# quality: peripheral
import datetime
import re
from pathlib import Path
import yaml
from yaml import MarkedYAMLError
from immuneML.dsl.InstructionParser import InstructionParser
from immuneML.dsl.OutputParser import OutputParser
from immuneML.dsl.definition_parsers.DefinitionParser import DefinitionParser
from immuneML.dsl.symbol_table.SymbolTable import SymbolTable
from immuneML.util.PathBuilder import PathBuilder
[docs]class ImmuneMLParser:
"""
Simple DSL parser from python dictionary or equivalent YAML for configuring repertoire / receptor_sequence
classification in the (simulated) settings
DSL example with hyper-parameter optimization:
.. highlight:: yaml
.. code-block:: yaml
definitions:
datasets:
d1:
format: MiXCR
params:
result_path: loaded_dataset/
region_type: IMGT_CDR3
path: path_to_files/
metadata_file: metadata.csv
encodings:
e1:
KmerFrequency
k: 3
e2:
Word2Vec:
vector_size: 16
context: sequence
ml_methods:
log_reg1:
LogisticRegression:
C: 0.001
reports:
r1:
SequenceLengthDistribution
preprocessing_sequences:
seq1:
- filter_chain_B:
ChainRepertoireFilter:
keep_chain: A
- filter_clonotype:
ClonesPerRepertoireFilter:
lower_limit: 1000
seq2:
- filter_clonotype:
ClonesPerRepertoireFilter:
lower_limit: 500
- filter_chain_A:
ChainRepertoireFilter:
keep_chain: B
instructions:
inst1:
type: TrainMLModel
settings:
- preprocessing: seq1
encoding: e1
ml_method: log_reg1
- preprocessing: seq2
encoding: e2
ml_method: log_reg1
assessment:
split_strategy: random
split_count: 1
training_percentage: 70
reports:
data: []
data_splits: []
encoding: []
models: []
selection:
split_strategy: k-fold
split_count: 5
reports:
data: []
data_splits: [r1]
encoding: []
models: []
labels:
- CD
dataset: d1
strategy: GridSearch
metrics: [accuracy, f1_micro]
optimization_metric: balanced_accuracy
reports: []
output: # this section can also be omitted, in that case output will be automatically HTML
format: HTML # or None
"""
[docs] @staticmethod
def parse_yaml_file(file_path: Path, result_path: Path = None, parse_func=None):
try:
with file_path.open("r") as file:
workflow_specification = yaml.safe_load(file)
ImmuneMLParser.check_keys(workflow_specification)
except yaml.YAMLError as exc:
problem_description = "\n--------------------------------------------------------------------------------\n" \
"There was a YAML formatting error in the supplied specification file. Please validate specification " \
"(you can use https://jsonformatter.org/yaml-validator) and try again."
raise MarkedYAMLError(context=str(exc), problem=problem_description, problem_mark=f"The error was {exc.problem_mark}.")
try:
if parse_func is None:
symbol_table, path = ImmuneMLParser.parse(workflow_specification, file_path, result_path)
else:
symbol_table, path = parse_func(workflow_specification, file_path, result_path)
except KeyError as key_error:
raise Exception(f"ImmuneMLParser: an error occurred during parsing the YAML specification. Missing key was '{key_error.args[0]}'. "
f"For more details, refer to the log above and check the documentation.") from key_error
return symbol_table, path
[docs] @staticmethod
def check_keys(specs: dict):
for key in specs.keys():
key_to_check = str(key)
assert re.match(r'^[A-Za-z0-9_]+$', key_to_check), \
f"ImmuneMLParser: the keys in the specification can contain only letters, numbers and underscore. Error with key: {key}"
if isinstance(specs[key], dict) and key not in ["column_mapping", "metadata_column_mapping"]:
ImmuneMLParser.check_keys(specs[key])
[docs] @staticmethod
def parse(workflow_specification: dict, file_path, result_path):
symbol_table = SymbolTable()
def_parser_output, specs_defs = DefinitionParser.parse(workflow_specification, symbol_table, result_path)
symbol_table, specs_instructions = InstructionParser.parse(def_parser_output, result_path)
app_output = OutputParser.parse(workflow_specification, symbol_table)
path = ImmuneMLParser._output_specs(file_path=file_path, result_path=result_path, definitions=specs_defs,
instructions=specs_instructions, output=app_output)
return symbol_table, path
@staticmethod
def _get_full_specs_filepath(file_path, result_path) -> Path:
file_name = f"full_{file_path.stem}.yaml"
if result_path is None:
folder = file_path.absolute().parent
return folder / file_name
else:
return result_path / file_name
@staticmethod
def _output_specs(file_path=None, result_path=None, definitions: dict = None, instructions: dict = None, output: dict = None) -> Path:
filepath = ImmuneMLParser._get_full_specs_filepath(file_path, result_path)
result = {"definitions": definitions, "instructions": instructions, "output": output}
result = ImmuneMLParser._paths_to_strings_recursive(result)
PathBuilder.build_from_objects(filepath.parent)
with filepath.open("w") as file:
yaml.dump(result, file)
print(f"{datetime.datetime.now()}: Full specification is available at {filepath}.\n", flush=True)
return filepath
@staticmethod
def _paths_to_strings_recursive(specs):
if isinstance(specs, Path):
return specs.as_posix()
elif isinstance(specs, dict):
return {key: ImmuneMLParser._paths_to_strings_recursive(value) for key, value in specs.items()}
elif isinstance(specs, list):
return [ImmuneMLParser._paths_to_strings_recursive(item) for item in specs]
else:
return specs