace("DAT_", "")
return {"variable": variable, "region": region, "year": year}
ts["variable"] = ts.index.map(
lambda x: convert_magicc7_to_openscm_variables(
x.split("__")[0].replace("DAT_", "")))
ts["region"] = ts.index.map(
lambda x: convert_magicc_to_openscm_regions(x.split("__")[1]))
ts["year"] = ts.index.map(lambda x: x.split("__")[2])
# Make sure all the year strings are four characters long. Not the best test,
# but as good as we can do for now.
if not (ts["year"].apply(len)
== 4).all(): # pragma: no cover # safety valve
raise NotImplementedError("Non-annual data not yet supported")
ts["year"] = ts["year"].astype(int)
ts = ts.reset_index(drop=True)
id_cols = {"variable", "region", "year"}
run_cols = set(ts.columns) - id_cols
ts = ts.melt(value_vars=run_cols, var_name="run_id", id_vars=id_cols)
ts["unit"] = "unknown"
new_index = list(set(ts.columns) - {"value"})
ts = ts.set_index(new_index)["value"].unstack("year")
paras = compact_table[para_cols]
paras.index.name = "run_id"
cols_to_merge = find_parameter_groups(paras.columns.tolist())
paras_clean = paras.copy()
# Aggregate the columns
for new_col, components in cols_to_merge.items():
components = sorted(components)
paras_clean.loc[:,
new_col] = tuple(paras[components].values.tolist())
paras_clean = paras_clean.drop(columns=components)
years = ts.columns.tolist()
ts = ts.reset_index().set_index("run_id")
out = pd.merge(ts, paras_clean, left_index=True,
right_index=True).reset_index()
id_cols = set(out.columns) - set(years)
out = out.melt(value_vars=years, var_name="year", id_vars=id_cols)
new_index = list(set(out.columns) - {"value"})
out = out.set_index(new_index)["value"].unstack("year")
out = out.T
column_headers = {
name.lower(): out.columns.get_level_values(name).tolist()
for name in out.columns.names
}
df = out.copy()
metadata = {}
return metadata, df, column_headers
def _write_datablock(self, output):
# for SCEN files, the data format is vitally important for the source code
# we have to work out a better way of matching up all these conventions/testing them, tight coupling between pymagicc and MAGICC may solve it for us...
lines = output.getvalue().split(self._newline_char)
# notes are everything except the first 6 lines
number_notes_lines = len(lines) - 6
def _gip(lines, number_notes_lines):
"""
Get the point where we should insert the data block.
"""
return len(lines) - number_notes_lines
region_order_db = get_region_order(
self._get_df_header_row("region"), scen7=self._scen_7
)
region_order_magicc = self._ensure_file_region_type_consistency(region_order_db)
# format is vitally important for SCEN files as far as I can tell
time_col_length = 11
first_col_format_str = ("{" + ":{}d".format(time_col_length) + "}").format
other_col_format_str = "{:10.4f}".format
# TODO: doing it this way, out of the loop, should ensure things
# explode if your regions don't all have the same number of emissions
# timeseries or does extra timeseries in there (that probably
# shouldn't raise an error, another one for the future), although the
# explosion will be cryptic so should add a test for good error
# message at some point
formatters = [other_col_format_str] * (
int(len(self.data_block.columns) / len(region_order_db))
+ 1 # for the years column
)
formatters[0] = first_col_format_str
variables = convert_magicc7_to_openscm_variables(
self._get_df_header_row("variable"), inverse=True
)
variables = _strip_emis_variables(variables)
special_scen_code = get_special_scen_code(
regions=region_order_magicc, emissions=variables
)
if special_scen_code % 10 == 0:
variable_order = PART_OF_SCENFILE_WITH_EMISSIONS_CODE_0
else:
variable_order = PART_OF_SCENFILE_WITH_EMISSIONS_CODE_1
for region_db, region_magicc in zip(region_order_db, region_order_magicc):
region_block_region = convert_magicc_to_openscm_regions(region_db)
region_block = self.data_block.xs(
region_block_region, axis=1, level="region", drop_level=False
)
region_block.columns = region_block.columns.droplevel("todo")
region_block.columns = region_block.columns.droplevel("region")
variables = region_block.columns.levels[0]
variables = convert_magicc7_to_openscm_variables(variables, inverse=True)
region_block.columns = region_block.columns.set_levels(
levels=_strip_emis_variables(variables), level="variable",
)
region_block = region_block.reindex(
variable_order, axis=1, level="variable"
)
variables = region_block.columns.get_level_values("variable").tolist()
variables = convert_magicc6_to_magicc7_variables(
[v.replace("_EMIS", "") for v in variables], inverse=True
)
units = convert_pint_to_fortran_safe_units(
region_block.columns.get_level_values("unit").tolist()
)
# column widths don't work with expressive units
units = [u.replace("_", "").replace("peryr", "") for u in units]
if not (region_block.columns.names == ["variable", "unit"]):
raise AssertionError(
"Unexpected region block columns: "
"{}".format(region_block.columns.names)
)
region_block = region_block.rename(columns=str).reset_index()
region_block.columns = [["YEARS"] + variables, ["Yrs"] + units]
region_block_str = region_magicc + self._newline_char
region_block_str += region_block.to_string(
index=False, formatters=formatters, sparsify=False
)
region_block_str += self._newline_char * 2
lines.insert(_gip(lines, number_notes_lines), region_block_str)
output.seek(0)
output.write(self._newline_char.join(lines))
return output