import logging
from typing import Callable, Dict, List, Mapping, Optional, Union
import ixmp
import pandas as pd
from message_ix import Scenario
from sdmx.model.v21 import Code
from message_ix_models.util import add_par_data, strip_par_data
from message_ix_models.util.ixmp import maybe_check_out, maybe_commit
from message_ix_models.util.scenarioinfo import ScenarioInfo, Spec
log = logging.getLogger(__name__)
def _add_unit(mp: ixmp.Platform, unit: str, comment: str) -> None:
"""Handle exceptions in :meth:`.Platform.add_unit`."""
# TODO move upstream to ixmp.JDBCBackend
log.info(f"Add unit {repr(unit)}")
try:
mp.add_unit(unit, comment)
except Exception as e: # pragma: no cover
if "Error assigning an unit-key-id mapping" in str(e) and "" == str(unit):
log.warning(f"…skip {repr(unit)} (ixmp.JDBCBackend with Oracle database)")
else:
raise
# FIXME Reduce complexity from 14 to ≤13
[docs]def apply_spec( # noqa: C901
scenario: Scenario,
spec: Union[Spec, Mapping[str, ScenarioInfo]],
data: Optional[Callable] = None,
**options,
):
"""Apply `spec` to `scenario`.
Parameters
----------
spec : .Spec
Specification of changes to make to `scenario`.
data : callable, optional
Function to add data to `scenario`. `data` can either manipulate the scenario
directly, or return a :class:`dict` compatible with :func:`.add_par_data`.
Other parameters
----------------
dry_run : bool
Don't modify `scenario`; only show what would be done. Default :obj:`False`.
Exceptions will still be raised if the elements from ``spec['required']`` are
missing; this serves as a check that the scenario has the required features for
applying the spec.
fast : bool
Do not remove existing parameter data; increases speed on large scenarios.
quiet : bool
Only show log messages at level ``ERROR`` and higher. If :obj:`False` (default),
show log messages at level ``DEBUG`` and higher.
message : str
Commit message.
See also
--------
.add_par_data
.strip_par_data
.Code
.ScenarioInfo
"""
dry_run = options.get("dry_run", False)
fast = options.get("fast", False)
log.setLevel(logging.ERROR if options.get("quiet", False) else logging.DEBUG)
if not dry_run:
try:
scenario.remove_solution()
except ValueError:
pass
maybe_check_out(scenario)
dump: Dict[str, pd.DataFrame] = {} # Removed data
# Sort the list of sets by the number of dimensions; this places basic (non-indexed)
# sets first. Elements for these sets must be added before elements for indexed
# sets that may reference them.
sets = sorted((len(scenario.idx_sets(s)), s) for s in scenario.set_list())
# Existing 'region' codes stored on the Platform associated with `scenario`
platform_regions = set(scenario.platform.regions()["region"])
for _, set_name in sets:
# Check whether this set is mentioned at all in the spec
if 0 == sum(map(lambda info: len(info.set[set_name]), spec.values())):
# Not mentioned; don't do anything
continue
log.info(f"Set {repr(set_name)}")
# Base contents of the set
base_set = scenario.set(set_name)
# Unpack a multi-dimensional/indexed set to a list of tuples
base = (
list(base_set.itertuples(index=False))
if isinstance(base_set, pd.DataFrame)
else base_set.tolist()
)
log.info(f" {len(base)} elements")
# log.debug(', '.join(map(repr, base))) # All elements; verbose
# Check for required elements
require = spec["require"].set[set_name]
log.info(f" Check {len(require)} required elements")
# Raise an exception about the first missing element
missing = list(filter(lambda e: e not in base, require))
if missing:
log.error(f" {len(missing)} elements not found: {missing!r}")
raise ValueError
# Remove elements and associated parameter values
for element in spec["remove"].set[set_name]:
strip_par_data(
scenario,
set_name,
element,
dry_run=dry_run,
dump=None if fast else dump,
)
# Add elements
add = [] if dry_run else spec["add"].set[set_name]
for element in add:
name = element.id if isinstance(element, Code) else element
scenario.add_set(set_name, name)
if set_name == "node" and name not in platform_regions:
scenario.platform.add_region(name, "region")
if len(add):
log.info(f" Add {len(add)} element(s)")
log.debug(" " + ellipsize(add))
log.info(" ---")
if not fast:
N_removed = sum(len(d) for d in dump.values())
log.info(f"{N_removed} total rows removed")
# Add units to the Platform before adding data
for unit in spec["add"].set["unit"]:
unit = unit if isinstance(unit, Code) else Code(id=unit, name=unit)
_add_unit(scenario.platform, unit.id, str(unit.name))
# Add data
if callable(data):
result = data(scenario, dry_run=dry_run)
if result:
# `data` function returned some data; use add_par_data()
add_par_data(scenario, result, dry_run=dry_run)
# Finalize
log.info("Commit results.")
maybe_commit(
scenario,
condition=not dry_run,
message=options.get("message", f"{__name__}.apply_spec()"),
)
[docs]def ellipsize(elements: List) -> str:
"""Generate a short string representation of `elements`.
If the list has more than 5 elements, only the first two and last two are shown,
with "..." between.
"""
if len(elements) > 5:
return ", ".join(map(str, elements[:2] + ["..."] + elements[-2:]))
else:
return ", ".join(map(str, elements))