def test_unicode(self):
"""Check that there are no exceptions raised when using unicode folders."""
tmpsource = tempfile.mkdtemp()
tmpdest = tempfile.mkdtemp()
with open(os.path.join(tmpsource, 'sąžininga'), 'w',
encoding='utf8') as fhandle:
fhandle.write('test')
with open(os.path.join(tmpsource, 'žąsis'), 'w',
encoding='utf8') as fhandle:
fhandle.write('test')
folder = Folder(tmpdest)
folder.insert_path(tmpsource, 'destination')
folder.insert_path(tmpsource, 'šaltinis')
self.assertEqual(sorted(folder.get_content_list()),
sorted(['destination', 'šaltinis']))
self.assertEqual(
sorted(folder.get_subfolder('destination').get_content_list()),
sorted(['sąžininga', 'žąsis']))
self.assertEqual(
sorted(folder.get_subfolder('šaltinis').get_content_list()),
sorted(['sąžininga', 'žąsis']))
folder = Folder(os.path.join(tmpsource, 'šaltinis'))
folder.insert_path(tmpdest, 'destination')
folder.insert_path(tmpdest, 'kitas-šaltinis')
self.assertEqual(sorted(folder.get_content_list()),
sorted(['destination', 'kitas-šaltinis']))
def presubmit(self, folder: Folder) -> CalcInfo:
"""Prepares the calculation folder with all inputs, ready to be copied to the cluster.
:param folder: a SandboxFolder that can be used to write calculation input files and the scheduling script.
:return calcinfo: the CalcInfo object containing the information needed by the daemon to handle operations.
"""
# pylint: disable=too-many-locals,too-many-statements,too-many-branches
from aiida.common.exceptions import PluginInternalError, ValidationError, InvalidOperation, InputValidationError
from aiida.common import json
from aiida.common.utils import validate_list_of_string_tuples
from aiida.common.datastructures import CodeInfo, CodeRunMode
from aiida.orm import load_node, Code, Computer
from aiida.plugins import DataFactory
from aiida.schedulers.datastructures import JobTemplate
computer = self.node.computer
inputs = self.node.get_incoming(link_type=LinkType.INPUT_CALC)
if not self.inputs.metadata.dry_run and self.node.has_cached_links(
): # type: ignore[union-attr]
raise InvalidOperation(
'calculation node has unstored links in cache')
codes = [_ for _ in inputs.all_nodes() if isinstance(_, Code)]
for code in codes:
if not code.can_run_on(computer):
raise InputValidationError(
'The selected code {} for calculation {} cannot run on computer {}'
.format(code.pk, self.node.pk, computer.label))
if code.is_local() and code.get_local_executable(
) in folder.get_content_list():
raise PluginInternalError(
f'The plugin created a file {code.get_local_executable()} that is also the executable name!'
)
calc_info = self.prepare_for_submission(folder)
calc_info.uuid = str(self.node.uuid)
scheduler = computer.get_scheduler()
# I create the job template to pass to the scheduler
job_tmpl = JobTemplate()
job_tmpl.shebang = computer.get_shebang()
job_tmpl.submit_as_hold = False
job_tmpl.rerunnable = False
job_tmpl.job_environment = {}
# 'email', 'email_on_started', 'email_on_terminated',
job_tmpl.job_name = f'aiida-{self.node.pk}'
job_tmpl.sched_output_path = self.options.scheduler_stdout
if self.options.scheduler_stderr == self.options.scheduler_stdout:
job_tmpl.sched_join_files = True
else:
job_tmpl.sched_error_path = self.options.scheduler_stderr
job_tmpl.sched_join_files = False
# Set retrieve path, add also scheduler STDOUT and STDERR
retrieve_list = calc_info.retrieve_list or []
if (job_tmpl.sched_output_path is not None
and job_tmpl.sched_output_path not in retrieve_list):
retrieve_list.append(job_tmpl.sched_output_path)
if not job_tmpl.sched_join_files:
if (job_tmpl.sched_error_path is not None
and job_tmpl.sched_error_path not in retrieve_list):
retrieve_list.append(job_tmpl.sched_error_path)
retrieve_list.extend(
self.node.get_option('additional_retrieve_list') or [])
self.node.set_retrieve_list(retrieve_list)
retrieve_singlefile_list = calc_info.retrieve_singlefile_list or []
# a validation on the subclasses of retrieve_singlefile_list
for _, subclassname, _ in retrieve_singlefile_list:
file_sub_class = DataFactory(subclassname)
if not issubclass(file_sub_class, orm.SinglefileData):
raise PluginInternalError(
'[presubmission of calc {}] retrieve_singlefile_list subclass problem: {} is '
'not subclass of SinglefileData'.format(
self.node.pk, file_sub_class.__name__))
if retrieve_singlefile_list:
self.node.set_retrieve_singlefile_list(retrieve_singlefile_list)
# Handle the retrieve_temporary_list
retrieve_temporary_list = calc_info.retrieve_temporary_list or []
self.node.set_retrieve_temporary_list(retrieve_temporary_list)
# the if is done so that if the method returns None, this is
# not added. This has two advantages:
# - it does not add too many \n\n if most of the prepend_text are empty
# - most importantly, skips the cases in which one of the methods
# would return None, in which case the join method would raise
# an exception
prepend_texts = [computer.get_prepend_text()] + \
[code.get_prepend_text() for code in codes] + \
[calc_info.prepend_text, self.node.get_option('prepend_text')]
job_tmpl.prepend_text = '\n\n'.join(prepend_text
for prepend_text in prepend_texts
if prepend_text)
append_texts = [self.node.get_option('append_text'), calc_info.append_text] + \
[code.get_append_text() for code in codes] + \
[computer.get_append_text()]
job_tmpl.append_text = '\n\n'.join(append_text
for append_text in append_texts
if append_text)
# Set resources, also with get_default_mpiprocs_per_machine
resources = self.node.get_option('resources')
scheduler.preprocess_resources(
resources, computer.get_default_mpiprocs_per_machine())
job_tmpl.job_resource = scheduler.create_job_resource(**resources)
subst_dict = {
'tot_num_mpiprocs': job_tmpl.job_resource.get_tot_num_mpiprocs()
}
for key, value in job_tmpl.job_resource.items():
subst_dict[key] = value
mpi_args = [
arg.format(**subst_dict) for arg in computer.get_mpirun_command()
]
extra_mpirun_params = self.node.get_option(
'mpirun_extra_params') # same for all codes in the same calc
# set the codes_info
if not isinstance(calc_info.codes_info, (list, tuple)):
raise PluginInternalError(
'codes_info passed to CalcInfo must be a list of CalcInfo objects'
)
codes_info = []
for code_info in calc_info.codes_info:
if not isinstance(code_info, CodeInfo):
raise PluginInternalError(
'Invalid codes_info, must be a list of CodeInfo objects')
if code_info.code_uuid is None:
raise PluginInternalError(
'CalcInfo should have the information of the code to be launched'
)
this_code = load_node(code_info.code_uuid, sub_classes=(Code, ))
this_withmpi = code_info.withmpi # to decide better how to set the default
if this_withmpi is None:
if len(calc_info.codes_info) > 1:
raise PluginInternalError(
'For more than one code, it is necessary to set withmpi in codes_info'
)
else:
this_withmpi = self.node.get_option('withmpi')
if this_withmpi:
this_argv = (mpi_args + extra_mpirun_params +
[this_code.get_execname()] +
(code_info.cmdline_params
if code_info.cmdline_params is not None else []))
else:
this_argv = [this_code.get_execname()
] + (code_info.cmdline_params if
code_info.cmdline_params is not None else [])
# overwrite the old cmdline_params and add codename and mpirun stuff
code_info.cmdline_params = this_argv
codes_info.append(code_info)
job_tmpl.codes_info = codes_info
# set the codes execution mode
if len(codes) > 1:
try:
job_tmpl.codes_run_mode = calc_info.codes_run_mode
except KeyError as exc:
raise PluginInternalError(
'Need to set the order of the code execution (parallel or serial?)'
) from exc
else:
job_tmpl.codes_run_mode = CodeRunMode.SERIAL
########################################################################
custom_sched_commands = self.node.get_option(
'custom_scheduler_commands')
if custom_sched_commands:
job_tmpl.custom_scheduler_commands = custom_sched_commands
job_tmpl.import_sys_environment = self.node.get_option(
'import_sys_environment')
job_tmpl.job_environment = self.node.get_option(
'environment_variables')
queue_name = self.node.get_option('queue_name')
account = self.node.get_option('account')
qos = self.node.get_option('qos')
if queue_name is not None:
job_tmpl.queue_name = queue_name
if account is not None:
job_tmpl.account = account
if qos is not None:
job_tmpl.qos = qos
priority = self.node.get_option('priority')
if priority is not None:
job_tmpl.priority = priority
max_memory_kb = self.node.get_option('max_memory_kb')
if max_memory_kb is not None:
job_tmpl.max_memory_kb = max_memory_kb
max_wallclock_seconds = self.node.get_option('max_wallclock_seconds')
if max_wallclock_seconds is not None:
job_tmpl.max_wallclock_seconds = max_wallclock_seconds
max_memory_kb = self.node.get_option('max_memory_kb')
if max_memory_kb is not None:
job_tmpl.max_memory_kb = max_memory_kb
submit_script_filename = self.node.get_option('submit_script_filename')
script_content = scheduler.get_submit_script(job_tmpl)
folder.create_file_from_filelike(io.StringIO(script_content),
submit_script_filename,
'w',
encoding='utf8')
subfolder = folder.get_subfolder('.aiida', create=True)
subfolder.create_file_from_filelike(io.StringIO(json.dumps(job_tmpl)),
'job_tmpl.json',
'w',
encoding='utf8')
subfolder.create_file_from_filelike(io.StringIO(json.dumps(calc_info)),
'calcinfo.json',
'w',
encoding='utf8')
if calc_info.local_copy_list is None:
calc_info.local_copy_list = []
if calc_info.remote_copy_list is None:
calc_info.remote_copy_list = []
# Some validation
this_pk = self.node.pk if self.node.pk is not None else '[UNSTORED]'
local_copy_list = calc_info.local_copy_list
try:
validate_list_of_string_tuples(local_copy_list, tuple_length=3)
except ValidationError as exception:
raise PluginInternalError(
f'[presubmission of calc {this_pk}] local_copy_list format problem: {exception}'
) from exception
remote_copy_list = calc_info.remote_copy_list
try:
validate_list_of_string_tuples(remote_copy_list, tuple_length=3)
except ValidationError as exception:
raise PluginInternalError(
f'[presubmission of calc {this_pk}] remote_copy_list format problem: {exception}'
) from exception
for (remote_computer_uuid, _, dest_rel_path) in remote_copy_list:
try:
Computer.objects.get(uuid=remote_computer_uuid) # pylint: disable=unused-variable
except exceptions.NotExistent as exception:
raise PluginInternalError(
'[presubmission of calc {}] '
'The remote copy requires a computer with UUID={}'
'but no such computer was found in the '
'database'.format(this_pk,
remote_computer_uuid)) from exception
if os.path.isabs(dest_rel_path):
raise PluginInternalError(
'[presubmission of calc {}] '
'The destination path of the remote copy '
'is absolute! ({})'.format(this_pk, dest_rel_path))
return calc_info
def prepare_for_submission(self, folder: folders.Folder):
"""Create input files from the input nodes passed to this instance of the `CalcJob`.
:param folder: an `aiida.common.folders.Folder` to temporarily write files on disk
:return: `aiida.common.datastructures.CalcInfo` instance
"""
# To be filled out below
local_copy_list = []
remote_copy_list = []
remote_symlink_list = []
# Create the subfolders for pseudopotentials and orbitals
folder.get_subfolder(self._PSEUDO_SUBFOLDER, create=True)
folder.get_subfolder(self._ORBITAL_SUBFOLDER, create=True)
# Get an uppercase-key-only version of the settings dictionary (also check for case-insensitive duplicates)
if 'settings' in self.inputs:
settings = uppercase_dict_keys(self.inputs.settings.get_dict(),
dict_name='settings')
else:
settings = {}
# Get an uppercase-key-only verion of the parameters dictionary (also check for case-insensitive duplicates)
parameters = uppercase_dict_keys(self.inputs.parameters.get_dict(),
dict_name='parameters')
# No reserved parameter keywords should be provided
self._check_reserved_keywords(parameters)
# Load parameter schema
with open(self._INPUT_SCHEMA, 'r') as stream:
schema = json.load(stream)
# Automatically generate input parameters for derived fields, e.g. structure -> Atoms.Unitvectors, etc.
parameters = self._generate_input_parameters(
self.inputs.structure, self.inputs.kpoints, parameters,
self.inputs.pseudos, self.inputs.orbitals,
self.inputs.orbital_configurations)
# Get a lowercase-value-only version of the parameters dictionary
parameters = lowercase_dict_values(parameters)
# Validate input parameters
self._validate_inputs(self.inputs.structure, self.inputs.kpoints,
parameters, self.inputs.pseudos,
self.inputs.orbitals,
self.inputs.orbital_configurations, schema)
# Get input file contents and lists of the pseudopotential and orbital files which need to be copied
input_file_content = write_input_file(parameters, schema)
local_copy_pseudo_list, local_copy_orbital_list = self._generate_local_copy_lists(
self.inputs.pseudos, self.inputs.orbitals)
local_copy_list += local_copy_pseudo_list
local_copy_list += local_copy_orbital_list
# Add output files to retrieve which have been specified to write in the input parameters
retrieve_list = []
if parameters.get('BAND_NKPATH', 0) > 0 and parameters.get(
'SCF_EIGENVALUESOLVER', 'band') == 'band':
retrieve_list.append(self._DATAFILE_BAND_FILE)
if parameters.get('MD_TYPE', 'nomd') != 'nomd':
retrieve_list.append(self._DATAFILE_MD_FILE)
retrieve_list.append(self._DATAFILE_MD2_FILE)
# Write input file
with folder.open(self._INPUT_FILE, 'w') as handle:
handle.write(input_file_content)
# Fill out the `CodeInfo`
codeinfo = datastructures.CodeInfo()
codeinfo.code_uuid = self.inputs.code.uuid
codeinfo.withmpi = True
codeinfo.cmdline_params = ([self._INPUT_FILE] +
list(settings.pop('CMDLINE', [])))
codeinfo.stdout_name = self._OUTPUT_FILE
# Fill out the `CalcInfo`
calcinfo = datastructures.CalcInfo()
calcinfo.uuid = str(self.uuid)
calcinfo.codes_info = [codeinfo]
calcinfo.local_copy_list = local_copy_list
calcinfo.remote_copy_list = remote_copy_list
calcinfo.remote_symlink_list = remote_symlink_list
calcinfo.retrieve_list = retrieve_list
calcinfo.retrieve_list.append(self._OUTPUT_FILE)
calcinfo.retrieve_list += settings.pop('ADDITIONAL_RETRIEVE_LIST', [])
# TODO: pop parser settings and report remaining unknown settings
return calcinfo