Add ability to cache ETL data sources (#2169)

* Add a rough prototype allowing a developer to pre-download data sources for all ETLs

* Update code to be more production-ish

* Move fetch to Extract part of ETL
* Create a downloader to house all downloading operations
* Remove unnecessary "name" in data source

* Format source files with black

* Fix issues from pylint and get the tests working with the new folder structure

* Clean up files with black

* Fix unzip test

* Add caching notes to README

* Fix tests (linting and case sensitivity bug)

* Address PR comments and add API keys for census where missing

* Merging comparator changes from main into this branch for the sake of the PR

* Add note on using cache (-u) during pipeline

data/data-pipeline/data_pipeline/etl/sources/census_acs/etl.py

@@ -8,12 +8,11 @@ from data_pipeline.etl.base import ExtractTransformLoad
 from data_pipeline.etl.sources.census_acs.etl_imputations import (
     calculate_income_measures,
 )
-from data_pipeline.etl.sources.census_acs.etl_utils import (
-    retrieve_census_acs_data,
-)
 from data_pipeline.score import field_names
 from data_pipeline.utils import get_module_logger
 from data_pipeline.utils import unzip_file_from_url
+from data_pipeline.etl.datasource import DataSource
+from data_pipeline.etl.datasource import CensusDataSource
 
 logger = get_module_logger(__name__)
 
@@ -28,6 +27,9 @@ class CensusACSETL(ExtractTransformLoad):
     MINIMUM_POPULATION_REQUIRED_FOR_IMPUTATION = 1
 
     def __init__(self):
+
+        self.census_acs_source = self.get_sources_path() / "acs.csv"
+
         self.TOTAL_UNEMPLOYED_FIELD = "B23025_005E"
         self.TOTAL_IN_LABOR_FORCE = "B23025_003E"
         self.EMPLOYMENT_FIELDS = [
@@ -311,6 +313,34 @@ class CensusACSETL(ExtractTransformLoad):
 
         self.df: pd.DataFrame
 
+    def get_data_sources(self) -> [DataSource]:
+        # Define the variables to retrieve
+        variables = (
+            [
+                self.MEDIAN_INCOME_FIELD,
+                self.MEDIAN_HOUSE_VALUE_FIELD,
+            ]
+            + self.EMPLOYMENT_FIELDS
+            + self.LINGUISTIC_ISOLATION_FIELDS
+            + self.POVERTY_FIELDS
+            + self.EDUCATIONAL_FIELDS
+            + self.RE_FIELDS
+            + self.COLLEGE_ATTENDANCE_FIELDS
+            + self.AGE_INPUT_FIELDS
+        )
+
+        return [
+            CensusDataSource(
+                source=None,
+                destination=self.census_acs_source,
+                acs_year=self.ACS_YEAR,
+                variables=variables,
+                tract_output_field_name=self.GEOID_TRACT_FIELD_NAME,
+                data_path_for_fips_codes=self.DATA_PATH,
+                acs_type="acs5",
+            )
+        ]
+
     # pylint: disable=too-many-arguments
     def _merge_geojson(
         self,
@@ -339,27 +369,15 @@ class CensusACSETL(ExtractTransformLoad):
             )
         )
 
-    def extract(self) -> None:
-        # Define the variables to retrieve
-        variables = (
-            [
-                self.MEDIAN_INCOME_FIELD,
-                self.MEDIAN_HOUSE_VALUE_FIELD,
-            ]
-            + self.EMPLOYMENT_FIELDS
-            + self.LINGUISTIC_ISOLATION_FIELDS
-            + self.POVERTY_FIELDS
-            + self.EDUCATIONAL_FIELDS
-            + self.RE_FIELDS
-            + self.COLLEGE_ATTENDANCE_FIELDS
-            + self.AGE_INPUT_FIELDS
-        )
+    def extract(self, use_cached_data_sources: bool = False) -> None:
 
-        self.df = retrieve_census_acs_data(
-            acs_year=self.ACS_YEAR,
-            variables=variables,
-            tract_output_field_name=self.GEOID_TRACT_FIELD_NAME,
-            data_path_for_fips_codes=self.DATA_PATH,
+        super().extract(
+            use_cached_data_sources
+        )  # download and extract data sources
+
+        self.df = pd.read_csv(
+            self.census_acs_source,
+            dtype={self.GEOID_TRACT_FIELD_NAME: "string"},
         )
 
     def transform(self) -> None: