Lightweight ETL for extracting CSV data from NYC DOHMH.
Clone this repository and then:
bundle install
and then:
ruby etl.rb
I tried to keep this ETL as flexible as possible, allowing to just edit Loader#setup_db with the columns needed. Loader.rb dynamically creates the insert purely based on the data passed into it. With a quick addition to the extractor to filter only certain keys one would be able to only store requested columns. The following code would need to be added to the extractor:
def transform_data(row)
filter_row(row) // new code
sanitize_input(row)
add_full_address(row)
return row
end
@columns = [
"CAMIS", "DBA", "BORO", "BUILDING", "STREET", "ZIPCODE", "PHONE",
"CUISINE DESCRIPTION", "INSPECTION DATE", "SCORE", "GRADE"
]
def filter_row(row)
row.select! { |k, v| @columns.include?(k) }
end
When it came to the choice of database and its design I went with sqlite for fairly simple reasons, it is fast and it is convenient for transferring to another project. I ensure that I do not use too much memory by splitting the database loading into chunks. I would have liked to have found a composite key that I could ensure was unique to a record but found that my assumptions that there could only be one violation_code per camis on a specific inspection_date was inaccurate. The downside to not having a composite key that still allows all records to be entered is that if this ETL were to be run multiple times on the same database there would be duplicated data.
I would have also liked to add some cli progress bars as these processes do tend to take some time; however, I decided my time was better spent elsewhere.
Below is the schema for database.db which contains one table:
dohmh_inspections
| column name | data type | details |
|---|---|---|
id |
int | PRIMARY KEY, AUTOINCREMENT |
camis |
int(10) | NOT NULL |
dba |
varchar(255) | DEFAULT NULL |
address |
varchar(255) | DEFAULT NULL |
boro |
varchar(255) | DEFAULT NULL |
building |
int(10) | NOT NULL |
street |
varchar(255) | DEFAULT NULL |
zipcode |
int(10) | NOT NULL |
phone |
varchar(10) | DEFAULT NULL |
action |
varchar(255) | DEFAULT NULL |
violation_code |
varchar(255) | DEFAULT NULL |
violation_description |
varchar(255) | DEFAULT NULL |
critical_flag |
varchar(255) | DEFAULT NULL |
cuisine_description |
varchar(255) | DEFAULT NULL |
inspection_date |
timestamp | NOT NULL |
score |
int(10) | DEFAULT NULL |
grade |
varchar(10) | DEFAULT NULL |
grade_date |
timestamp | DEFAULT NULL |
inspection_type |
varchar(255) | DEFAULT NULL |
record_date |
timestamp | DEFAULT NULL |
created_at |
timestamp | NOT NULL, DEFAULT CURRENT_TIMESTAMP |
updated_at |
timestamp | NOT NULL, DEFAULT CURRENT_TIMESTAMP |