Relaxing interface to SQLite

pip install dumptruck==0.1.6



DumpTruck is a document-like interface to a SQLite database.

Quick start

Install, save data and retrieve it using default settings.


pip2 install dumptruck || pip install dumptruck


Open the database connection by initializing the a DumpTruck object

from dumptruck import DumpTruck

dt = DumpTruck()


The simplest insert call looks like this.


This saves a new row with "Thomas" in the "firstname" column and "Levine" in the "lastname" column. It uses the table "dumptruck" inside the database "dumptruck.db". It creates or alters the table if it needs to.

If you insert one row, DumpTruck.insert returns the rowid of the row.

dt.insert({"foo", "bar"}, "new-table") == 1

If you insert many rows, DumpTruck.insert returns a list of the rowids of the new rows.

dt.insert([{"foo", "one"}, {"foo", "two"}], "new-table") == [2, 3]

If there are UNIQUE constraints on the table (perhaps from create_index) then insert will fail if these constraints are violated. You can use upsert (with the same syntax) to replace the existing row instead.


Once the database contains data, you can retrieve them.

data = dt.dump()

The data come out as a list of ordered dictionaries, with one dictionary per row.

Slow start


You can specify a few of keyword arguments when you initialize the DumpTruck object. For example, if you want the database file to be bucket-wheel-excavators.db, you can use this.

dt = DumpTruck(dbname="bucket-wheel-excavators.db")

It actually takes up to four keyword arguments.

DumpTruck(dbname='dumptruck.db', auto_commit = True, vars_table = "_dumptruckvars", adapt_and_convert = True)
  • dbname is the database file to save to; the default is dumptruck.db.
  • vars_table is the name of the table to use for DumpTruck.get_var and DumpTruck.save_var; default is _dumptruckvars. Set it to None to disable the get_var and save_var methods.
  • auto_commit is whether changes to the database should be automatically committed; if it is set to False, changes must be committed with the commit method or with the commit keywoard argument.
  • adapt_and_convert is whether types should be converted automatically; with this on dates get inserted as dates, lists as lists, &c.


As discussed earlier, the simplest insert call looks like this.

dt.insert({"firstname": "Thomas", "lastname": "Levine"})

Different tables

By default, that saves to the table dumptruck. You can specify a different table; this saves to the table diesel-engineers.

dt.insert({"firstname": "Thomas", "lastname": "Levine"}, "diesel-engineers")

Multiple rows

You can also pass a list of dictionaries.

    {"firstname": "Thomas", "lastname": "Levine"},
    {"firstname": "Julian", "lastname": "Assange"}

Complex objects

You can even pass nested structures; dictionaries, sets and lists will automatically be dumped to JSON.

    {"title":"The Elements of Typographic Style","authors":["Robert Bringhurst"]},
    {"title":"How to Read a Book","authors":["Mortimer Adler","Charles Van Doren"]}

Your data will be stored as JSON. When you query it, it will come back as the original Python objects.

And if you have some crazy object that can't be JSONified, you can use the dead-simple pickle interface.

# This fails
data = {"weirdthing": {range(100): None}

# This works
from dumptruck import Pickle
data = Pickle({"weirdthing": {range(100): None})

It automatically pickles and unpickles your complex object for you.


Column names and table names automatically get quoted if you pass them without quotes, so you can use bizarre table and column names, like no^[hs!'e]?'sf_"&'

Null values

None dictionary values are always equivalent to non-existence of the key. That is, these insert commands are equivalent.

dt = DumpTruck()
dt.insert({ u'foo': 8, u'bar': None})
dt.insert({ u'foo': 8})

Passing an empty dictionary creates a new row with all NULL values.

# These all create a row with all NULL values.
dt.insert({u'potato': None})

More precisely, they set the values to the default values via this SQL.


Passing an empty list to insert inserts zero rows (rather than one); this command does nothing.


You can pass zero rows or empty rows to DumpTruck.insert, but you'll get an error if you try passing them to DumpTruck.create_table.


You can use normal SQL to retrieve data from the database.

data = dt.execute('SELECT * FROM `diesel-engineers`')

The data come back as a list of dictionaries, one dictionary per row. They are coerced to different python types depending on their database types.

Individual values

It's often useful to be able to quickly and easily save one metadata value. For example, you can record which page the last run of a script managed to get up to.

dt.save_var('last_page', 27)
27 == dt.get_var('last_page')

It's stored in a table that you can specify when initializing DumpTruck. If you don't specify one, it's stored in _dumptruckvars.

If you want to save anything other than an int, float or string type, use json or pickle.


DumpTruck provides specialized wrapper around some common commands.

DumpTruck.tables returns a set of all of the tables in the database.


DumpTruck.drop drops a table.


DumpTruck.dump returns the entire particular table as a list of dictionaries.


It's equivalent to running this:

dt.execute('SELECT * from `coal`;')

Creating empty tables

When working with relational databases, one typically defines a schema before populating the database. You can use the DumpTruck.insert method like this by calling it with create_only = True.

For example, if the table tools does not exist, the following call will create the table tools with the columns toolName and weight, with the types TEXT and INTEGER, respectively, but will not insert the dictionary values ("jackhammer" and 58) into the table.

dt.create_table({"toolName":"jackhammer", "weight": 58}, "tools")

If you are concerned about the order of the tables, pass an OrderedDict.

dt.create_table(OrderedDict([("toolName", "jackhammer"), ("weight", 58)]), "tools")

The columns will be created in the specified order.



DumpTruck contains a special method for creating indices. To create an index, first create an empty table. (See "Creating empty tables" above.) Then, use the DumpTruck.create_index method.

dt.create_index(['toolName'], 'tools')

This will create a non-unique index on the column tool. To create a unique index, use the keyword argument unique = True.

dt.create_index(['toolName'], 'tools', unique = True)

You can also specify multi-column indices.

dt.create_index(['toolName', 'weight'], 'tools')

DumpTruck names these indices according to the names of the relevant table and columns. The index created in the previous example might be named dt__tools_toolName_weight.

Other index manipulation

DumpTruck does not implement special methods for viewing or removing indices, but here are the relevant SQLite SQL commands.

The following command lists indices on the tools table.

dt.execute('PRAGMA index_list(tools)')

The following command gives more information about the index named dt__tools_toolName_weight.

dt.execute('PRAGMA index_info(dt__tools_toolName_weight)')

And this one deletes the index.

dt.execute('DROP INDEX dt__tools_toolName_weight')

For more information on indices and, particularly, the PRAGMA commands, check the SQLite documentation.

Delaying commits

By default, the insert, get_var, drop and execute methods automatically commit changes. You can stop one of them from committing by passing commit=False to the method. Commit manually with the commit method. For example:

dt = DumpTruck()
dt.insert({"name":"Bagger 293","manufacturer":"TAKRAF","height":95}, commit=False)
dt.save_var('page_number', 42, commit=False)