def install(benchmark, solver_names, dataset_names, force=False,
recreate=False, env_name='False', confirm=False):
# Check that the dataset/solver patterns match actual dataset
benchmark = Benchmark(benchmark)
print(f"Installing '{benchmark.name}' requirements")
benchmark.validate_dataset_patterns(dataset_names)
benchmark.validate_solver_patterns(solver_names)
# Get a list of all conda envs
default_conda_env, conda_envs = list_conda_envs()
# If env_name is False (default), installation in the current environement.
if env_name == 'False':
env_name = None
# incompatible with the 'recreate' flag to avoid messing with the
# user environement
if recreate:
msg = "Cannot recreate conda env without using options " + \
"'-e/--env' or '--env-name'."
raise RuntimeError(msg)
# check if any current conda environment
if default_conda_env is not None:
# ask for user confirmation to install in current conda env
if not confirm:
click.confirm(
f"Install in the current env '{default_conda_env}'?",
abort=True
)
else:
raise RuntimeError("No conda environment is activated.")
else:
# If env_name is True, the flag `--env` has been used. Create a conda
# env specific to the benchmark. Else, use the <env_name> value.
if env_name == 'True':
env_name = f"benchopt_{benchmark.name}"
else:
# check provided <env_name>
# (to avoid empty name like `--env-name ""`)
if len(env_name) == 0:
raise RuntimeError("Empty environment name.")
# avoid recreating 'base' conda env`
if env_name == 'base' and recreate:
raise RuntimeError(
"Impossible to recreate 'base' conda environment."
)
# create environment if necessary
create_conda_env(env_name, recreate=recreate)
# If force is True (default is False), it forces the reinstallation of
# selected solvers (all solvers from the benchmark by default)
forced_solvers = ()
if force:
if len(solver_names) > 0:
forced_solvers = solver_names
else:
forced_solvers = benchmark.list_benchmark_solver_names()
# same for datasets
forced_datasets = ()
if force:
if len(dataset_names) > 0:
forced_datasets = dataset_names
else:
forced_datasets = benchmark.list_benchmark_dataset_names()
# install required datasets
print("# Datasets", flush=True)
benchmark.install_required_datasets(
dataset_names, forced_datasets=forced_datasets, env_name=env_name
)
# install required solvers
print("# Solvers", flush=True)
benchmark.install_required_solvers(
solver_names, forced_solvers=forced_solvers, env_name=env_name
)
def get_datasets(ctx, args, incomplete):
skip_import()
benchmark = Benchmark(args[1])
datasets = benchmark.list_benchmark_dataset_names()
datasets = [d.lower() for d in datasets]
return [d for d in datasets if incomplete.lower() in d]