Human-Centric-Machine-Learning/stackexchange-dump-to-postgres

README

StackOverflow data to postgres

This is a quick script to move the Stackoverflow data from the StackExchange data dump (Sept '14) to a Postgres SQL database.

Schema hints are taken from a post on Meta.StackExchange and from StackExchange Data Explorer.

Quickstart

Install requirements, create a new database (e.g. beerSO below), and use load_into_pg.py script:

$ pip install -r requirements.txt
...
Successfully installed argparse-1.2.1 libarchive-c-2.9 lxml-4.5.2 psycopg2-binary-2.8.4 six-1.10.0
$ createdb beerSO
$ python load_into_pg.py -s beer -d beerSO

This will download compressed files from archive.org and load all the tables at once.

Advanced Usage

You can use a custom database name as well. Make sure to explicitly give it while executing the script later.

Each table data is archived in an XML file. Available tables varies accross history. load_into_pg.py knows how to handle the following tables:

Badges.
Posts.
Tags (not present in earliest dumps).
Users.
Votes.
PostLinks.
PostHistory.
Comments.

You can download manually the files to the folder from where the program is executed: Badges.xml, Votes.xml, Posts.xml, Users.xml, Tags.xml. In some old dumps, the cases in the filenames are different.

Then load each file with e.g. python load_into_pg.py -t Badges.

After all the initial tables have been created:

$ psql beerSO < ./sql/final_post.sql

For some additional indexes and tables, you can also execute the the following;

$ psql beerSO < ./sql/optional_post.sql

If you give a schema name using the -n switch, all the tables will be moved to the given schema. This schema will be created in the script.

The paths are not changed in the final scripts sql/final_post.sql and sql/optional_post.sql. To run them, first set the search_path to your schema name: SET search_path TO <myschema>;

Caveats and TODOs

It prepares some indexes and views which may not be necessary for your analysis.
The Body field in Posts table is NOT populated by default. You have to use --with-post-body argument to include it.
The EmailHash field in Users table is NOT populated.

Sept 2011 data dump

The tags.xml is missing from the data dump. Hence, the PostTag and UserTagQA tables will be empty after final_post.sql.
The ViewCount in Posts is sometimes equal to an empty value. It is replaced by NULL in those cases.

Acknowledgement

@madtibo made significant contributions by adding jsonb and Foreign Key support.
@bersace brought the dependencies and the README.md instructions into 2020s.
@rdrg109 simplified handling of non-public schemas and fixed bugs associated with re-importing tables.