Converting Django project from MySQL to Mongo, any major pitfalls? - python

I want to try Mongodb w/ mongoengine. I'm new to Django and databases and I'm having a fit with Foreign Keys, Joins, Circular Imports (you name it). I know I could eventually work through these issues but Mongo just seems like a simpler solution for what I am doing. My question is I'm using a lot of pluggable apps (Imagekit, Haystack, Registration, etc) and wanted to know if these apps will continue to work if I make the switch. Are there any known headaches that I will encounter, if so I might just keep banging my head with MySQL.

There's no reason why you can't use one of the standard RDBMSs for all the standard Django apps, and then Mongo for your app. You'll just have to replace all the standard ways of processing things from the Django ORM with doing it the Mongo way.
So you can keep urls.py and its neat pattern matching, views will still get parameters, and templates can still take objects.
You'll lose querysets because I suspect they are too closely tied to the RDBMS models - but they are just lazily evaluated lists really. Just ignore the Django docs on writing models.py and code up your database business logic in a Mongo paradigm.
Oh, and you won't have the Django Admin interface for easy access to your data.

You might want to check out django-nonrel, which is a young but promising attempt at a NoSQL backend for Django. Documentation is lacking at the moment, but it works great if you just work it out.

I've used mongoengine with django but you need to create a file like mongo_models.py for example. In that file you define your Mongo documents. You then create forms to match each Mongo document. Each form has a save method which inserts or updates whats stored in Mongo. Django forms are designed to plug into any data back end ( with a bit of craft )
BEWARE: If you have very well defined and structured data that can be described in documents or models then don't use Mongo. Its not designed for that and something like PostGreSQL will work much better.
I use PostGreSQL for relational or well structured data because its good for that. Small memory footprint and good response.
I use Redis to cache or operate in memory queues/lists because its very good for that. great performance providing you have the memory to cope with it.
I use Mongo to store large JSON documents and to perform Map and reduce on them ( if needed ) because its very good for that. Be sure to use indexing on certain columns if you can to speed up lookups.
Don't circle to fill a square hole. It won't fill it.
I've seen too many posts where someone wanted to swap a relational DB for Mongo because Mongo is a buzz word. Don't get me wrong, Mongo is really great... when you use it appropriately. I love using Mongo appropriately

Upfront, it won't work for any existing Django app that ships it's models. There's no backend for storing Django's Model data in mongodb or other NoSQL storages at the moment and, database backends aside, models themselves are somewhat of a moot point, because once you get in to using someones app (django.contrib apps included) that ships model-template-view triads, whenever you require a slightly different model for your purposes you either have to edit the application code (plain wrong), dynamically edit the contents of imported Python modules at runtime (magical), fork the application source altogether (cumbersome) or provide additional settings (good, but it's a rare encounter, with django.contrib.auth probably being the only widely known example of an application that allows you to dynamically specify which model it will use, as is the case with user profile models through the AUTH_PROFILE_MODULE setting).
This might sound bad, but what it really means is that you'll have to deploy SQL and NoSQL databases in parallel and go from an app-to-app basis--like Spacedman suggested--and if mongodb is the best fit for a certain app, hell, just roll your own custom app.
There's a lot of fine Djangonauts with NoSQL storages on their minds. If you followed the streams from the past Djangocon presentations, every year there's been important discussions about how Django should leverage NoSQL storages. I'm pretty sure, in this year or the next, someone will refactor the apps and models API to pave the path to a clean design that can finally unify all the different flavors of NoSQL storages as part of the Django core.

I have recently tried this (although without Mongoengine). There are a huge number of pitfalls, IMHO:
No admin interface.
No Auth django.contrib.auth relies on the DB interface.
Many things rely on django.contrib.auth.User. For example, the RequestContext class. This is a huge hindrance.
No Registration (Relies on the DB interface and django.contrib.auth)
Basically, search through the django interface for references to django.contrib.auth and you'll see how many things will be broken.
That said, it's possible that MongoEngine provides some support to replace/augment django.contrib.auth with something better, but there are so many things that depend on it that it's hard to say how you'd monkey patch something that much.

Primary pitfall (for me): no JOINs!

Related

Handling repetitive content within django apps

I am currently building a tool in Django for managing the design information within an engineering department. The idea is to have a common catalogue of items accessible to all projects. However, the projects would be restricted based on user groups.
For each project, you can import items from the catalogue and change them within the project. There is a requirement that each project must be linked to a different database.
I am not entirely sure how to approach this problem. From what I read, the solution I came up with is to have multiple django apps. One represents the common catalogue of items (linked to its own database) and then an app for each project(which can write and read from its own database but it can additionally read also from the common items catalogue database). In this way, I can restrict what user can access what database/project. However, the problem with this solution is that it is not DRY. All projects look the same: same models, same forms, same templates. They are just linked to different database and I do not know how to do this in a smart way (without copy-pasting entire files cause I think managing this would be a pain).
I was thinking that this could be avoided by changing the database label when doing queries (employing the using attribute) depending on the group of the authenticated user. The problem with this is that an user can have access to multiple projects. So, I am again at a loss.
It looks for me that all you need is a single application that will manage its access properly.
If the requirement is to have separate DBs then I will not argue that, but ... there is always small chance that separate tables in 1 DB is what they will accept
Django apps don't segregate objects, they are a way of structuring your code base. The idea is that an app can be re-used in other projects. Having a separate app for your catalogue of items and your projects is a good idea, but having them together in one is not a problem if you have a small codebase.
If I have understood your post correctly, what you want is for the databases of different departments to be separate. This is essentially a multi-tenancy question which is a big topic in itself, there are a few options:
Code separation - all of your projects/departments exist in a single database and schema but are separate by code that filters departments depending on who the end user is (literally by using Django .filters()). This is easy to do but there is a risk that data could be leaked to the wrong user if you get your code wrong. I would recommend this one for your use-case.
Schema separation - you are still using a single database but each department has its own schema. You would need to use Postgresql for this but once a schema has been set, there is far less chance that data is going to be visible to the wrong user. There are some Django libraries such as django-tenants that can do a lot of the heavy lifting.
Database separation - each department has their own database. There is even less of a chance that data will be leaked but you have to manage multi-databases and it is more difficult to scale. You can manage this through django as there is support for multi-databases.
Application separation - each department not only has their own database but their own application instance. The separation is absolute but again you need to manage multiple applications on a host like Heroku, which is even less scalable.

Django: Bypassing the database abstraction

I have been playing arround with django for a couple of days and it seems great, but I find it a pain if I want to change the structure of my database, I then am stuck with a few rather awkward options.
Is there a way to completely bypass djangos database abstraction so if I change the structure of the database I dont have to guess what model would have generated it or use a tool (south or ...) to change things?
I essentially want this: https://docs.djangoproject.com/en/dev/topics/db/sql/ (Raw SQL Queries) but instead of refering to a model, refering to an external database.
Could I just create an empty model and then only perform raw queries on it? (and set up the DB externally)
Thanks
P.S. I dont really mind if I have separate databases for the admin stuff and the app data
It's in your question already, just read the docs article from here: Executing custom SQL directly

Building a DSL query language

i'm working on a project (written in Django) which has only a few entities, but many rows for each entity.
In my application i have several static "reports", directly written in plain SQL. The users can also search the database via a generic filter form. Since the target audience is really tech-savvy and at some point the filter doesn't fit their needs, i think about creating a query language for my database like YQL or Jira's advanced search.
I found http://sourceforge.net/projects/littletable/ and http://www.quicksort.co.uk/DeeDoc.html, but it seems that they only operate on in-memory objects. Since the database can be too large for holding it in-memory, i would prefer that the query is translated in SQL (or better a Django query) before doing the actual work.
Are there any library or best practices on how to do this?
Writing such a DSL is actually surprisingly easy with PLY, and what ho—there's already an example available for doing just what you want, in Django. You see, Django has this fancy thing called a Q object which make the Django querying side of things fairly easy.
At DjangoCon EU 2012, Matthieu Amiguet gave a session entitled Implementing Domain-specific Languages in Django Applications in which he went through the process, right down to implementing such a DSL as you desire. His slides, which include all you need, are available on his website. The final code (linked to from the last slide, anyway) is available at http://www.matthieuamiguet.ch/media/misc/djangocon2012/resources/compiler.html.
Reinout van Rees also produced some good comments on that session. (He normally does!) These cover a little of the missing context.
You see in there something very similar to YQL and JQL in the examples given:
groups__name="XXX" AND NOT groups__name="YYY"
(modified > 1/4/2011 OR NOT state__name="OK") AND groups__name="XXX"
It can also be tweaked very easily; for example, you might want to use groups.name rather than groups__name (I would). This modification could be made fairly trivially (allow . in the FIELD token, by modifying t_FIELD, and then replacing . with __ before constructing the Q object in p_expression_ID).
So, that satisfies simple querying; it also gives you a good starting point should you wish to make a more complex DSL.
I've faced exactly this problem - a large database which needs searching. I made some static reports and several fancy filters using django (very easy with django) just like you have.
However the power users were clamouring for more. I decided that there already was a DSL that they all knew - SQL. The question was how to make it secure enough.
So I used django permissions to give the power users permission to make SQL queries in a new table. I then made a view for the not-quite-so-power users to use these queries. I made them take optional parameters. The queries were run using Python's lower level DB-API which django is using under the hood for its ORM anyway.
The real trick was opening a read only database connection to run these queries just to make sure that no updates were ever run. I made a read only connection by creating a different user in the database with lower permissions and opening a specific connection for that in the view.
TL;DR - SQL is the way to go!
Depending on the form of your data, the types of queries your users need to use, and the frequency that your data is updated, an alternative to the pure SQL solution suggested by Nick Craig-Wood is to index your data in Solr and then run queries against it.
Solr is an added layer of complexity (configuration, data synchronization) but it is super-fast, can handle large datasets, and provides a (relatively) intuitive query language.
You could write your own SQL-ish language using pyparsing, actually. There is even pretty verbose example you could extend.

Is there any python web app framework that provides database abstraction layer for SQL and NoSQL?

Is it even possible to create an abstraction layer that can accommodate relational and non-relational databases? The purpose of this layer is to minimize repetition and allows a web application to use any kind of database by just changing/modifying the code in one place (ie, the abstraction layer). The part that sits on top of the abstraction layer must not need to worry whether the underlying database is relational (SQL) or non-relational (NoSQL) or whatever new kind of database that may come out later in the future.
There's a Summer of Code project going on right now to add non-relational support to Django's ORM. It seems to be going well and chances are good that it will be merged into core in time for Django 1.3.
You could use stock Django and Django-nonrel ( http://www.allbuttonspressed.com/projects/django-nonrel ) together to get a quite unified experience. Some limits apply, read docs carefully though, remembering Spolsky's "All abstractions are leaky".
Yo may also check web2py, they support relational databases and GAE on the core.
Regarding App Engine, all existing attempts limit you in some way (web2py doesn't support transactions or namespaces and probably many other stuff, for example). If you plan to work with GAE, use what GAE provides and forget looking for a SQL-NoSQL holy grail. Existing solutions are inevitably limited and affect performance negatively.
Thank you for all the answers. To summarize the answers, currently only web2py and Django supports this kind of abstraction.
It is not about a SQL-NoSQL holy grail, using abstraction can make the apps more flexible. Lets assume that you started a project using NoSQL, and then later on you need to switch over to SQL. It is desirable that you only make changes to the codes in a few spots instead of all over the place. For some cases, it does not really matter whether you store the data in a relational or non-relational db. For example, storing user profiles, text content for dynamic page, or blog entries.
I know there must be a trade off by using the abstraction, but my question is more about the existing solution or technical insight, instead of the consequences.

How to generate basic CRUD functionality in python given a database tables

I want to develop a desktop application using python with basic crud operation. Is there any library in python that can generate a code for CRUD functionality and user interface given a database table.
Hopefully, this won't be the best option you end up with, but, in the tradition of using web-interfaces for desktop applications, you could always try django. I would particularLY take a look at the inspectdb command, which will generate the ORM code for you.
The advantage is that it won't require that much code to get off the ground, and if you just want to use it from the desktop, you don't need a webserver; you can use the provided test server. The bundled admin site is easy to get off the ground, and flexible up to a point; past which people seem to invest a lot of time battling it (probably a testimony to how helpful it is at first).
There are many disadvantages, not the least of which is the possibility of having to use html/javascript/css when you want to start customizing a lot.
If it were me, I would consider borrowing django's ORM, but then again, I'm already familiar with it.
Having said that, I like working with it, it's usable outside the framework, and it will give you mysql, postgres, or sqlite support. You could also hook up the django admin site to your models and have a web-based editor.
There are surely other ORMs and code generators out there too (I hope some python gurus will point some out, I'm kind of curious).
If you want something really small and simple, I like the Autumn ORM.
If you use the Django ORM, you can use the automatically-generated Django admin interface, which is really nice. It's basically a web-based GUI for browsing and editing records in your database.
If you think you will need advanced SQL features, SQLAlchemy is a good way to go. I suspect for a desktop application, Django or Autumn would be better.
There are other Python ORMs, such as Storm. Do a Google search on "python ORM". See also the discussion on this web site: What are some good Python ORM solutions?

Categories

Resources