The Federal Energy Regulatory Commission (FERC) has moved to collecting and distributing data using XBRL. XBRL is primarily designed for financial reporting, and has been adopted by regulators in the US and other countries. Much of the tooling in the XBRL ecosystem is targeted towards filers, and rendering individual filings in a human readable way, but there is very little targeted towards accessing and analyzing large collections of filings. This tool is designed to provide that functionality for FERC XBRL data. Specifically, it can extract data from a set of XBRL filings, and write that data to a SQLite database whose structure is generated from an XBRL Taxonomy. While each XBRL instance contains a reference to a taxonomy, this tool requires a path to a single taxonomy that will be used to interpret all instances being processed. This means even if instances were created from different versions of a Taxonomy, the provided taxonomy will be used when processing all of these instances, so the output database will have a consistent structure. For more information on the technical details of the XBRL extraction, see the docs.
We are currently using this tool to extract and publish the following FERC data:
To install using conda, run the following command, and activate the environment.
$ conda env create -f environment.ymlActivate:
$ conda activate ferc-xbrl-extractorThis tool can be used as a library, as it is in PUDL,
or there is a CLI provided for interacting with XBRL data. The only required options
for the CLI are a path to the filings to be extracted, and a path to the output
SQLite database. The path to the filings can point to a directory full of XBRL
Filings, a single XBRL filing, or a zipfile with XBRL filings. If
the path to the database points to an existing database, the --clobber option
can be used to drop all existing data before performing the extraction.
$ xbrl_extract {path_to_filings} {path_to_database}This repo contains a small selection of FERC Form 1 filings from 2021, along with
an archive of taxonomies in the examples directory. To test the tool on these
filings, use the command:
$ xbrl_extract examples/ferc1-2021-sample.zip --db-path ./ferc1-2021-sample.sqlite \
--taxonomy examples/ferc1-xbrl-taxonomies.zipThe tool expects the --taxonomy option to point to a zipfile containing archived
taxonomies produced by the pudl-archiver.
The extractor will parse all taxonomies in the archive, then use the taxonomy referenced
in each filing while parsing it.
Parsing XBRL filings can be a time consuming and CPU heavy task, so this tool
implements some basic multiprocessing to speed this up. It uses a
process pool
to do this. There are two options for configuring the process pool, --batch-size
and --workers. The batch size configures how many filings will be processed by
each child process at a time, and workers specifies how many child processes to
create in the pool. It may take some experimentation to get these options
optimally configured. The following command will use 5 worker processes to process
batches of 50 filings at a time.
$ xbrl_extract examples/ferc1-2021-sample.zip .--db-path /ferc1-2021-sample.sqlite \
--taxonomy examples/ferc1-xbrl-taxonomies.zip
--workers 5 \
--batch-size 50There are also several options included for extracting metadata from the taxonomy.
First is the --datapackage-path command to save a
frictionless datapackage
descriptor as JSON, which annotates the generated SQLite database. There is also the
--metadata-path option, which writes more extensive taxonomy metadata to a json
file, grouped by table name. See the ferc_xbrl_extractor.arelle_interface module
for more info on the extracted metadata. To create both of these files using the example
filings and taxonomy, run the following command.
$ xbrl_extract examples/ferc1-2021-sample.zip .--db-path /ferc1-2021-sample.sqlite \
--taxonomy examples/ferc1-xbrl-taxonomies.zip
--metadata-path metadata.json \
--datapackage-path datapackage.json