A repository of journalist's lookup tables. Designed for programmatic access using tools such as agate-lookup.
Anyone may contribute a lookup table by sending a pull request to this repository.
Each folder is a key that can be used for a lookup. Within that folder are CSV files. The name of the CSV file is the name of the value that it maps to. The CSV itself will contain two columns, one with the key and another with the value. For example, usps/state.csv contains a CSV file that looks like this:
usps,state
AL,Alabama
AK,Alaska
AZ,Arizona
...
Sometimes the mapping from a key to value varies over time. For example, NAICS codes change every five years. In this case, a version specifier may be included in the filename. For example, naics/description.2007.csv is the 2007 version of the code mapping and naics/description/2012.csv is the 2012 version.
It may also be useful to be able to map two keys to a single value. For example, you might want to look up population by state and year. In those cases key folders can be nested and the CSV can contain more than one key column. For example, usps/year/population.csv contains a CSV that looks like this:
usps,year,state
AL,2015,4858979
AL,2014,4846411
AL,2013,4830533
...
Each CSV table must be accompanied by a YAML file. That file must have an identical filename, plus the .yml extension. For example, the table fips/state.csv must be accompanied by fips/state.csv.yml. This file should contain the following metadata:
data: A description of the data, including any notes necessary to use it correctly.
version: A description of the specific version of the data.
sources:
- A list of sources for the data, such as "United States Census Bureau", including URLs whenever possible
contributors:
- The name <and email of anyone who has contributed to this table>
columns:
key_column_name: Agate column type, such as "Text" or "Number"
value_column_name: Agate column type, such as "Text" or "Number"See naics/description.2007.csv.yaml for an example of a complete metadata file.
Anyone may submit a pull request to add a table to this repository, however, the following rules will guide inclusion of any data:
- The data must have journalistic value.
- The data must be from an authoritative source.
- The CSV must be in "standardized" CSV format. (Run through in2csv.)
- All keys must be unique. (No split/combine crosswalks.)
- All keys must be durable identifiers, not names.
- All filenames and keys must use
snake_case. - Periods must not be used in filenames or keys except as defined above.
- Four digit years must be used everywhere.
- Each CSV must be 250KB or less.
If people are going to rely on the tables in this dataset then there must be a log every error. If you find an error in any data, please send a pull request with a correction. That same pull request must also add an entry to ERRORS.md describing precisely the nature of the error.