namlook / mongokit Goto Github PK

I tried setting up a default pylons project to use mongokit as directed here:
http://namlook.github.com/mongokit/pylons.html

However it gives me the error:

Traceback (most recent call last):
File "/usr/bin/paster", line 18, in
command.run()
File "/usr/lib/pymodules/python2.6/paste/script/command.py", line 84, in run
invoke(command, command_name, options, args[1:])
File "/usr/lib/pymodules/python2.6/paste/script/command.py", line 123, in invoke
exit_code = runner.run(args)
File "/usr/lib/pymodules/python2.6/paste/script/command.py", line 218, in run
result = self.command()
File "/usr/lib/pymodules/python2.6/paste/script/serve.py", line 276, in command
relative_to=base, global_conf=vars)
File "/usr/lib/pymodules/python2.6/paste/script/serve.py", line 313, in loadapp
*_kw)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 204, in loadapp
return loadobj(APP, uri, name=name, *_kw)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 224, in loadobj
global_conf=global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 248, in loadcontext
global_conf=global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 278, in _loadconfig
return loader.get_context(object_type, name, global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 409, in get_context
section)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 431, in _context_from_use
object_type, name=use, global_conf=global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 361, in get_context
global_conf=global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 248, in loadcontext
global_conf=global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 285, in _loadegg
return loader.get_context(object_type, name, global_conf)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 561, in get_context
object_type, name=name)
File "/usr/lib/pymodules/python2.6/paste/deploy/loadwsgi.py", line 587, in find_egg_entry_point
possible.append((entry.load(), protocol, entry.name))
File "/usr/lib/python2.6/dist-packages/pkg_resources.py", line 1954, in load
entry = import(self.module_name, globals(),globals(), ['name'])
File "/home/ciferkey/Projects/runnr-site/runnrsite/config/middleware.py", line 11, in
from runnrsite.config.environment import load_environment
File "/home/ciferkey/Projects/runnr-site/runnrsite/config/environment.py", line 8, in
import runnrsite.lib.app_globals as app_globals
File "/home/ciferkey/Projects/runnr-site/runnrsite/lib/app_globals.py", line 2, in
from ekeet.models import register_models
ImportError: No module named ekeet.models

I haven't been able to find any help anywhere else. Do you have any ideas as to why this is happening?

I am bringing data into Mongo from an external process using YAML. Because of that, I needed to loosen up some of the validation that MongoKit was doing for me and have MongoKit fill in some gaps in my data. So I added the following config variables:

autofill_incomplete_records = False
allow_nonrequired_to_be_missing = False
allow_unspecified_fields = False
auto_promote_str_and_int = False

The first one is the major enhancement. It tells MongoKit to accept a sparse record from Mongo and just fill in the missing data for me. This turned out to be really easy to do since namlook already had most of the work done to make this work. This is great for creating initial data, and for automatically adding to your structure without having to do an actual migrate.

In my case, if my collection is empty, I drop a bunch of documents in with only the 'url' filled out. When I read through the records, MongoKit will fix them up and add everything I left out.

One set of data I am storing is a dict of attributes on a corresponding file. Some files have 5 attributes while others have 30. Some are numeric, some dates, etc. So I want to specify the types for those I care about, but I don't even want to list many of the attributes in my structure. This is where allow_nonrequired_to_be_missing and allow_unspecified_fields come in. If allow_nonrequired_to_be_missing is True, validate will not complain about a missing field unless it is on the required fields list. If allow_unspecified_fields in True, validate won't complain about extra fields it doesn't know about.

Finally, since the YAML output often contains strings rather than unicode, I added auto_promote_str_and_int. I don't really care if a field should be unicode but it actually contains a str. I also don't care if it is supposed to be long but actually has an int. So set this to True and MongoKit will accept str for unicode and int for long.

I only had to change about a dozen lines of code in schema_document.py and a few lines in document.py. I'm happy to share but I don't know what the standard procedure is for submitting possible enhancements. All of the new code still passes all of the unit tests that 0.6 passed so nothing should change for current users unless they turn on some of my switches. (I'm not married to any of my config names, BTW.)

Is there an easy way for me to submit my changes for review?

Cheers,

Scott Maxwell
Code Cobblers, Inc.

Currently mongokit uses the structure dictionary to decide which class of document to cast to when autoreferencing. That can cause problems when you want to have a document autoreference another document, if that referenced document's class is not known ahead of time.
Below is a some code where verification fails because the Run document casts run.result into its base class when you save run.
Rather that using the structure property to cast an autoreferenced document to its proper class, why not include a _type field that is used to select which class to cast a document to. Along with that, a generic, mongokit 'document reference type' would need to be created to allow one document to reference another, without specifying what kind of document it is. When the document is loaded from the database, it would be automatically converted into the correct class based on the _type field.

http://gist.github.com/636106

see http://groups.google.com/group/mongokit/browse_thread/thread/a1d26733b555ae25

i have a structure like this:
structure = { 'meta': dict }

when i set indexes like this:

indexes = [{'field': 'meta.title', 'checked': False}]

i get the following exception:
ValueError: Error in indexes: can't find meta.title in structure

i thought 'checked': False would prevent this?

i can't put 'title' in the structure for 2 reason:

1. it's optionnal
2. it would conflict with the fact that meta contains a dict of arbitrary key/value

this would allow something like this :

con = Connection()

@con.register
class MyDoc(Document):
  structure = {....}

Note : This patch seems to be quite straightforward to implement. However, it might be a bit trickier to implement for Pylons users. Since the connection is initialized inside Globals, the models must not be imported before that.

Christopher Grebs wrote a patch to http://bytebucket.org/EnTeQuAk/mongokit-pymongo-1.6/changeset/5a3cb3fc8d1d

see https://gist.github.com/875662

We should be able to do this:

class MyDoc(Document):
    structure = {
        "tags": Set(unicode),
    }

>>> doc = db.col.MyDoc()
>>> doc['tags'] = [u'foo', u'bar', u'foo']
>>> doc.save()

>>> fetched_doc = db.col.get_from_id(doc['_id'])
>>> fetched_doc['tags']
set([u'foo', u'bar'])

the following should work without exception:

from mongokit import Connection
con = Connection()

@con.register
class Root(Document):
    __database__ = "test"

@con.register
class DocA(Root):
   __collection__ = "doca"
  structure = {'title:unicode}

Add the possibility to install tip by easy_install/pip using "mongokit==dev".

issue history : http://bitbucket.org/namlook/mongokit/issue/23/provide-nightly-build-install-for

See http://groups.google.com/group/mongokit/browse_thread/thread/f282f4dee0dfe1ce
Maybe we should call this descriptor "process" as it will process the value before saving it...

This will allow to accept either unicode and str types

So if you need to use another mongo tools or another ODM in another languages, your data won’t be pollued by MongoKit’s staff.

should be

So if you need to use other mongo tools or ODMs in other languages, your data won’t be polluted by MongoKit’s stuff.

cron:/# easy_install mongokit
Searching for mongokit
Reading http://pypi.python.org/simple/mongokit/
Reading http://bitbucket.org/namlook/mongokit/
Best match: mongokit 0.5.13
Downloading http://pypi.python.org/packages/source/m/mongokit/mongokit-0.5.13.tar.gz#md5=e8d65011e112571968c32d1a92a812ee
Processing mongokit-0.5.13.tar.gz
Running mongokit-0.5.13/setup.py -q bdist_egg --dist-dir /tmp/easy_install-xjR6j4/mongokit-0.5.13/egg-dist-tmp-BkZU97
Traceback (most recent call last):
File "/usr/bin/easy_install", line 8, in
load_entry_point('setuptools==0.6c12dev-r85381', 'console_scripts', 'easy_install')()
File "/usr/lib/python2.5/site-packages/easy_install.py", line 1712, in main

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 1700, in with_ei_usage

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 1716, in <lambda>

  File "/usr/lib/python2.5/distutils/core.py", line 151, in setup
dist.run_commands()
  File "/usr/lib/python2.5/distutils/dist.py", line 974, in run_commands
self.run_command(cmd)
  File "/usr/lib/python2.5/distutils/dist.py", line 994, in run_command
cmd_obj.run()
  File "/usr/lib/python2.5/site-packages/easy_install.py", line 211, in run

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 446, in easy_install

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 476, in install_item

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 655, in install_eggs

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 930, in build_and_install

  File "/usr/lib/python2.5/site-packages/easy_install.py", line 919, in run_setup

  File "/usr/lib/python2.5/site-packages/setuptools-0.6c12dev_r85381-py2.5.egg/setuptools/sandbox.py", line 62, in run_setup
  File "/usr/lib/python2.5/site-packages/setuptools-0.6c12dev_r85381-py2.5.egg/setuptools/sandbox.py", line 105, in run
  File "/usr/lib/python2.5/site-packages/setuptools-0.6c12dev_r85381-py2.5.egg/setuptools/sandbox.py", line 64, in <lambda>
  File "setup.py", line 66, in <module>
  File "/usr/lib/python2.5/distutils/core.py", line 151, in setup
dist.run_commands()
  File "/usr/lib/python2.5/distutils/dist.py", line 974, in run_commands
self.run_command(cmd)
  File "/usr/lib/python2.5/distutils/dist.py", line 994, in run_command
cmd_obj.run()
  File "/usr/lib/python2.5/site-packages/setuptools-0.6c12dev_r85381-py2.5.egg/setuptools/command/bdist_egg.py", line 167, in run
  File "/usr/lib/python2.5/distutils/cmd.py", line 333, in run_command
self.distribution.run_command(command)
  File "/usr/lib/python2.5/distutils/dist.py", line 994, in run_command
cmd_obj.run()
  File "/usr/lib/python2.5/site-packages/setuptools-0.6c12dev_r85381-py2.5.egg/setuptools/command/egg_info.py", line 170, in run
  File "/usr/lib/python2.5/site-packages/setuptools-0.6c12dev_r85381-py2.5.egg/setuptools/command/egg_info.py", line 379, in write_pkg_info
  File "/usr/lib/python2.5/distutils/dist.py", line 1076, in write_pkg_info
self.write_pkg_file(pkg_info)
  File "/usr/lib/python2.5/distutils/dist.py", line 1101, in write_pkg_file
file.write('Description: %s\n' % long_desc)
UnicodeEncodeError: 'ascii' codec can't encode character u'\u2019' in position 1292: ordinal not in range(128)

via Peter Bengtsson

I've upgraded to Pymongo 1.9 and MongoKit 0.5.13-dev and this code
used to work just fine before. Here's my structure (simplified)::

class BaseDocument(Document):
   structure = {
     'add_date': datetime.datetime,
     'modify_date': datetime.datetime,
   }

   default_values = {
     'add_date': datetime.datetime.now,
     'modify_date': datetime.datetime.now
   }
   use_autorefs = True
   use_dot_notation = True

class User(BaseDocument):
   structure = {
     'guid': unicode,
     'username': unicode,
     'email': unicode,
     'password': unicode,
     'first_name': unicode,
     'last_name': unicode,
   }

   required_fields = ['guid']
   default_values = {
     'guid': lambda:unicode(uuid.uuid4()),
   }

   indexes = [
     {'fields': 'guid',
      'unique': True},
   ]

class Event(BaseDocument):
   structure = {
     'user': User,
     'title': unicode,
   }
   required_fields = ['user', 'title']

   indexes = [
     {'fields': ['user']},
   ]

And here is my unit test that now breaks:

   def test_create_event(self):
       user = self.db.users.User()
       user.save()
       event = self.db.events.Event()
       event.user = user
       event.title = u"Test"
       print event.user
       event.validate()
       event.save()

Traceback (most recent call last):
 File "/home/peterbe/dev/TORNADO/worklog/tests/test_models.py", line
52, in test_create_event
   event.validate()
 File "/home/peterbe/virtualenvs/worklog/lib/python2.6/site-packages/
mongokit-0.5.13_dev-py2.6.egg/mongokit/document.py", line 249, in
validate
   super(Document, self).validate()
 File "/home/peterbe/virtualenvs/worklog/lib/python2.6/site-packages/
mongokit-0.5.13_dev-py2.6.egg/mongokit/schema_document.py", line 351,
in validate
   self._validate_required(self, self.structure)
 File "/home/peterbe/virtualenvs/worklog/lib/python2.6/site-packages/
mongokit-0.5.13_dev-py2.6.egg/mongokit/schema_document.py", line 772,
in _validate_required
   self._raise_exception(RequireFieldError, req, "%s is required" %
req)
 File "/home/peterbe/virtualenvs/worklog/lib/python2.6/site-packages/
mongokit-0.5.13_dev-py2.6.egg/mongokit/schema_document.py", line 519,
in _raise_exception
   raise exception(message)
RequireFieldError: user is required

It succeeds if I make 'user' NOT a required field.

MongoKit only allows INDEX_ASCENDING and INDEX_DESCENDING. However, since 1.5.1 mongodb also has a GEO2D index specifier. Please allow this.

The type of strings used in a pymongo collection.find() need to match. The original code doesn't find just inserted IDs, because it's not looking for a unicode string.
The one liner patch (excuse the cut/paste) below corrects the issue for me (it now finds the relevant record).
StopIteration occurs due to the generator iterating over an empty query result set when its .next() method is called.
The test did not fail before; I'm presently working on modifying the test so that it fails before the patch, below, is applied. I'll send a pull request at that time if you'd prefer.

$ hg diff diff -r 43dd8afd9305 mongokit/versioned_document.py --- a/mongokit/versioned_document.py  Mon Aug 09 16:12:14 2010 +0200 +++ b/mongokit/versioned_document.py Thu Aug 26 23:53:36 2010 +1000 @@ -97,7 +97,7 @@
def get_last_revision_id(self): last_doc = self.versioning_collection.find(
- {'id':self['_id']}).sort('revision', -1).next()
+ {'id': unicode(self['_id'])}).sort('revision', -1).next() if last_doc: return last_doc['revision']

was "remove automatique index creation and add ensure_index method on Document"

see http://groups.google.com/group/mongokit/t/50c0c9ab4a313010

http://gist.github.com/636126

from Christian Joudrey
We could add a dict called "structure_aliases" to Document which would serve as an alias when performing writes and reads.
For instance:

class User(Document):
    structure = {
      'u': unicode,
      'p': unicode
    }
    required_fields = [
      'u', 'p'
    ]
    structure_aliases = {
      'u': 'username',
      'p': 'password'
    }

The size of a collection can dramatically increase in size when there are many columns with long names.
Structure aliases would allow one to perform read and writes using "username" and "password" as column names, but they would really be stored as "u" and "p" in MongoDb thus saving (in theory) 14 bytes per row.

for more details see : http://groups.google.com/group/mongokit/t/c2e483cafa0b29cc

Sample code:

from mongokit import Connection, Document, DocumentMigration

class TestDocMigration(DocumentMigration):
  pass

class TestDoc(Document):
  structure = {'name': unicode}
  migration_handler = TestDocMigration

con = Connection()
con.register([TestDoc])

collection = con['test'].foos

a = collection.TestDoc()
a['name'] = a

b = collection.TestDoc({'name': 'b'})

The latter syntax (with b) doesn't seem to be documented, but it works until you add migrations to your document. Once you add a migration handler, a KeyError is raised because MongoKit thinks the document has already been saved to the db:

...
File "python2.6/site-packages/mongokit/document.py", line 132, in _migrate
  old_doc = self.collection.get_from_id(self['_id'])
KeyError: '_id'

I suppose the solution is to either (a) discourage the use of the constructing documents with an initial structure — I'm not sure how I started doing that since it doesn't seem to be documented anyway — or (b) prevent migration of documents which haven't yet been saved.

MongoKit is great. Keep up the good work!

When I try to setup pylons with mongokit I get the error

from mongokit.ext.pylons_env import MongoPylonsEnv
ImportError: No module named ext.pylons_env

from the line:

from mongokit.ext.pylons_env import MongoPylonsEnv

from this document: http://bytebucket.org/namlook/mongokit/wiki/html/pylons_support.html
Have mongokit and mongokit-pylons not been fully merged yet? Though I tried using the origin mongokit-pylons and i still get the same error. Any help would be appreciated.

Add constant SELF to indicate documents referencing itself:

http://gist.github.com/636121

At the moment all Document classes must use MongoDocumentCursor.

It should be possible to instead have a cursor_class class property (with MongoDocumentCursor as default) to allow a custom cursor class to be used instead - for example like the "objects" property in the Django ORM.

As auto referencing is currently implement, there can be no circular references.
This is not allowed currently: Doc1 referencing Doc2 and Doc2 referencing Doc1
Circular references are common in graph data structures. Currently, to implement circular references the user needs to implement their own dbref functions, and can not use the mongokit's auto referencing feature.
Graphs are a common data structure as can be seen in the numerous non-sql graph databases. Adding the ability to handle circular references in mongokit would greatly enhance its usefulness.

class BlogPostMigration(DocumentMigration):
   def allmigration01__add_update_field_and_fill_it(self):
       self.target = {'blog_post.update_date':{'$exists':False}, 'blog_post':{'$exists':True}}
       if not self.status:
           for doc in self.collection.find(self.target):
               self.update = {'$set':{'blog_post.update_date': self.doc['blog_post']['creation_date']}}
               self.collection.update(self.target, self.update, multi=True, safe=True)

It's doc['blog_post']['creation_date'] not self.doc

via Phillip Oldham:

Would it be possible to enable mongokit to use the python with statement?

For instance:

>>> from mongokit import *
>>> class MyObj(MongoDocument):
...   structure = {
...     "settings": { unicode: OR(unicode, bool, int) }
...   }
... 
>>> document = MyObj()
>>> with document.settings:
...   foo = 'bar'
... 
>>> print document.settings.foo
bar

Not completely sure that's accurate, but even using:

>>> with document.settings as s: 
...   s.foo = 'bar'

would help keep things a little neater when working with larger objects/structures.

Hello,
in the section about 'indexes', the field 'checked'
http://namlook.github.com/mongokit/tutorial.html#indexes
should be under ttl

also a typo: grep 'liste' and replace by 'list'

see http://stackoverflow.com/questions/5229470/mongokit-schema-index-on-fields-in-array

MongoKit tip, testcase in attachment:

gauss@pmain:~/Projects/mongokit_test$ ./test1.py 
Traceback (most recent call last):
  File "./test1.py", line 20, in 
    a.save()
  File "/home/gauss/python-site-packages/mongokit/document.py", line 362, in save
    self.validate(auto_migrate=False)
  File "/home/gauss/python-site-packages/mongokit/document.py", line 174, in validate
    super(Document, self).validate()
  File "/home/gauss/python-site-packages/mongokit/schema_document.py", line 290, in validate
    self._validate_doc(self, self.structure)
  File "/home/gauss/python-site-packages/mongokit/schema_document.py", line 538, in _validate_doc
    self._validate_doc(doc[key], struct[key],  new_path)
  File "/home/gauss/python-site-packages/mongokit/schema_document.py", line 549, in _validate_doc
    self._validate_doc(obj, struct, path)
  File "/home/gauss/python-site-packages/mongokit/schema_document.py", line 517, in _validate_doc
    "missed fields : %s" % struct_doc_diff )
  File "/home/gauss/python-site-packages/mongokit/schema_document.py", line 472, in _raise_exception
    raise exception(message)
mongokit.schema_document.StructureError: missed fields : ['a2']

But a2 isn't marked as required.

see http://groups.google.com/group/mongokit/browse_thread/thread/242aaaa73cb4b457

I'm pretty new to MongoKit so please bear with me if I'm confused, but it doesn't seem to be possible to specify default values for embedded objects. Using this Document:

class UserInfo(Document):
    use_dot_notation = True
    structure = {
        'email_addresses': [{
            'email': unicode,
            'verified': bool,
        }],
    }
    required_fields = []
    default_values = {
        'email_addresses': [{
            'verified': True,
        }],
    }

raises this exception at object-instantiation time:

Traceback (most recent call last):
  File "./mongofun.py", line 3, in <module>
    from core.model import con
  File "...core/model.py", line 12, in <module>
    class UserInfo(Document):
  File ".../lib/python2.7/site-packages/mongokit/document.py", line 76, in __new__
    return SchemaProperties.__new__(cls, name, bases, attrs)        
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 154, in __new__
    cls._validate_descriptors(attrs)
  File ".../lib/python2.7/site-packages/mongokit/document.py", line 80, in _validate_descriptors
    SchemaProperties._validate_descriptors(attrs)
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 174, in _validate_descriptors
    raise ValueError("Error in default_values: can't find %s in structure" % dv )
ValueError: Error in default_values: can't find email_addresses in structure

Changing the required_fields parameter to this instead:

default_values = { 'email_addresses.verified': False, }

raises this exception at validation time, because MongoKit hasn't used the defaults to fill out the fields:

Traceback (most recent call last):
  File "./mongofun.py", line 23, in <module>
    user.validate()
  File ".../lib/python2.7/site-packages/mongokit/document.py", line 247, in validate
    super(Document, self).validate()
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 348, in validate
    self._validate_doc(self, self.structure)
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 585, in _validate_doc
    self._validate_doc(doc[key], struct[key],  new_path)
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 596, in _validate_doc
    self._validate_doc(obj, struct, path)
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 564, in _validate_doc
    "missed fields : %s" % struct_doc_diff )
  File ".../lib/python2.7/site-packages/mongokit/schema_document.py", line 519, in _raise_exception
    raise exception(message)
mongokit.schema_document.StructureError: missed fields : ['verified']

This issue has been brought up before, and the solution suggested there indicates that it's supposed to work:

https://groups.google.com/forum/#!topic/mongokit/ZjPjE9CgO0E

Via structure introspection, we can guess the type of field and then generate some fixtures on the fly.
The api could be :

>>> connection.mydb.mycol.Doc.fixtures(how_many=10)

this will load to mycol 10 Doc objects with random value.
Thanks to Vladimir Dronnikov for the original idea...

Would it be usefull to enable/disable safe mode in general ?

class MyDoc(Document):
   safe_insert = True # by default
   structure = {...}

This will be implemented into the pymongo driver (http://jira.mongodb.org/browse/PYTHON-181) and maybe a MongoKish API should be built...

see http://groups.google.com/group/mongokit/t/65b99d22d2523c66

hello you should tag released version, it will automatically make an archive in the download section.

i got this exception

Exception Value:
init() got an unexpected keyword argument 'from_son'
Exception Location: /usr/local/lib/python2.6/dist-packages/mongokit-0.5.12-py2.6.egg/mongokit/database.py in _fix_outgoing, line 73

it seems Document constructor doesn't have this argument:
def init(self, doc=None, gen_skel=True, collection=None, lang='en', fallback_lang='en', generate_index=True):

see details : http://groups.google.com/group/mongokit/t/d06e28de5db0939

pymongo encodes uuid.UUID types as bson.binary.UUID_SUBTYPE and it would be nice if mongokit supported this too.

pymongo support: https://gist.github.com/933686
mongokit: https://gist.github.com/933670

hello,
i think this is a bug:
default_values = {'blog_post.created_at':datetime.utcnow()}

should be datetime.utcnow (the callable)

otherwise all document will have the same date

It would be useful if there was a .simplify() method on Document objects, which would recursively convert the object to the base data-types as defined in the document's structure, eg:

class Foo(Document):
    structure = {
      'title': unicode,
      'properties': {unicode:None}
     }

x = Foo()
x['title'] = u'my document'
x['properties'] = {u'active': True}
x.save()

print type(x) # mongokit.Document
print type(x['_id']) # pymongo.objectid.ObjectId
print type(x['properties']) # mongokit.helpers.DotedDict

y = x.simplify()
print type(x) # dict
print type(x['_id']) # unicode
print type(x['properties']) # dict

There's a test case for this issue here: https://gist.github.com/795871

It appears that using a custom type in a structure dict with a type as the key causes a key error to be raised. For example:

structure = {
    'dates' : {unicode: CustomDate()}
}

would cause an error on validation, but not:

structure = {
    'dates' : {unicode: datetime.datetime}
}

The exception raised is:

Traceback (most recent call last):
  File "testcase.py", line 32, in <module>
    doc.validate()
  File "/Users/allanc/externals/mongokit/mongokit/document.py", line 247, in validate
    super(Document, self).validate()
  File "/Users/allanc/externals/mongokit/mongokit/schema_document.py", line 347, in validate
    self._process_custom_type('bson', self, self.structure)
  File "/Users/allanc/externals/mongokit/mongokit/schema_document.py", line 656, in _process_custom_type
    self._process_custom_type(target, doc[key], struct[key], new_path, root_path)
  File "/Users/allanc/externals/mongokit/mongokit/schema_document.py", line 642, in _process_custom_type
    if not isinstance(doc[key], struct[key].python_type) and doc[key] is not None:
KeyError: <type 'unicode'>

http://namlook.github.com/mongokit/tutorial.html#adding-complex-validation

It suggest code that doesn't work. Here's a better piece of example code:

def validate(self, *args, **kwargs):
    assert self['foo'] > self['bar']
    assert self['baz'] == unicode(self['foo'])
    super(MyDoc, self).validate(*args, **kwargs)

This is necessary to prevent some buildout issues:

via Andrew Degtiariov:

I have pymongo 1.9.0, but there is buildout.
It overrides python path:

#!/opt/python/bin/python -S

import sys
sys.path[0:0] = [
    '/opt/project/parts/buildout',
    ]


import os
path = sys.path[0]
if os.environ.get('PYTHONPATH'):
    os.environ['BUILDOUT_ORIGINAL_PYTHONPATH'] = os.environ['PYTHONPATH']
    path = os.pathsep.join([path, os.environ['PYTHONPATH']])
os.environ['PYTHONPATH'] = path
import site # imports custom buildout-generated site.py

import zc.buildout.buildout

if __name__ == '__main__':
    zc.buildout.buildout.main()

pymongo installed into /opt/project/eggs directory, so ./bin/buildout known nothing about pymongo during processing setup.py.
Could you make version independent from external imports? E.g. could you make it as string as you was doing in previous versions of MongoKit?