mirror of
https://github.com/DOI-DO/j40-cejst-2.git
synced 2025-07-23 10:10:17 -07:00
Add ETL Contract Checks (#619)
* Adds dev dependencies to requirements.txt and re-runs black on codebase * Adds test and code for national risk index etl, still in progress * Removes test_data from .gitignore * Adds test data to nation_risk_index tests * Creates tests and ETL class for NRI data * Adds tests for load() and transform() methods of NationalRiskIndexETL * Updates README.md with info about the NRI dataset * Adds to dos * Moves tests and test data into a tests/ dir in national_risk_index * Moves tmp_dir for tests into data/tmp/tests/ * Promotes fixtures to conftest and relocates national_risk_index tests: The relocation of national_risk_index tests is necessary because tests can only use fixtures specified in conftests within the same package * Fixes issue with df.equals() in test_transform() * Files reformatted by black * Commit changes to other files after re-running black * Fixes unused import that caused lint checks to fail * Moves tests/ directory to app root for data_pipeline * Adds new methods to ExtractTransformLoad base class: - __init__() Initializes class attributes - _get_census_fips_codes() Loads a dataframe with the fips codes for census block group and tract - validate_init() Checks that the class was initialized correctly - validate_output() Checks that the output was loaded correctly * Adds test for ExtractTransformLoad.__init__() and base.py * Fixes failing flake8 test * Changes geo_col to geoid_col and changes is_dataset to is_census in yaml * Adds test for validate_output() * Adds remaining tests * Removes is_dataset from init method * Makes CENSUS_CSV a class attribute instead of a class global: This ensures that CENSUS_CSV is only set when the ETL class is for a non-census dataset and removes the need to overwrite the value in mock_etl fixture * Re-formats files with black and fixes broken tox tests
This commit is contained in:
parent
1f78920f63
commit
d1273b63c5
13 changed files with 358 additions and 32 deletions
|
@ -1,8 +1,17 @@
|
|||
from pathlib import Path
|
||||
from typing import Optional
|
||||
|
||||
import pandas as pd
|
||||
import yaml
|
||||
|
||||
from data_pipeline.config import settings
|
||||
from data_pipeline.utils import unzip_file_from_url, remove_all_from_dir
|
||||
from data_pipeline.utils import (
|
||||
unzip_file_from_url,
|
||||
remove_all_from_dir,
|
||||
get_module_logger,
|
||||
)
|
||||
|
||||
logger = get_module_logger(__name__)
|
||||
|
||||
|
||||
class ExtractTransformLoad:
|
||||
|
@ -17,7 +26,8 @@ class ExtractTransformLoad:
|
|||
GEOID_TRACT_FIELD_NAME (str): The common column name for a Census Tract identifier
|
||||
"""
|
||||
|
||||
DATA_PATH: Path = settings.APP_ROOT / "data"
|
||||
APP_ROOT: Path = settings.APP_ROOT
|
||||
DATA_PATH: Path = APP_ROOT / "data"
|
||||
TMP_PATH: Path = DATA_PATH / "tmp"
|
||||
FILES_PATH: Path = settings.APP_ROOT / "files"
|
||||
GEOID_FIELD_NAME: str = "GEOID10"
|
||||
|
@ -26,11 +36,51 @@ class ExtractTransformLoad:
|
|||
EXPECTED_MAX_CENSUS_BLOCK_GROUPS: int = 220405
|
||||
EXPECTED_MAX_CENSUS_TRACTS: int = 73076
|
||||
|
||||
def get_yaml_config(self) -> None:
|
||||
def __init__(self, config_path: Path) -> None:
|
||||
"""Inits the class with instance specific variables"""
|
||||
|
||||
# set by _get_yaml_config()
|
||||
self.NAME: str = None
|
||||
self.SOURCE_URL: str = None
|
||||
self.GEOID_COL: str = None
|
||||
self.GEO_LEVEL: str = None
|
||||
self.SCORE_COLS: list = None
|
||||
self.FIPS_CODES: pd.DataFrame = None
|
||||
self.OUTPUT_PATH: Path = None
|
||||
self.CENSUS_CSV: Path = None
|
||||
|
||||
self._get_yaml_config(config_path)
|
||||
|
||||
def _get_yaml_config(self, config_path: Path) -> None:
|
||||
"""Reads the YAML configuration file for the dataset and stores
|
||||
the properies in the instance (upcoming feature)"""
|
||||
# parse the yaml config file
|
||||
try:
|
||||
with open(config_path, "r", encoding="utf-8") as file:
|
||||
config = yaml.safe_load(file)
|
||||
except (FileNotFoundError, yaml.YAMLError) as err:
|
||||
raise err
|
||||
|
||||
pass
|
||||
# set dataset specific attributes
|
||||
census_dir = self.DATA_PATH / "census" / "csv"
|
||||
if config["is_census"]:
|
||||
csv_dir = census_dir
|
||||
else:
|
||||
self.CENSUS_CSV = census_dir / "us.csv"
|
||||
self.FIPS_CODES = self._get_census_fips_codes()
|
||||
csv_dir = self.DATA_PATH / "dataset"
|
||||
|
||||
# parse name and set output path
|
||||
name = config.get("name")
|
||||
snake_name = name.replace(" ", "_").lower() # converts to snake case
|
||||
output_dir = snake_name + (config.get("year") or "")
|
||||
self.OUTPUT_PATH = csv_dir / output_dir / "usa.csv"
|
||||
self.OUTPUT_PATH.parent.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
# set class attributes
|
||||
attrs = ["NAME", "SOURCE_URL", "GEOID_COL", "GEO_LEVEL", "SCORE_COLS"]
|
||||
for attr in attrs:
|
||||
setattr(self, attr, config[attr.lower()])
|
||||
|
||||
def check_ttl(self) -> None:
|
||||
"""Checks if the ETL process can be run based on a the TLL value on the
|
||||
|
@ -44,9 +94,9 @@ class ExtractTransformLoad:
|
|||
extract_path: Path = None,
|
||||
verify: Optional[bool] = True,
|
||||
) -> None:
|
||||
"""Extract the data from
|
||||
a remote source. By default it provides code to get the file from a source url,
|
||||
unzips it and stores it on an extract_path."""
|
||||
"""Extract the data from a remote source. By default it provides code
|
||||
to get the file from a source url, unzips it and stores it on an
|
||||
extract_path."""
|
||||
|
||||
# this can be accessed via super().extract()
|
||||
if source_url and extract_path:
|
||||
|
@ -70,3 +120,53 @@ class ExtractTransformLoad:
|
|||
"""Clears out any files stored in the TMP folder"""
|
||||
|
||||
remove_all_from_dir(self.TMP_PATH)
|
||||
|
||||
# TODO: Add test for this
|
||||
def _get_census_fips_codes(self) -> pd.DataFrame:
|
||||
"""Loads FIPS codes for each Census block group and tract"""
|
||||
|
||||
# check that the census data exists
|
||||
if not self.CENSUS_CSV.exists():
|
||||
logger.info("Census data not found, please run download_csv first")
|
||||
# load the census data
|
||||
df = pd.read_csv(
|
||||
self.CENSUS_CSV, dtype={self.GEOID_FIELD_NAME: "string"}
|
||||
)
|
||||
# extract Census tract FIPS code from Census block group
|
||||
df[self.GEOID_TRACT_FIELD_NAME] = df[self.GEOID_FIELD_NAME].str[0:11]
|
||||
return df[[self.GEOID_FIELD_NAME, self.GEOID_TRACT_FIELD_NAME]]
|
||||
|
||||
# TODO: Create tests
|
||||
def validate_output(self) -> None:
|
||||
"""Checks that the output of the ETL process adheres to the contract
|
||||
expected by the score module
|
||||
|
||||
Contract conditions:
|
||||
- Output is saved as usa.csv at the path specified by self.OUTPUT_PATH
|
||||
- The output csv has a column named GEOID10 which stores each of the
|
||||
Census block group FIPS codes in data/census/csv/usa.csv
|
||||
- The output csv has a column named GEOID10_TRACT which stores each of
|
||||
Census tract FIPS codes associated with each Census block group
|
||||
- The output csv has each of the columns expected by the score and the
|
||||
name and dtype of those columns match the format expected by score
|
||||
"""
|
||||
# read in output file
|
||||
# and check that GEOID cols are present
|
||||
assert self.OUTPUT_PATH.exists(), f"No file found at {self.OUTPUT_PATH}"
|
||||
df_output = pd.read_csv(
|
||||
self.OUTPUT_PATH,
|
||||
dtype={
|
||||
self.GEOID_FIELD_NAME: "string",
|
||||
self.GEOID_TRACT_FIELD_NAME: "string",
|
||||
},
|
||||
)
|
||||
|
||||
# check that the GEOID cols in the output match census data
|
||||
geoid_cols = [self.GEOID_FIELD_NAME, self.GEOID_TRACT_FIELD_NAME]
|
||||
for col in geoid_cols:
|
||||
assert col in self.FIPS_CODES.columns
|
||||
assert self.FIPS_CODES.equals(df_output[geoid_cols])
|
||||
|
||||
# check that the score columns are in the output
|
||||
for col in self.SCORE_COLS:
|
||||
assert col in df_output.columns, f"{col} is missing from output"
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue