An instance of Collection is a container for multiple documents, where a document is an associative array of key-value pairs. Examples are the job state point, or the job document.

The Collection class is used internally to manage and search data space indexes which are generated on-the-fly. But you can also use such a container explicitly for managing document data.

Creating collections

To create an empty collection, simply call the default constructor:

from signac import Collection

collection = Collection()

You can then add documents with the signac.Collection.insert_one() method.

By default, the collection is stored purely in memory. But you can use the signac.Collection container also to manage collections directly on disk. For this, simply open a file like this:

with"my_collection.txt") as collection:

A collection file by default is openend in append plus mode, that means it is opened for both reading and writing. The open() function accepts all standard file open modes, such as r for read-only, etc.

Large collections can also be stored in a compressed format using gzip for efficiency. To use a compressed collection, simply pass in a compression level from 1-9 as a compresslevel argument to the signac.Collection constructor:

from signac import Collection

collection = Collection(compresslevel=9)

Searching collections

To search a collection, use the signac.Collection.find() method. As an example, to search all documents where the value a is equal to 42, execute:

for doc in collection.find({"a": 42}):

The signac.Collection.find() method uses the framework-wide query API.

Command Line Interface

To manage and search a collection file directly from the command line, create a Python script with the following content:

from signac import Collection

with"my_collection.txt") as c:

Storing the code above in a file called and then executing it will allow you to search for all or specific documents within the collection, directly from the command line $ python

For more information on how to use the command line interface, execute: $ python --help.