datafreeze

datafreeze creates static extracts of SQL databases for use in interactive web applications. SQL databases are a great way to manage relational data, but exposing them on the web to drive data apps can be cumbersome. Often, the capacities of a proper database are not actually required, a few static JSON files and a bit of JavaScript can have the same effect. Still, exporting JSON by hand (or with a custom script) can also become a messy process.

With datafreeze, exports are scripted in a Makefile-like description, making them simple to repeat and replicate.

Installation

The easiest way to install datafreeze is to retrieve it from the Python package index using pip:

pip install datafreeze

Basic Usage

Calling DataFreeze is simple, the application is called with a freeze file as its argument:

datafreeze Freezefile.yaml

Freeze files can be either written in JSON or in YAML. The database URI indicated in the Freezefile can also be overridden via the command line:

datafreeze --db sqlite:///foo.db Freezefile.yaml

Example Freezefile.yaml

A freeze file is composed of a set of scripted queries and specifications on how their output is to be handled. An example could look like this:

common:

  database: "postgresql://user:password@localhost/operational_database"
  prefix: my_project/dumps/
  format: json

exports:

  - query: "SELECT id, title, date FROM events"
    filename: "index.json"

  - query: "SELECT id, title, date, country FROM events"
    filename: "countries/{{country}}.csv"
    format: csv

  - query: "SELECT * FROM events"
    filename: "events/{{id}}.json"
    mode: item

  - query: "SELECT * FROM events"
    filename: "all.json"
    format: tabson

An identical JSON configuration can be found in this repository.

Options in detail

The freeze file has two main sections, common and exports. Both accept many of the same arguments, with exports specifying a list of exports while common defines some shared properties, such as the database connection string.

The following options are recognized:

• database is a database URI, including the database type, username and password, hostname and database name. Valid database types include sqlite, mysql and postgresql (requires psycopg2).
• prefix specifies a common root directory for all extracted files.
• format identifies the format to be generated, csv, json and tabson are supported. tabson is a condensed JSON representation in which rows are not represented by objects but by lists of values.
• query needs to be a valid SQL statement. All selected fields will become keys or columns in the output, so it may make sense to define proper aliases if any overlap is to be expected.
• mode specifies whether the query output is to be combined into a single file (list) or whether a file should be generated for each result row (item).
• filename is the output file name, appended to prefix. All occurences of {{field}} are expanded to a fields value to allow the generation of file names e.g. by primary key. In list mode, templating can be used to group records into several buckets, e.g. by country or category.
• wrap can be used to specify whether the output should be wrapped in a results hash in JSON output. This defaults to true for list-mode output and false for item-mode.

Contributors

dataset is written and maintained by Friedrich Lindenberg, Gregor Aisch and Stefan Wehrmeyer. We're standing on the shoulders of giants.