base — Simple Document Interface


This module provides a simple document interface that stores its data in the MongoDB database, while saving you from dealing with the details of MongoDB operations. It provides a base.Document base class, which you subclass to instantiate to represent a specific document collection, e.g. Paper. It provides the four standard CRUD operations:

p = Paper(docData=dict(title='My Paper', authors=[person._id],
                       summary='Lorem ipsem...')) # CREATE doc

p = Paper(docID) # READ doc
print p.title, p.summary # data bound as attributes
for person in p.authors: # get linked objects
    print, len(person.subscriptions)

p.update(dict(year=2009, pages='45-60')) # UPDATE doc

p.delete() # DELETE from collection in database

What It Isn’t

Note: this is not an Object-Relational Mapper. ORM objects have to worry about data synchronization, in both directions. That is, if you change an attribute on the object, that must be saved to the database so anyone else who may be viewing this document will see the new value. Conversely, this object needs to know if any change occurred in the database (and update its attribute values accordingly). One way of saying this is that the ORM is itself responsible for ensuring synchronization over (any number of) object instances (across many threads? many processes? many machines connected to the same database?). We did not want this complexity and potential inefficiency.

What It Is

Instead, we provide the convenience of an object interface to the data, with the simplicity of the “stateless” CRUD model.

  • the database is solely responsible for synchronization and atomicity (as it is designed to be);
  • you only interact with the database through the four CRUD methods, so to change a document you must call its update() method;
  • you use an object to render a page view (or whatever you’re doing), and then throw it away. The spnet server is stateless; state is stored in the database, not the spnet server.

Embedded Documents

NoSQL documents can contain any structure of data, i.e. sub-documents “embedded” in a top-level document. This module provides convenient ways of declaring that a field (sub-document) should itself be represented as a Document object (with the same CRUD interface). This allows us to create a modular (object) interface to all these documents, enabling our code to work with them in the same way regardless of whether they are top-level documents or embedded subdocuments. The module provides both base.EmbeddedDocument and base.ArrayDocument classes.

Linked Documents

Document fields that reference another document (“foreign key”) are automatically transformed into Document objects for you. To make this efficient, the linked document data is actually retrieved only when you explicitly request that attribute of the linking Document. This enables Documents to have a rich interface of attribute links to all other relevant documents without paying any “database transactions” price. I.e. unless you actually view a specific attribute, it incurs no database query.

Document classes


The base.Document class is a base class for defining a MongoDB collection. You use this by declaring a subclass, with (at minimum) a coll attribute set to the pymongo.Collection you want it to use as its MongoDB datastore. Typically you also “decorate” your subclass with base.LinkDescriptor() attributes to link it to other document collections, custom methods and other useful data. For a simple example, see core.GplusSubscriptions in, or core.Paper for a more complex example.

class base.Document(fetchID=None, docData=None, insertNew=True)

Base class for defining a specific document collection.

fetchID, if not None, is the unique ID for a specific document you want to retrieve.

docData, if not None, is a dictionary of data representing the document.

insertNew=True will INSERT docData as a new record in the MongoDB collection associated with your specific Document subclass. If docData has an _id key, it will be used as the MongoDB ID of the inserted document; otherwise a unique ID will be auto-generated for you.

insertNew=False will simply initialize an object representing docData, without storing anything to MongoDB.


All MongoDB documents have a unique identifier _id field.


Update the MongoDB document record atomically, setting just the specific field values given by updateDict.


Delete this record from the MongoDB collection.

classmethod Document.find(queryDict={}, fields=None, idOnly=True, sortKeys=None, limit=None, **kwargs)

A convenience wrapper for the pymongo.Collection.find() method.

queryDict, if not empty, specifies a MongoDB query to perform on your collection.

fields, if not None, specifies the fields to be returned from matching records, in the usual MongoDB way.

idOnly=True makes it simply iterate over the database ID values of matching records (overrides the fields argument).

sortKeys and limit allow you to specify both the order and maximum number of results to be returned. Uses the efficient new MongoDB aggregation framework.

kwargs are passed through to pymongo.Collection.find().

classmethod Document.find_obj(queryDict={}, **kwargs)

Same as find(), but returns an object represenation of each matching record.


Gets a desired value via precedence rules defined by your subclass’ _get_value_attrs attribute, which must be a list of subdocument field names. It finds the first of these subdocuments that actually exists in this document, and calls its method named get_STEM() (where STEM is the value of the stem argument, i.e. in the default case this would be get_spnet_url()). If none of the subdocuments is found, it calls the same method name on itself.


Base class for defining subclasses that represent subdocuments embedded as a specified field in “parent” documents in a specified MongoDB collection. Your subclass must define the following:

  • _dbfield attribute must be a string of the form 'field.subfield', where field is the name of the field in the top-level document that will store the embedded document, and subfield is the name of the field in the embedded document that will store its ID value, which must be unique across all documents in this collection.

Your subclass may define the following optional methods:

  • _query_external(self, fetchID): must return docData for the embedded document, based on querying some external resource using the embedded document’s ID (fetchID).

  • _insert_parent(self, docData): must create a parent

    document (i.e. make a docData dictionary representing it (based on the docData from your embedded document), and save it in the database), and return a document object representing it.

class base.EmbeddedDocument(fetchID=None, docData=None, parent=None, insertNew=True)

Same arguments as for base.Document except as follows:

parent, if not None, must be either the ID of the parent document (containing this embedded document), or an object representing that parent document.

insertNew=’findOrInsert’ will make it first query the database to see if the specified document already exists (in which case it returns an object representing it). If not, it stores docData to the database, as a field of a specified parent document. If parent is not None, the new document will be embedded in the record specified by parent. If parent is None, then your subclass must implement an _insert_parent() method that will create a parent document (i.e. create a docData dictionary representing it, and save it in the database), and return a document object representing it.

Note that findOrInsert first queries the database using either fetchID, or if that is None, by extracting the fetchID value from docData[subfield], where subfield is obtained from self._dbfield (see above).


Update the MongoDB document record atomically, setting just the specific field values given by updateDict.


Delete this subdocument from its parent document.


Base class for defining subclasses that represent subdocuments stored as an array in a specified field of a “parent” document in a specified MongoDB collection. Your subclass must define at least the following:

  • _dbfield attribute must be a string of the form 'field.subfield', where field is the name of the field in the top-level document that will store the array, and subfield is the name of the field in each array document that will store its ID value.
class base.ArrayDocument(fetchID=None, docData=None, parent=None, insertNew=True)

Same arguments as for base.EmbeddedDocument except as follows:

fetchID, if not None, must be a tuple (parentID, subID), where parentID gives the ID of the parent (top-level) document containing this ArrayDocument, and subID gives gives the ID stored in the subdocument’s subfield field (see _dbfield attr explanation above).

insertNew=’findOrInsert’ does not support use of _insert_parent() or _query_external() methods. Instead, you must provide both parent and docData arguments.

classmethod ArrayDocument.find_obj_in_parent(parent, subID)

Returns the desired ArrayDocument whose fetchID is (parent._id, subID), from the already retrieved parent object, without performing any database query.

parent must be base.Document instance representing document containing this subdocument.


Same as base.ArrayDocument, but for the case where each subdocument has an identifier field that is unique, i.e. no other subdocument in the same named array in any document in the same collection will have the same value of this identifier field. This means that we can just use this subID as its fetchID (instead of a (parentID, subID) tuple as in the regular base.ArrayDocument case).

“Foreign Key” Convenience Classes

This module also provides several classes that provide convenient ways of linking an object to other objects.

class base.LinkDescriptor(attr, fetcher, noData=False, missingData=False, **kwargs)

Returns an attribute descriptor that a subclass can use to define an object attribute that will only be retrieved when someone actually tries to get its value (i.e. getattr).

attr is the name of the attribute.

fetcher must be a callable that will actually retrieve the desired object(s).

if noData is True, getattr(doc, attr) will simply return fetcher(doc, **kwargs). Otherwise, it will return fetcher(doc, data, **kwargs), where data is the raw data stored in the database as fieldname attr.

if missingData is False, an exception will be raised if no raw data for attr is stored in the database record. Otherwise in that case it will simply return the value specified by missingData.

You should not write directly to doc.attr (i.e. setattr); its data should come only from the database. If you want to change the raw data stored as attr in the database, use doc.update({attr:newValue}) in the usual CRUD way.

The module also provides a number of convenient classes to use as fetchers:

class base.FetchObj(klass, **kwargs)

For retrieving the document object specified by a “foreign key” giving its ID. When a base.FetchObj instance is used as a fetcher function, it interprets data as a fetchID, i.e. it returns klass(fetchID=data, **kwargs).

class base.FetchList(klass, **kwargs)

For retrieving a list of document objects specified by a list of foreign keys. I.e. returns [klass(fetchID, **kwargs) for fetchID in data].

class base.FetchQuery(klass, queryFunc, **kwargs)

For retrieving the results of a database query on the mongoDB collection klass.coll. queryFunc does not perform the query, it simply formulates the query to be performed (as always in mongoDB, the query is a dictionary). When a base.FetchQuery instance is used as a fetcher function, requesting getattr(doc, attr) will call queryFunc(doc, **kwargs), and will finally return list(klass.find_obj(query)) where query was the value returned by queryFunc.

You must always use the LinkDescriptor(noData=True) option when using this fetcher.

class base.FetchParent(klass, **kwargs)

Retrieves object representing parent document containing this subdocument.

EmbeddedDocument wrapping

This class converts the raw database data for a field representing a subdocument, into its associated base.EmbeddedDocument object.

class base.SaveAttr(klass, arg='parent', postprocess=None, **kwargs)

Returns klass(docData=data, **kwargs) where klass is the desired subdocument class to apply, and data is the raw database data for this subdocument.

ArrayDocument wrapping

This class converts the raw database data for a field representing an array, into a list of base.ArrayDocument objects.

class base.SaveAttrList(klass, arg='parent', postprocess=None, **kwargs)

Returns [klass(docData=d, **kwargs) for d in data] where klass is the desired subdocument class to apply, and data is the raw database data for this array.