DB I/O

A simple Python module for the following database operations: importing from CSV, querying to CSV, or querying to a table in a database. All database-specific knowledge is abstracted.

Installation

Use pip or setup utils:

pip install dbio

Additionally, you will need the Python DB API modules for the database(s) that you wish to use. Any databases that are currently supported have the modules listed in the extras_require dictionary in setup.py. For example, if you wish to install the modules needed for using MySQL and Vertica, you can augment the above command:

pip install dbio[MySQL, Vertica]

Note that some Python DB API modules may have OS-dependent pre-requisite packages to install. Consult the module's installation guide if setup fails.

Usage

To run an operation from the command line:

dbio dbio_args op_name op_args

For general help, use dbio -h. For operation specific help and a list of args, use dbio op_name -h. All commands have CSV formatting options that can be specified via the command-line. For the full list, view dbio -h.

To call an operation within Python:

import dbio
dbio.op_name(op_args, op_kwargs)

All operations require an SQLAlchemy URL for each database used in the operation.

For more detailed information on calling scripts within Python, check out the documentation.

Logging is supported via the Python logging module.

Tests can be run with

python setup.py test -a 'py.test args'

Operations

Replicate

dbio replicate query_db_url load_db_url query table

Runs query against the database pointed to by query_db_url, loading the results into table in the database pointed to by load_db_url. The rows in the loading database that existed before replication are removed.

Optional flags:

• -f: indicates that query is the name of a file.
• -a: runs in append mode. Rows that were in the loading database before replication are preserved.
• -z: analyzes table for query optimization after completing the load.
• -i: drops or disable indices while loading, recreating them afterwards.
• -nf: does not use mkfifo(). Use this if mkfifo() is not available.
• -s: expects an table named 'table_staging' to already exist.
• -rc: performs a check to ensure that the query rowcount matches the load table rowcount.
• -dt: Adds direct keyword to Vertica copy commands used in loading data. supported by your OS (e.g. Windows).

How it Works

Replicate creates a Unix first-in-first-out object (a.k.a. a "named pipe"). Using the subprocess module, Load and Query are then simultaneously executed, with the loading operation acting as the pipe reader, and the query object acting as the pipe writer. This allows query results to be streamed directly into the database's preferred method of import.

For a detailed explanation, see this blog post.

Load

dbio load db_url table filename

Loads the contents of a csv file at filename into table in the database pointed to by load_db_url. The rows in the loading database that existed before replication are removed.

Optional flags:

• -a: runs in append mode. Rows that were in the loading database before replication are preserved.
• -z: analyzes table after completing the load.
• -i: drops or disable indices while loading, recreating them afterwards.
• -s: expects a table named 'table_staging' to already exist.
• -rc: the number of rows to ensure are present in the table after loading.
• -dt: Adds direct keyword to Vertica copy commands used in loading data.

• csv flags:

  • -qc: character to enclose fields. If not included, fields are not enclosed.
  • -ns: string to replace NULL fields. Defaults to "NULL".
  • -d: field separation character. Defaults to ",".
  • -esc: escape character. Defaults to "".
  • -l: record terminator. Defaults to "n".
  • -e: character encoding. Defaults to "utf-8"

Query

dbio query db_url query filename

Runs query against the database pointed to by query_db_url, placing the results into a CSV file at filename.

Optional flags:

• -f: indicates that query is the name of a file.
• -b: specify batch_size, which determines the number of rows. to store in memory before writing to the file. Defaults to 1,000,000.

• csv flags:

  • -qc: character to enclose fields. If not included, fields are not enclosed.
  • -ns: string to replace NULL fields. Defaults to "NULL".
  • -d: field separation character. Defaults to ",".
  • -esc: escape character. Defaults to "".
  • -l: record terminator. Defaults to "n".
  • -e: character encoding. Defaults to "utf-8"

Databases

MySQL

Include 'MySQL' in the list of extras when installing.

Requires MySQL-python.

PostgreSQL

Include 'PostgreSQL' in the list of extras when installing.

Requires psycopg2.

SQLite

Included in the Python standard library. Note that the SQLite python library has no method designed for bulk-loading from CSV, so batch insert statements are used, which may cause bottlenecks that are not present for other databases.

Vertica:

Include 'Vertica' in the list of extras when installing.

Requires vertica-python and sqlalchemy-vertica-python

Alternatively, there is support for using pyodbc to drive the connection.

Additional Databases

To add support for a new database:

1. Create a class inside the dbio.databases subpackage that extends dbio.databases.base.Exportable and/or dbio.databases.base.Importable depending on desired functionality. The DB must have a valid SQLAlchemy Dialect. Existing supported databases are listed here, but SQLAlchemy also supports registering new dialects.
2. Add a corresponding import and mapping dictionary entry into databases.__init__.py

Examples

Query a Vertica database and stream the results into a MySQL table with a schema matching the results:

dbio replicate "vertica+vertica_python://user:pwd@host:port/database" "mysql://user:pwd@host:port/database" "SELECT * FROM vertica_table" mysql_table

Load foo.csv with "|" field delimiters into a PostgreSQL table:

dbio load "postgresql://user:pwd@host:port/database" foo_table foo.csv -d "|"

Query a SQLite table using a query file and write the results to a CSV with NULL represented by "NULL" and lines terminated with "\n".

dbio query "sqlite:///path/to/sqlite/db/file.db" foo_query.sql foo.csv -f -ns NULL -l "\n"

Changelog

• 0.5.0: Adding in ability to add Vertica DIRECT hint when loading data into Vertica
• 0.4.9: Deleting staging table prior to creating it to ensure that function does not fail for that reason.
• 0.4.8: Pinning requirement and "extras" requirements to specific versions where applicable
• 0.4.7: When a new staging table is created with the postgres db, grants are copied over as well
• 0.4.6: Making version number read from all dbio/__init__.py so it doesn't need to copied and pasted everywhere.
• 0.4.5: Explicit SQLAlchemy transaction creation.
• 0.4.4: Added SQL logic to Vertica and MySQL to raise errors when loading has any issues, including truncated data.
• 0.4.3: Minor logging additions.
• 0.4.2: Added a rowcount check option and randomized pipe names.
• 0.4.1: Support for existing staging tables.
• 0.4.0: Support for temporary index disabling.
• 0.3.4: Link to documentation in README.rst.
• 0.3.3: Added public documentation and minor fixes.
• 0.3.2: Fixed minor vertica.py bugs.
• 0.3.1: Fixed critical CLI bug.
• 0.3.0: Handle replication CSV formatting automatically.
• 0.2.0: Add ANALYZE support.
• 0.1.1: Initial public release.