Query the Data Delivery Network
Query the DDNThe easiest way to query any data on Splitgraph is via the "Data Delivery Network" (DDN). The DDN is a single endpoint that speaks the PostgreSQL wire protocol. Any Splitgraph user can connect to it at data.splitgraph.com:5432 and query any version of over 40,000 datasets that are hosted or proxied by Splitgraph.
For example, you can query the cardiac_surgery_by_surgeon_beginning_2008 table in this repository, by referencing it like:
"health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb:latest"."cardiac_surgery_by_surgeon_beginning_2008"or in a full query, like:
SELECT
":id", -- Socrata column ID
"year_of_hospital_discharge", -- Year or range of years included in analysis
"facility_id", -- New York State (NYS) Facility Identification Number. Facility ID ‘0000’ represents the sum (or average) of all the reported data in NYS
"number_of_deaths", -- Number of cases resulting in death during the hospitalization or after discharge but within 30-days
"observed_mortality_rate", -- The observed number of deaths divided by the total number of cases
"nys_physician_license_number", -- New York State (NYS) License Number of the physician that performed the procedure
"physician_name", -- Name of the physician that performed the procedure. Physician information is presented for each physician who a) performed 200 or more procedures during the three year analysis and/or b) performed at least one PCI in each of the three years. The results for physicians not meeting the above criteria are grouped together and reported as “All others” in the hospital in which the procedures are performed
"expected_mortality_rate", -- The sum of the predicted probabilities of death for all patients divided by the total number of patients
"region", -- Region of hospital performing the procedure
"risk_adjusted_mortality_rate", -- The best estimate, based on the statistical model, of what the provider’s mortality rate would have been if the provider had a mix of patients similar to the statewide mix. It is obtained by first dividing the observed mortality rate by the expected mortality rate, and then multiplying that quotient by the statewide mortality rate
"number_of_cases", -- Number of procedures performed
"procedure", -- Type of procedure performed: All PCI, Non-Emergency PCI, Emergency PCI, CABG, or Valve Surgery. PCI stands for Percutaneous Coronary Intervention, sometimes also called coronary angioplasty or coronary stenting. CABG stands for Coronary Artery Bypass Graft surgery
"lower_limit_of_confidence_interval", -- Lower 95% confidence limit for risk adjusted mortality rate
"hospital_name", -- Hospital Name
"detailed_region", -- Region of hospital performing the procedure; New York City hospitals are classified by county
"national_provider_id", -- National Provider Identifier of the physician that performed the procedure.
"upper_limit_of_confidence_interval", -- Upper 95% confidence limit for risk adjusted mortality rate
"comparison_results" -- A statistical test was performed to determine if each hospital performed significantly better or worse than average for the given year. Result may be: Significantly higher than statewide rate, Significantly lower than statewide rate, Not different than the statewide rate
FROM
"health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb:latest"."cardiac_surgery_by_surgeon_beginning_2008"
LIMIT 100;Connecting to the DDN is easy. All you need is an existing SQL client that can connect to Postgres. As long as you have a SQL client ready, you'll be able to query health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb with SQL in under 60 seconds.
Query Your Local Engine
bash -c "$(curl -sL https://github.com/splitgraph/splitgraph/releases/latest/download/install.sh)"Read the installation docs.
Splitgraph Cloud is built around Splitgraph Core (GitHub), which includes a local Splitgraph Engine packaged as a Docker image. Splitgraph Cloud is basically a scaled-up version of that local Engine. When you query the Data Delivery Network or the REST API, we mount the relevant datasets in an Engine on our servers and execute your query on it.
It's possible to run this engine locally. You'll need a Mac, Windows or Linux system to install sgr, and a Docker installation to run the engine. You don't need to know how to actually use Docker; sgrcan manage the image, container and volume for you.
There are a few ways to ingest data into the local engine.
For external repositories, the Splitgraph Engine can "mount" upstream data sources by using sgr mount. This feature is built around Postgres Foreign Data Wrappers (FDW). You can write custom "mount handlers" for any upstream data source. For an example, we blogged about making a custom mount handler for HackerNews stories.
For hosted datasets (like this repository), where the author has pushed Splitgraph Images to the repository, you can "clone" and/or "checkout" the data using sgr cloneand sgr checkout.
Cloning Data
Because health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb:latest is a Splitgraph Image, you can clone the data from Spltgraph Cloud to your local engine, where you can query it like any other Postgres database, using any of your existing tools.
First, install Splitgraph if you haven't already.
Clone the metadata with sgr clone
This will be quick, and does not download the actual data.
sgr clone health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xbCheckout the data
Once you've cloned the data, you need to "checkout" the tag that you want. For example, to checkout the latest tag:
sgr checkout health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb:latestThis will download all the objects for the latest tag of health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb and load them into the Splitgraph Engine. Depending on your connection speed and the size of the data, you will need to wait for the checkout to complete. Once it's complete, you will be able to query the data like you would any other Postgres database.
Alternatively, use "layered checkout" to avoid downloading all the data
The data in health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb:latest is 0 bytes. If this is too big to download all at once, or perhaps you only need to query a subset of it, you can use a layered checkout.:
sgr checkout --layered health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb:latestThis will not download all the data, but it will create a schema comprised of foreign tables, that you can query as you would any other data. Splitgraph will lazily download the required objects as you query the data. In some cases, this might be faster or more efficient than a regular checkout.
Read the layered querying documentation to learn about when and why you might want to use layered queries.
Query the data with your existing tools
Once you've loaded the data into your local Splitgraph Engine, you can query it with any of your existing tools. As far as they're concerned, health-data-ny-gov/cardiac-surgery-by-surgeon-beginning-2008-dk4z-k3xb is just another Postgres schema.