gamechanger/mongothon

A lightweight Mongo object-document mapping abstraction layer over PyMongo

mongothon's Issues

Optimistic locking for updates

Transaction with database

Hi Guys

Please Is the library allows transactions ?

Import error

I'm importing create_model:

from mongothon import create_model

I get:

Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/runpy.py", line 170, in _run_module_as_main
    "__main__", mod_spec)
  File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "src/__main__.py", line 3, in <module>
    from database.mongodb.Collection import Collection
  File "src/database/mongodb/Collection.py", line 1, in <module>
    from mongothon import create_model
  File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/site-packages/mongothon/__init__.py", line 3, in <module>
    from document import Document
ImportError: No module named 'document'

Any ideas?

[QUESTION] Connect to DB using Mongothon

Traits

Would be useful to be able to declare traits like in monufacture for commonly used items like timestamps, versioning fields etc.

Need to thing how this might work with virtual fields etc.

Make subdocument nodes independently validatable

Subnodes in documents have associated sub-schemas, so why not make them validatable?

This would go a long way to making atomic updates less scary as a frequent requirement is to add a subnode to an embedded array. Being able to validate that subnode prior to the atomic update would be useful.

Model.find_one is returning empty model instead of None

This might not be strictly "wrong", and the return value is still falsy since it evaluates as an empty dict, but I suspect that this difference in behavior from pymongo might trip some people up.

Document the will_reload and did_reload standard events

Dirty tracking

Document should track the dirty state of fields. Here's a sketch of the idea:

    user = User.find_by_id(uid)
    user['first_name'] = 'bob'
    user['email'] = '[email protected]'
    user['referrers'].append('[email protected]')
    user['prefs']['alerts'] = True

    user.changed('first_name')      # => True
    user.changed()                  # => {'first_name', 'email'}
    user.changes                    # => {'first_name': ('robert', 'bob'), 'email': ('[email protected]', '[email protected]')}
    user.change('first_name')       # => ('robert', 'bob')
    user.was('first_name')          # => 'robert'

    user['prefs'].changed('alerts') # => True

    user['referrers'].added         # => ['[email protected]']
    user['referrers'].removed       # => []

    user.save()

    user.changed()                  # => []
    user.previous.changes           # => {'first_name': ('robert', 'bob'), 'email': ('[email protected]', '[email protected]')}
    user.previous.changed()         # => {'first_name', 'email'}

Virtual fields

Mongothon used to have virtual fields, and they sucked, mainly because they were part of the model.

Instead, we should extend Schema to allow virtual fields to be declared with the model's schema and be accessible as a dictionary accessor.

Sketch

# Create a schema
user_schema = Schema({
    'first_name': {'type': unicode, 'required': True},
    'last_name': {'type': unicode, 'required': True},
})

# Register virtual field getter
@user_schema.get('display_name')
def display_name(doc):
    return "{} {}".format(doc['first_name'], doc['last_name'])


# Consume virtual field
print user['display_name']

Document.__init__ doesn't allow keyword args

With the python dict constructor you can do:

d = dict({'foo': 'bar'}, baz='bat')
d == {'foo': 'bar', 'baz': 'bat'}

Document.__init__ currently only allows for a single (dictionary) argument. Should we change this?

is_new? and dirty? tracking on models

Support for declaring indexes

Schemas (or possibly models) should allow indexes to be declared.

will_reload / did_reload events

Atomic operation builder DSL

Example:

BlogPost.update(id, atop.increment('likes').push(tags='cooking'))

Rather than...

BlogPost.update(id, {"$inc": {"likes": 1}, "$push": {"tags": "cooking"}})

Advantages:

• Slightly better readability
• "Compile-time" safety

Python3 support

It would be nice to Mongothon had python3 support.

Thanks!

Update to Schemer 0.2.0

Model instance update errors out on multi-character keyword arguments

A model instance update of the form

car = Car()
car.update(maker='porsche')

raises a ValueError due to a dangling comma in Document.update

See

mongothon/mongothon/document.py

Line 23 in 6c91cb3

for key, in kwargs:

Single-character kwargs actually work fine since they get unpacked into a length-1 tuple.

Expose Schemer validators

To save needing to import schemer.validators, we should create a proxy at mongothon.validators, potentially adding Mongo-specific validators in future.

Rename Model.update

The update method on Model proxies to the underlying pymongo collection's update method. This is great, except it hides the update method on the underlying dict type.

We should rename it to something else so the dict version is still usable.

cc @byels

Refactor schemas out into separate library for reuse

Schemas are useful for more than just Mongo!

Make ScopeBuilder smarter

Right now ScopeBuilder requires an explicit call to execute() or a __getitem__ call to actually invoke the query. It should be a lot smarter:

• Implement __iter__ forwarding to an underlying cursor, so we don't force Python to use the slow "try getting until we get an index error" method when iterating.
• Forward any cursor methods to an underlying cursor
• Eliminate the need for execute entirely. ScopeBuilder should function both as a builder and a proxy to the underlying query.

Query building DSL

What's needed is something similar to mongoose.js query building syntax.

Example:

Person.find(query.where('name.last').equals('Ghost')
                 .where('age').gt(17).lt(66)
                 .where('likes').in(['vaporizing', 'talking'])
                 .limit(10)
                 .sort('-occupation')
                 .select('name occupation'))

No License

Hi, thanks for your work on this project.

There is no license attached to the project, do which license do you intend to release this code under?

Thanks.