How do you track record relations in NoSQL? - nosql

I am trying to figure out the equivalent of foreign keys and indexes in NoSQL KVP or Document databases. Since there are no pivotal tables (to add keys marking a relation between two objects) I am really stumped as to how you would be able to retrieve data in a way that would be useful for normal web pages.
Say I have a user, and this user leaves many comments all over the site. The only way I can think of to keep track of that users comments is to
Embed them in the user object (which seems quite useless)
Create and maintain a user_id:comments value that contains a list of each comment's key [comment:34, comment:197, etc...] so that that I can fetch them as needed.
However, taking the second example you will soon hit a brick wall when you use it for tracking other things like a key called "active_comments" which might contain 30 million ids in it making it cost a TON to query each page just to know some recent active comments. It also would be very prone to race-conditions as many pages might try to update it at the same time.
How can I track relations like the following in a NoSQL database?
All of a user's comments
All active comments
All posts tagged with [keyword]
All students in a club - or all clubs a student is in
Or am I thinking about this incorrectly?

All the answers for how to store many-to-many associations in the "NoSQL way" reduce to the same thing: storing data redundantly.
In NoSQL, you don't design your database based on the relationships between data entities. You design your database based on the queries you will run against it. Use the same criteria you would use to denormalize a relational database: if it's more important for data to have cohesion (think of values in a comma-separated list instead of a normalized table), then do it that way.
But this inevitably optimizes for one type of query (e.g. comments by any user for a given article) at the expense of other types of queries (comments for any article by a given user). If your application has the need for both types of queries to be equally optimized, you should not denormalize. And likewise, you should not use a NoSQL solution if you need to use the data in a relational way.
There is a risk with denormalization and redundancy that redundant sets of data will get out of sync with one another. This is called an anomaly. When you use a normalized relational database, the RDBMS can prevent anomalies. In a denormalized database or in NoSQL, it becomes your responsibility to write application code to prevent anomalies.
One might think that it'd be great for a NoSQL database to do the hard work of preventing anomalies for you. There is a paradigm that can do this -- the relational paradigm.

The couchDB approach suggest to emit proper classes of stuff in map phase and summarize it in reduce.. So you could map all comments and emit 1 for the given user and later print out only ones. It would require however lots of disk storage to build persistent views of all trackable data in couchDB. btw they have also this wiki page about relationships: http://wiki.apache.org/couchdb/EntityRelationship.
Riak on the other hand has tool to build relations. It is link. You can input address of a linked (here comment) document to the 'root' document (here user document). It has one trick. If it is distributed it may be modified at one time in many locations. It will cause conflicts and as a result huge vector clock tree :/ ..not so bad, not so good.
Riak has also yet another 'mechanism'. It has 2-layer key name space, so called bucket and key. So, for student example, If we have club A, B and C and student StudentX, StudentY you could maintain following convention:
{ Key = {ClubA, StudentX}, Value = true },
{ Key = {ClubB, StudentX}, Value = true },
{ Key = {ClubA, StudentY}, Value = true }
and to read relation just list keys in given buckets. Whats wrong with that? It is damn slow. Listing buckets was never priority for riak. It is getting better and better tho. btw. you do not waste memory because this example {true} can be linked to single full profile of StudentX or Y (here conflicts are not possible).
As you see it NoSQL != NoSQL. You need to look at specific implementation and test it for yourself.
Mentioned before Column stores look like good fit for relations.. but it all depends on your A and C and P needs;) If you do not need A and you have less than Peta bytes just leave it, go ahead with MySql or Postgres.
good luck

user:userid:comments is a reasonable approach - think of it as the equivalent of a column index in SQL, with the added requirement that you cannot query on unindexed columns.
This is where you need to think about your requirements. A list with 30 million items is not unreasonable because it is slow, but because it is impractical to ever do anything with it. If your real requirement is to display some recent comments you are better off keeping a very short list that gets updated whenever a comment is added - remember that NoSQL has no normalization requirement. Race conditions are an issue with lists in a basic key value store but generally either your platform supports lists properly, you can do something with locks, or you don't actually care about failed updates.
Same as for user comments - create an index keyword:posts
More of the same - probably a list of clubs as a property of student and an index on that field to get all members of a club

You have
"user": {
"userid": "unique value",
"category": "student",
"metainfo": "yada yada yada",
"clubs": ["archery", "kendo"]
}
"comments": {
"commentid": "unique value",
"pageid": "unique value",
"post-time": "ISO Date",
"userid": "OP id -> THIS IS IMPORTANT"
}
"page": {
"pageid": "unique value",
"post-time": "ISO Date",
"op-id": "user id",
"tag": ["abc", "zxcv", "qwer"]
}
Well in a relational database the normal thing to do would be in a one-to-many relation is to normalize the data. That is the same thing you would do in a NoSQL database as well. Simply index the fields which you will be fetching the information with.
For example, the important indexes for you are
Comment.UserID
Comment.PageID
Comment.PostTime
Page.Tag[]
If you are using NosDB (A .NET based NoSQL Database with SQL support) your queries will be like
SELECT * FROM Comments WHERE userid = ‘That user’;
SELECT * FROM Comments WHERE pageid = ‘That user’;
SELECT * FROM Comments WHERE post-time > DateTime('2016, 1, 1');
SELECT * FROM Page WHERE tag = 'kendo'
Check all the supported query types from their SQL cheat sheet or documentation.

Although, it is best to use RDBMS in such cases instead of NoSQL, yet one possible solution is to maintain additional nodes or collections to manage mapping and indexes. It may have additional cost in form of extra collections/nodes and processing, but it will give an solution easy to maintain and avoid data redundancy.

Related

Why there are two refs in declaring one-to-many association in mongoose?

I'm very new in mongodb, see this one-to-many example
As per my understanding
This example says that a person can write many stories or a story belongs_to a person , I think storing the person._id in stories collection was enough
why the person collection has the field stories
cases for fetching data
case 1
Fetch all stories of a person whose id is let us say x
solution: For this just fire a query in story collection where author = x
case 2
Fetch the author name of a particular story
solution: For this we have author field story collection
TL;DR
Put simply: Because there is no such notion as explicit relations in MongoDB.
Mongoose can not know how you want to resolve the relationship. Will the search be from a given story object and the author is to find? Or will the search be to find all stories for an author object? So it makes sure that it can resolve the relation regardless.
Note that there is a problem with that approach, and a big one. Say we are not talking of a one-to-few relation as in this example, but a "One-To-A-Shitload"™ relation. Since BSON documents have a size limit of 16MB, you have a limit of relations you can manage this way. Quite some, but there will be an artificial limit.
How to solve this: Instead of using an ODM, do proper modelling yourself. Since you know your use cases. I will give you an example below.
Detailed
Let us first elaborate your cases a bit:
For a given user (aka "we already have all the data of that user document"), what are his or her stories?
List all stories together with the user name on an overview page.
For a selected ("given") story, what are the authors details?
And just for demonstration purposes: A given user wants to change the name under which a story is displayed, be it his user name or natural name (it happens!) or even pseudonym.
Ok, and now lets put mongoose aside for now and let us think about how we could implement this ourselves. Keeping in mind that
Data modelling in MongoDB is deriving your model from the questions which come from your use cases so that they most common use cases are covered in the most efficient way.
As opposed to RDBMS modelling, where you identify your entities, their properties and relations and then jump through some hoops to get your questions answered somehow.
So, looking at our user stories, I guess we can agree that 2 is the most common use case, 3 and 1 next and 4 is rather rare compared to the other ones.
So now we can start
Modelling
We model the data involved in our most common use cases first.
So, we want to make the query for stories the most efficient one. And we want to sort the stories in descending order of submission. Simple enough:
{
_id: new ObjectId(),
user: "Name to Display",
story: "long story cut short",
}
Now lets say you want to display your stories, 10 of them:
db.stories.find({}).sort({_id:-1}).limit(10)
No relation, all the data we need, a single query, used the default index on _id for sorting. Since a timestamp is part of the ObjectId and it is the most significant part, we can use it to sort the stories by time. The question "Hey, but what if one changes his or her user name?" usually comes now. Simple:
db.stories.update({"user":oldname},{$set:{"user":newname}},{multi:true})
Since this is a rare use case, it only has to be doable and does not have to be extremely efficient. However, later we will see that we have to put an index on user anyway.
Talking of authors: Here it really depends on how you want to do it. But I will show you how I tend to model something like that:
{
_id: "username",
info1: "foo",
info2: "bar",
active: true,
...
}
We make use of some properties of _id here: It is a required index with a unique constraint. Just what we want for usernames.
However it comes with a caveat: _id is immutable. So if somebody wants to change his or her username, we need to copy the original user to a document with the _id of the new user name and change the user property in our stories accordingly. The advantage of this way of doing it that even when the update for changing usernames (see above) should fail during its runtime, each and every story can still be related to a user. If the update is successful, I tend to log out the user and have him log in with the new username again.
In case you want to have a distinction between username and displayed name, it all becomes even easier:
{
_id: "username",
displayNames: ["Foo B. Baz","P.S. Eudonym"],
...
}
Then you use the display name in your stories, of course.
Now let us see how we can get the user details of a given story. We know the author's name so it is as easy as:
db.authors.find({"_id":authorNameOfStory})
or
db.authors.find({"displayNames": authorNameOfStory})
Finding all stories for a given user is quite simple, too. It is either:
db.stories.find({"name":idFieldOfUser})
or
db.stories.find({"name":{$in:displayNamesOfUser}})
Now we have all your our use cases covered, now we can make them even more efficient with
Indexing
An obvious index is on the story models user field, so we do it:
db.stories.ensureIndex({"name":1})
If you are good with the "username as _id" way only, you are done with indexing. Using display names, you obviously need to index them. Since you most likely want display names and pseudonyms to be unique, it is a bit more complicated:
db.authors.ensureIndex({"displayNames":1},{sparse:true, unique:true})
Note: We need to make this as sparse index in order to prevent unnecessary errors when somebody has not decided for a display name or pseudonym yet. Make sure you explicitly add this field to an author document only when a user decides for a display name. Otherwise, it would evaluate to null server side , which is a valid value and you will get a constraint violation error, namely "E1100 duplicate key".
Conclusion
We have covered all your use cases with relations handled by the application thereby simplifying our data model a great deal and have the most efficient queries for our most common use cases. Every use case is covered with a single query, taking into account the information we already have at the time we are doing the query.
Note that there is no artificial limit on how many stories a user can publish since we use implicit relations to our advantage.
As for more complicated queries ("How many stories does each user submit per month?"), use the aggregation framework. That is what it is there for.

nosql wishlist models - Struggle between reference and embedded document

I got a question about modeling wishlists using mongodb and mongoose. The idea is I need a user beeing able to have many different wishlists which contain many wishes, each wish making a reference to a single article
I was thinking about it and because a wishlist only belong to a single user I thought using embedded document for that.
Same for the wish beeing embedded to a wishlist.
So I got something like that
var UserSchema = new Schema({
...
wishlists: [wishlistSchema]
...
})
var WishlistSchema = new Schema({
...
wishes: [wishSchema]
...
})
but my question is what to do with the article ? should I use a reference or should I copy the article's data in an embedded document.
If I use embedded document I got an update problem. When the article's price change, to update every wish referencing this article become a struggle. But to access those wishes's article is a piece of cake.
If I use reference, The update is not a problem anymore but I got a probleme when I filter the wish depending on their article criteria ( when I filter the wishes depending on a price, category etc .. ).
I think the second way is probably the best but I don't know how if it's possible to build a query to filter the wish depending on the article's field. I tried a lot of things using population but nothing works very well when you need to populate depending on a nested object field. ( for exemple getting wishes where their article respond to certain conditions ).
Is this kind of query doable ?
Sry for the loooong question and for my bad English :/ but any advice would be great !
In my experience in dealing with NoSQL database (mongo, mainly), when designing a collection, do not think of the relations. Instead, think of how you would display, page, and retrieve the documents.
I would prefer embedding and updating multiple schema when there's a change, as opposed to doing a ref, for multiple reasons.
Get would be fast and easy and filter is not a problem (like you've said)
Retrieve operations usually happen a lot more often than updates and with proper indexing, you wouldn't really have to bother about performance.
It leverages on NoSQL's schema-less nature and you'll be less prone restructuring due to requirement changes (new sorting, new filters, etc)
Paging would be a lot less of a hassle, and UI would not be restricted with it's design with paging and limit.
Joining could become expensive. Redundant data might be a hassle to update but it's always better than not being able to display a data in a particular way because your schema is normalized and joining is difficult.
I'd say that the rule of thumb is that only split them when you do not need to display them together. It is not impossible to join them back if you do, but definitely more troublesome.

How to deal with many users update over Document driven nosql database

Problem
Starting with nosql document database I figured out lots of new possibilities, however, I see some pitfalls, and I would like to know how can I deal with them.
Suppose I have a product, and this product can be sold in many regions. There is one responsible for each region (with access to CMS). Each responsible modifies the products accordingly regional laws and rules.
Since Join feature isn't supported as we know it on relational databases, the document should be design in a way it contains all the needed information to build our selection statements and selection result to avoid round trips to the database.
So my first though was to design a document that follows more or less this structure:
{
type : "product",
id : "product_id",
title : "title",
allowedAge : 12,
regions : {
'TX' : {
title : "overriden title",
allowedAge : 13
},
'FL' : {
title : "still another title"
}
}
}
But I have the impression that this approach will generate conflicts while updating the document. Suppose we have a lot of users updating lots of document through a CMS. When same document is updated, the last update overwrites the updates done before, even the users are able to modify just fragments of this document (in this case the responsible should be able to modify just the regional data).
How to deal with this situation?
One possible solution I think of would be partial document updates. Positive: reducing the data overwriting from different operations, Negative: lose the optimistic lock feature since locking if done over a document not a fragment of such.
Is there another approach for the problem?
In this case you can use 3 solutions:
Leave current document structure and always check CAS value on update operations. If CAS doesn't match - call store function again. (But as you say if you have a lot of users it can be very slow).
Separate doc in several parts that could be updated independently, and then on app-side combine them together. This will result in increasing view calls (one for get main doc, another call to get i.e. regions, etc.). If you have many "parts" it will also reduce performance.
See this doc. It's about simulating joins in couchbase. There is also good example written by #Tug Grall.
If you're not bounded to using Couchbase (not clear from your question if it's general or specific to it) - look also into MongoDB. It supports partial updates on documents and also other atomic operations (like increments and array operations), so it might suite your use case better (checkout possible update operations on mongo - http://docs.mongodb.org/manual/core/update/ )

Blogs and Blog Comments Relationship in NoSQL

Going off an example in the accepted answer here:
Mongo DB relations between objects
For a blogging system, "Posts should be a collection. post author might be a separate collection, or simply a field within posts if only an email address. comments should be embedded objects within a post for performance."
If this is the case, does that mean that every time my app displays a blog post, I'm loading every single comment that was ever made on that post? What if there are 3,729 comments? Wouldn't this brutalize the database connection, SQL or NoSQL? Also there's the obvious scenario in which when I load a blog post, I want to show only the first 10 comments initially.
Document databases are not relational databases. You CANNOT first build the database model and then later on decide on various interesting ways of querying it. Instead, you should first determine what access patterns you want to support, and then design the document schemas accordingly.
So in order to answer your question, what we really need to know is how you intend to use the data. Displaying comments associated with a post is a distinctly different scenario than displaying all comments from a particular author. Each one of those requirements will dictate a different design, as will supporting them both.
This in itself may be useful information to you (?), but I suspect you want more concrete answers :) So please add some additional details on your intended usage.
Adding more info:
There are a few "do" and "don'ts" when deciding on a strategy:
DO: Optimize for the common use-cases. There is often a 20/80 breakdown where 20% of the UX drives 80% of the load - the homepage/landing page is a classic example. First priority is to make sure that these are as efficient as possible. Make sure that your data model allows either A) loading those in either a single IO request or B) is cache-friendly
DONT: don't fall into the dreaded "N+1" trap. This pattern occurs when you data model forces you to make N calls in order to load N entities, often preceded by an additional call to get the list of the N IDs. This is a killer, especially together with #3...
DO: Always cap (via the UX) the amount of data which you are willing to fetch. If the user has 3729 comments you obviously aren't going to fetch them all at once. Even it it was feasible from a database perspective, the user experience would be horrible. Thats why search engines use the "next 20 results" paradigm. So you can (for example) align the database structure to the UX and save the comments in blocks of 20. Then each page refresh involves a single DB get.
DO: Balance the Read and Write requirements. Some types of systems are read-heavy and you can assume that for each write there will be many reads (StackOverflow is a good example). So there it makes sense to make writes more expensive in order to gain benefits in read performance. For example, data denormalization and duplication. Other systems are evenly balanced or even write heavy and require other approaches
DO: Use the dimension of TIME to your advantage. Twitter is a classic example: 99.99% of tweets will never be accessed after the first hour/day/week/whatever. That opens all kinds of interesting optimization possibilities in the your data schema.
This is just the tip of the iceberg. I suggest reading up a little on column-based NoSQL systems (such as Cassandra)
Not sure if this answers you question, but anyhow you can throttle the amount of blog comments in two ways:
Load only the last 10 , or range of blog comments using $slice operator
db.blogs.find( {_id : someValue}, { comments: { $slice: -10 } } )
will return last 10 comments
db.blogs.find( {_id : someValue}, { comments: { $slice: [-10, 10] } } )
will return next 10 comments
Use capped array to save only the last n blog posts using capped arrays

MongoDB Schema Design With or Without Null Placeholders

I still getting used to using a schema-less document oriented database and I am wondering what a generally accepted practice is regarding schema designs within an application model.
Specifically I'm wondering whether it is a good practice to use enforce a schema within the application model when saving to mongodb like this:
{
_id: "foobar",
name: "John"
billing: {
address: "8237 Landeau Lane",
city: "Eden Prairie",
state: "MN",
postal: null
}
balance: null,
last_activity: null
}
versus only storing the fields that are used like this:
{
_id: "foobar",
name: "John"
billing: {
address: "8237 Landeau Lane",
city: "Eden Prairie",
state: "MN"
}
}
The former is self-descriptive which I like, while the latter makes no assumptions on the mutability of the model schema.
I like the first option because it makes it easy to see at a glance what fields are used by the model yet currently unspecified, but it seems like it would be a hassle to update every document to reflect a new schema design if I wanted to add an extra field, like favorite_color.
How do most veteran mongodb users handle this?
I would suggest second approach.
You can always see the intended structure if you look at your entity class in the source code. Or do you use dynamic language, and don't create an entity?
You save a lot of space per record, because you don't have to store null column names. This may not be expensive on small collections. But on large, with millions of records, I would even go to shorten the names of fields.
As you already mentioned. By specifying optional column names, you create a pattern, which, if you want to follow, you'll have to update all existing records when you add a new field. This is, again, a bad idea for a big DB.
In any case it all goes down your db size. If you don't target for many GBs or TBs of data, then both approaches are fine. But, if you predict, that your DB may grow really large, I would do anything to cut the size. Spending 30-40% of storage for column names is a bad idea.
I prefer the first option, it is easier to code within the application and requires much less state holders and functions to understand how things should work.
As for adding a new field over time you don't need to update all your records to support this new field like you would in SQL all you need to do is write the new field into your model application side and support this field being null if it is not returned from MongoDB.
A good example is in PHP.
I have a class of user at first with only one field, name
class User{
public $name;
}
6 months down the line and 60,000 users later I want to add, say, address. All I do is add that variable to my application model:
class User{
public $name;
public $address = array();
}
This now works exactly like adding a new null field to SQL without having to actually add it to every row on-demand.
It is a very reactive design, don't update what you don't need to. If that row gets used it will get updated, if not then who cares.
So eventually your rows actually become a mix and match between option 1 and 2 but it is really a reactive option 1.
Edit
On the storage side you have also got to think of pre-allocation and movement of documents.
Say the amount of a set record now is only a third of the doc but then suddenly, from the user updating the doc with all of the fields, you now have extra fragmentation from the movement of your docs.
Normally when you are defining a schema like this you are defining one that will eventually grow and apply to that user in most cases (much like an SQL schema does).
This is something to take into consideration that even though storage might be lower in the short term it could cause fragmentation and slow querying due to that fragmentation and you could easily find yourself having to run compacts or repairDbs due to the problems you now face.
I should mention that both of those functions I said above are not designed to be run regularly and have a significant performance problem to them while they run on a production environment.
So really with the structure above you don't need to add a new field across all documents and you will most likely get less movement and problems in the long run.
You can fix the performance problems of consistently growing documents by using power of 2 sizes padding, but then this is collection wide which means that even your fully filled documents will use up at least double their previous space and you small documents will probably be using as much space as your full documents would have on a padding factor of 1.
Aka you lose space, not gain it.