Skip to content
This repository has been archived by the owner. It is now read-only.


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


PyPI status CircleCI status Maintainability Test Coverage Gemnasium

Generate a random sample of rows from a relational database that preserves referential integrity - so long as constraints are defined, all parent rows will exist for child rows.

Good for creating test/development databases from production. It's slow, but how often do you need to generate a test/development database?


rdbms-subsetter <source SQLAlchemy connection string> <destination connection string> <fraction of rows to use>


rdbms-subsetter postgresql://:@/bigdb postgresql://:@/littledb 0.05

Valid SQLAlchemy connection strings are described here.

rdbms-subsetter promises that each child row will have whatever parent rows are required by its foreign keys. It will also try to include most child rows belonging to each parent row (up to the supplied --children parameter, default 3 each), but it can't make any promises. (Demanding all children can lead to infinite propagation in thoroughly interlinked databases, as every child record demands new parent records, which demand new child records, which demand new parent records... so increase --children with caution.)

When row numbers in your tables vary wildly (tens to billions, for example), consider using the -l flag, which sets row number targets by a logarithmic formula. When -l is set, if f is the fraction specified, and the original table has n rows, then each new table's row target will be:

math.pow(10, math.log10(n)*f)

A fraction of 0.5 seems to produce good results, converting 10 rows to 3, 1,000,000 to 1,000, and 1,000,000,000 to 31,622.

Rows are selected randomly, but for tables with a single primary key column, you can force rdbms-subsetter to include specific rows (and their dependencies) with force=<tablename>:<primary key value>. The children, grandchildren, etc. of these rows are exempted from the --children limit.

rdbms-subsetter only performs the INSERTS; it's your responsibility to set up the target database first, with its foreign key constraints. The easiest way to do this is with your RDBMS's dump utility. For example, for PostgreSQL,

pg_dump --schema-only -f schemadump.sql bigdb
createdb littledb
psql -f schemadump.sql littledb

Rows are taken from the schema visible by default to your database connection. You can also include rows from non-default schemas with the --schema=<name> parameter (which can be used multiple times). Currently the target database must contain the corresponding tables in its own schema of the same name (moving between schemas of different names is not yet supported).

You can restrict the tables included in the sample via the --table (-t) and --exclude-table (-T) parameters (which can be used multiple times). These parameters take a table name or pattern with wildcards (*), and supports both qualified names (i.e. schema.table) and simple names. When both -t and -T are given, the behavior is to include just the tables that match at least one -t switch, but no -T switches.

When target db write buffering is enabled (which it is by default), subset extraction for complex schemas that have tangled foreign key relationships may fail. To disable write buffering, set the buffer parameter to 0:

rdbms-subsetter  postgresql://:@/bigdb postgresql://:@/littledb 0.05 -b 0

Configuration file

If you need to honor relationships that aren't actually defined as foreign-key constraints in the database - for example, if you are using MySQL MyISAM and can't define constraints - you can specify a configuration file with --config. The config file should specify constraints in JSON. For example,

"constraints": {
"(child_table_name)": [
"referred_schema": null, "referred_table": "(name of parent table)", "referred_columns": ["(constraint col 1 in parent)", "(constraint col 2 in parent)",], "constrained_columns": ["(constrained col 1 in child)", "(constrained col 2 in child)",],



}, "tables": [ "SCHEMA1.TABLE1", "SCHEMA2.TABLE2" ], "schemas": [ "SCHEMA1", "SCHEMA2" ]


Optionally, you can qualify child_table_name, i.e. schema_name.child_table_name. Cross-schema constraints are also supported.

rdbms-subsetter treats these constraints like real foreign keys and fetches parent and child rows as described above.

tables and schemas are optional.

tables are merged with the --table elements passed on commandline.

schemas are merged with the --schema elements passed on commandline.

Signal handlers

If you provide a python module with appropriate signal handling functions, and specify that module when calling the script like --import=my.signals.signal_handlers, then any signal handlers that you have registered in your module will be called when the corresponding signals are sent during the DB subsetting process.

At the moment, the only signal is subsetter.SIGNAL_ROW_ADDED.

An example signal handling module:

from blinker import signal
import subsetter

row_added_signal = signal(subsetter.SIGNAL_ROW_ADDED)
def row_added(source_db, **kwargs):
   print("row_added called with source db: {}, and kwargs: {}".format(source_db, kwargs))


This signal will be sent when a new row has been selected for adding to the target database. The associated signal handler should have the following signature:

def row_added(source_db, **kwargs):

source_db is a subsetter.Db instance.

kwargs contains:

  • target_db: a subsetter.Db instance.
  • source_row: an sqlalchemy.engine.RowProxy with the values from the row that will be inserted.
  • target_table: an sqlalchemy.Table.
  • prioritized: a bool representing whether of not all child, grandchild, etc. rows should be included.


pip install rdbms-subsetter

Then the DB-API2 module for your RDBMS; for example, for PostgreSQL,

pip install psycopg2


Will consume memory roughly equal to the size of the extracted database. (Not the size of the source database!)


See also


Generates a subset of a relational database that respects foreign key constraints



Code of conduct





No releases published


No packages published