dbflows 0.2.4

Database incremental exports, transfers, imports, ETL, creation / management

A Python/CLI tool for:

1. Exporting database tables to compressed CSV files.
2. Transferring tables from from one database server to another.
3. Loading database data (from both files and Python)
4. Creating/Managing Postgresql/TimescaleDB tables, views, materialized views, functions, procedures, continuous aggregates, scheduled tasks.
5. Checking for mismatched attributes between SQLAlchemy tables/models and actual tables in a database.

Currently only Postgresql and Postgresql-based databases (e.g. TimescaleDB) are supported.

Install

pip install dbflows

If using the export functionality (export database tables to compressed CSV files), then you will additionally need to have the psql executable available. To install psql:

# enable PostgreSQL package repository
sudo sh -c 'echo "deb http://apt.postgresql.org/pub/repos/apt $(lsb_release -cs)-pgdg main" > /etc/apt/sources.list.d/pgdg.list'
wget -qO- https://www.postgresql.org/media/keys/ACCC4CF8.asc | sudo tee /etc/apt/trusted.gpg.d/pgdg.asc &>/dev/null
# replace `16` with the major version of your database
sudo apt update && sudo apt install -y postgresql-client-16

Export

Features:

• File splitting. Create separate export files based on a 'slice column' (an orderable column. e.g. datetime, integer, etc) and/or 'partition column' (a categorical column. e.g. name string).
• Incremental exports (export only data not yet exported). This works for both single file and multiple/split file output.

Examples

from dbflows import export
import sqlalchemy as sa
# the table to export data from
my_table = sa.Table(
    "my_table", 
    sa.MetaData(schema="my_schema"), 
    sa.Column("inserted", sa.DateTime),
    sa.Column("category", sa.String),
    sa.Column("value", sa.Float),
)
# one or more save locations (2 in this case)
save_locs = ["s3://my-bucket/my_table_exports", "/path/to/local_dir/my_table_exports"]
# database URL
url = "postgres://user:password@hostname:port/database-name"

Export entire table to a single file.

export(
    table=my_table,
    engine=url, # or sa.engine
    save_locs=save_locs
)

CLI equivalent:

db export table \
my_table.my_schema \
postgres://user:password@hostname:port/database-name` \
s3://my-bucket/my_table_exports \
/path/to/local_dir/my_table_exports

Export 500 MB CSVs, sorted and sliced on inserted datetime column.

export(
    table=my_table,
    engine=url, # or sa.engine
    save_locs=save_locs,
    slice_column=my_table.c.inserted,
    file_max_size="500 MB"
)

Create a CSV export for each unique category in table.

export(
    table=my_table,
    engine=url, # or sa.engine
    save_locs=save_locs,
    partition_column=my_table.c.category
)

CLI equivalent:

db export table \
my_table.my_schema \
postgres://user:password@hostname:port/database-name` \
# save to one or more locations (s3 paths or local)
s3://my-bucket/my_table_exports \ 
/path/to/local_dir/my_table_exports \ 
--partition-column category # or "-p category"

export 500 MB CSVs for each unique category, sorted and sliced on inserted datetime column.

export(
    table=my_table,
    engine=url, # or sa.engine
    save_locs=save_locs,
    slice_column=my_table.c.inserted,
    file_max_size="500 MB",
    partition_column=my_table.c.category,
)

Loading/Importing

Loading from Python objects

Create a PgLoader instance for your table and use the load method to load batches of rows.

Loading from CSV files

Use import_csvs to load CSV with parallel worker threads. This internally uses timescaledb-parallel-copy which can be installed with: go install github.com/timescale/timescaledb-parallel-copy/cmd/timescaledb-parallel-copy@latest