diff --git a/acmc/parse.py b/acmc/parse.py
index b1a541464253974cd14f2d9c2bc8a270bcb2e6e8..7c470dbefbc30b67fb1ace75dd16da5c87c4dcdc 100644
--- a/acmc/parse.py
+++ b/acmc/parse.py
@@ -10,15 +10,18 @@ logger = lc.setup_logger()
PHEN_CODE_ERROR_FILE = "code_errors.csv"
-class InvalidCodesException(Exception):
- """Custom exception class raised when invalid codes are found that cannot be resolved by processing"""
+class CodesError():
def __init__(self, message, codes=None, codes_file=None, mask=None, code_type=None):
- super().__init__(message)
-
# initialise class variables with provided parameters
for key, value in locals().items():
if key != "self":
setattr(self, key, value)
+
+class InvalidCodesException(Exception):
+ """Custom exception class raised when invalid codes are found that cannot be resolved by processing"""
+ def __init__(self, error):
+ super().__init__(error.message)
+ self.error = error
class Proto():
"""
@@ -66,7 +69,7 @@ class Proto():
codes = fix(codes, codes_file)
logger.debug(f"Check: Fixed")
except InvalidCodesException as ex:
- errors.append(ex)
+ errors.append(ex.error)
else:
logger.debug(f"Check: passed")
@@ -96,11 +99,11 @@ class Read2(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty {codes_file}",
+ InvalidCodesException(CodesError(f"Code list is empty",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
),
(
# check codes <5 characters, if too short pads it with . (dots) to reach 5 characters
@@ -119,23 +122,23 @@ class Read2(Proto):
"Alphanumeric Dot",
lambda codes : codes.str.match(r"^[a-zA-Z0-9.]+$"),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Illegal code format, not alphanumeric dot: {codes_file}",
+ InvalidCodesException(CodesError(f"Illegal code format, not alphanumeric dot",
codes=codes,
codes_file=codes_file,
mask=codes.str.match(r"^[a-zA-Z0-9.]+$"),
- code_type=self.name))
+ code_type=self.name)))
),
(
# checks code exists in self.db (the Read2 dataset). If missing log invalid codes.
"In Database",
lambda codes : self.in_database(codes, self.db, self.name),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Codes do not exist in database {codes_file}",
+ InvalidCodesException(CodesError(f"Codes do not exist in database",
codes=codes,
codes_file=codes_file,
mask=self.in_database(codes, self.db, self.name),
- code_type=self.name))
- ),
+ code_type=self.name)))
+ )
]
class Read3(Proto):
@@ -147,11 +150,11 @@ class Read3(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty {codes_file}",
+ InvalidCodesException(CodesError(f"Code list is empty",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
),
(
"Too Short",
@@ -167,21 +170,21 @@ class Read3(Proto):
"Alphanumeric Dot",
lambda codes : codes.str.match(r"^[a-zA-Z0-9.]+$"),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Alphanumeric Dot",
+ InvalidCodesException(CodesError(f"QA Alphanumeric Dot",
codes=codes,
codes_file=codes_file,
check_regex=codes.str.match(r"^[a-zA-Z0-9.]+$"),
- code_type=self.name))
+ code_type=self.name)))
),
(
"In Database",
lambda codes : self.in_database(codes, self.db, self.name),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA In Database",
+ InvalidCodesException(CodesError(f"QA In Database",
codes=codes,
codes_file=codes_file,
check_regex=self.in_database(codes, self.db, self.name),
- code_type=self.name))
+ code_type=self.name)))
),
]
@@ -194,21 +197,21 @@ class Icd10(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty {codes_file}",
+ InvalidCodesException(CodesError(f"Code list is empty {codes_file}",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
),
(
"Too Short",
lambda codes : ~(codes.str.len() < 3),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Too Short",
+ InvalidCodesException(CodesError(f"QA Too Short",
codes=codes,
codes_file=codes_file,
mask=~(codes.str.len() < 3),
- code_type=self.name))
+ code_type=self.name)))
),
(
"Has Dot",
@@ -220,21 +223,21 @@ class Icd10(Proto):
"Alphanumeric Capital",
lambda codes : codes.str.match(r"^[A-Z0-9]+$"),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Alphanumeric Capital",
+ InvalidCodesException(CodesError(f"QA Alphanumeric Capital",
codes=codes,
codes_file=codes_file,
mask=codes.str.match(r"^[A-Z0-9]+$"),
- code_type=self.name))
+ code_type=self.name)))
),
(
"In Database",
lambda codes : ~(~self.in_database(codes, self.db, self.name) & ~self.in_database(codes, self.db, self.name + "_alt")),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA In Database",
+ InvalidCodesException(CodesError(f"QA In Database",
codes=codes,
codes_file=codes_file,
mask=~(~self.in_database(codes, self.db, self.name) & ~self.in_database(codes, self.db, self.name+"_alt")),
- code_type=self.name))
+ code_type=self.name)))
)
# (
# "ICD10 Regex",
@@ -265,31 +268,31 @@ class Snomed(Proto):
"Too Short",
lambda codes : ~(codes.str.len() < 6),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Too Short",
+ InvalidCodesException(CodesError(f"QA Too Short",
codes=codes,
codes_file=codes_file,
mask=~(codes.str.len() < 6),
- code_type=self.name))
+ code_type=self.name)))
),
(
"Too Long",
lambda codes : ~(codes.str.len() > 18),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Too Long",
+ InvalidCodesException(CodesError(f"QA Too Long",
codes=codes,
codes_file=codes_file,
mask=~(codes.str.len() > 18),
- code_type=self.name))
+ code_type=self.name)))
),
(
"Numeric",
lambda codes : codes.str.match(r"[0-9]+$"),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Numeric",
+ InvalidCodesException(CodesError(f"QA Numeric",
codes=codes,
codes_file=codes_file,
mask=codes.str.match(r"[0-9]+$"),
- code_type=self.name))
+ code_type=self.name)))
),
# (
# "Is Integer",
@@ -300,11 +303,11 @@ class Snomed(Proto):
"In Database",
lambda codes : self.in_database(codes, self.db, self.name),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA In Database",
+ InvalidCodesException(CodesError(f"QA In Database",
codes=codes,
codes_file=codes_file,
mask=self.in_database(codes, self.db, self.name),
- code_type=self.name))
+ code_type=self.name)))
)
]
@@ -317,21 +320,21 @@ class Opcs4(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty {codes_file}",
+ InvalidCodesException(CodesError(f"Code list is empty",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
),
(
"In Database",
lambda codes : self.in_database(codes, self.db, self.name),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA In Database",
+ InvalidCodesException(CodesError(f"QA In Database",
codes=codes,
codes_file=codes_file,
mask=self.in_database(codes, self.db, self.name),
- code_type=self.name))
+ code_type=self.name)))
)
]
@@ -343,21 +346,21 @@ class Atc(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty",
+ InvalidCodesException(CodesError(f"Code list is empty",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
),
(
"Alphanumeric Capital",
lambda codes : codes.str.match(r"^[A-Z0-9]+$"),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"QA Alphanumeric Capital",
+ InvalidCodesException(CodesError(f"QA Alphanumeric Capital",
codes=codes,
codes_file=codes_file,
mask=codes.str.match(r"^[A-Z0-9]+$"),
- code_type=self.name))
+ code_type=self.name)))
),
]
@@ -369,11 +372,11 @@ class Med(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty {codes_file}",
+ InvalidCodesException(CodesError(f"Code list is empty",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
)
]
@@ -385,11 +388,11 @@ class Cprd(Proto):
"Not Empty",
lambda codes : pd.Series([len(codes) > 0]),
lambda codes, codes_file : self.raise_exception(
- InvalidCodesException(f"Code list is empty {codes_file}",
+ InvalidCodesException(CodesError(f"Code list is empty",
codes=codes,
codes_file=codes_file,
mask=None,
- code_type=self.name))
+ code_type=self.name)))
)
]
diff --git a/acmc/phen.py b/acmc/phen.py
index df354e2ae48ff5eacfc54dda13dbc7ddc275316e..41c3520effa300d6c9200a77d758754ba59773ae 100644
--- a/acmc/phen.py
+++ b/acmc/phen.py
@@ -328,28 +328,6 @@ def process_actions(df, file):
return df
-def log_invalid_code(codes, mask, code_type=None, file_path=None, cause=None):
- # print("ERROR WITH CODES", file_path, codes[~mask])
-
- errors = pd.DataFrame([])
- errors["CONCEPT"] = codes[~mask].astype(str)
- errors["VOCABULARY"] = code_type
- errors["SOURCE"] = file_path
- errors["CAUSE"] = cause
-
- #append to error log csv
- if os.path.exists(log_errors_path):
- print("FILE EXISTS")
- df_error = pd.read_csv(log_errors_path)
- df_error = pd.concat([df_error, errors])
- df_error.to_csv(log_errors_path, index=False)
- else:
- print("FILE NOT EXIST")
- df_error = errors
- df_error.to_csv(log_errors_path, index=False)
-
- return codes[mask]
-
# Perform QA Checks on columns individually and append to df
def preprocess_codes(df, file, target_code_type=None, codes_file=None):
""" Parses each column individually - Order and length will not be preserved! """
@@ -378,8 +356,8 @@ def preprocess_codes(df, file, target_code_type=None, codes_file=None):
# process codes, validating them using parser and returning the errors
codes, errors = code_type_parser.process(codes, codes_file)
if len(errors) > 0:
- code_errors = code_errors.append(errors)
- logger.warning(f"Code validation failed with {len(errors)} errors")
+ code_errors.extend(errors)
+ logger.warning(f"Codes validation failed with {len(errors)} errors")
# add metadata columns
out = pd.concat([out, pd.DataFrame({code_type_name: codes}).join(metadata_df)], ignore_index=True)
@@ -443,6 +421,17 @@ def sql_row_exist(conn, table, column, value):
return exists
+def write_code_errors(code_errors, code_errors_path):
+ err_df = pd.DataFrame([
+ {"CONCEPT": ", ".join(err.codes[~err.mask].tolist()),
+ "VOCABULARY": err.code_type,
+ "SOURCE": err.codes_file,
+ "CAUSE": err.message} for err in code_errors])
+
+ err_df = err_df.drop_duplicates() # Remove Duplicates from Error file
+ err_df = err_df.sort_values(by=["SOURCE", "VOCABULARY", "CONCEPT"])
+ err_df.to_csv(code_errors_path, index=False, mode="w")
+
def map(phen_dir, target_code_type):
logger.info(f"Processing phenotype: {phen_dir}")
logger.debug(f"Target coding format: {target_code_type}")
@@ -462,7 +451,7 @@ def map(phen_dir, target_code_type):
# Create output dataframe
out = pd.DataFrame([])
- code_errors []
+ code_errors = []
# Process each folder in codes section
for folder in codes:
@@ -486,7 +475,10 @@ def map(phen_dir, target_code_type):
file, codes_file=str(codes_file_path.resolve()),
target_code_type=target_code_type)
- code_errors = code_errors.append(errors)
+ logger.debug(f" Length of errors from preprocess {len(errors)}")
+ if len(errors) > 0:
+ code_errors.extend(errors)
+ logger.debug(f" Length of code_errors {len(code_errors)}")
# partition table by categorical column
if ("actions" in file and "divide_col" in file["actions"] and len(df) > 0):
@@ -521,10 +513,13 @@ def map(phen_dir, target_code_type):
logger.warning(f"File {file} has no output after preprocessing in config {str(config_path.resolve())}")
if(len(code_errors) > 0):
- logger.error(f"The map processing has {len(code_errors)} errors)
+ logger.error(f"The map processing has {len(code_errors)} errors")
+ error_filename = f"{target_code_type}-code-errors.csv"
+ write_code_errors(code_errors, phen_path / MAP_DIR / error_filename)
# Check there is output from processing
if len(out.index) == 0:
+ logger.error(f"No output after map processing")
raise Exception(f"No output after map processing, check config {str(config_path.resolve())}")
# Final processing
@@ -543,9 +538,7 @@ def map(phen_dir, target_code_type):
# Save output to map directory
output_filename = target_code_type + '.csv'
-
map_path = phen_path / MAP_DIR / output_filename
-
out.to_csv(map_path, index=False)
logger.info(f"Saved mapped concepts to {str(map_path.resolve())}")
@@ -574,8 +567,6 @@ def map(phen_dir, target_code_type):
shutil.copy(trud.VERSION_PATH, phen_path / trud.VERSION_FILE)
shutil.copy(omop.VERSION_PATH, phen_path / omop.VERSION_FILE)
- logger.debug(f"Saved concept_sets to {str(concept_set_path.resolve())}")
-
logger.info(f"Phenotype processed successfully")
def publish(phen_dir):