Related
I was wondering if you have a website with a dozen different types of listings (Shops, Restaurants, Clubs, Hotels, Events) that require different fields, is there a benefit of creating a table with columns defined like so
Example Shop:
shop_id | name | X | Y | city | district | area | metro | station | address | phone | email | website | opening_hours
Or a more abstract approach similar to this:
object_id | name
---------------
1 | Messy Joe's
2 | Bate's Motel
type_id | name
---------------
1 | hotel
2 | restaurant
object_id | type_id
---------------
1 | 2
2 | 1
field_id | name | field_type
---------------
1 | address | text
2 | opening_hours | date
3 | speciality | text
type_id | field_id
---------------
1 | 1
1 | 2
2 | 1
2 | 3
object_id | field_id | value
1 | 1 | 1st street....
1 | 3 | English Cuisine
Of course it can be more abstract if value's are predefined (Example: specialties could have their own list)
If I take the abstract approach it can be very flexible, but queries will be more complex with a lot of joins.
But I don't know if this affects the performance, executing these 'more complex' queries.
I would be interested to know what are the up and downsides of both methods. I can just imagine for myself, but I don't have the experience to confirm this.
Certain issues need to be clarified and resolved before we can enter into a reasonable discussion.
Pre-requisite Resolution
Labels
In a profession that demands precision, it is important that we use precise labels, to avoid confusion, and so that we can communicate without having to use long-winded descriptions and qualifiers.
What you have posted as FixedTables, is Unnormalised. Fair enough, it may be an attempt at Third Normal form, but in fact it is a flat file, Unnormalised (not "denormalised). What you have posted as AbstractTables is, to be precise, Entity-Attribute-Value, which is almost, but not quite, Sixth Normal form, and is therefore more Normalised than 3NF. Assuming it is done correctly, of course.
The Unnormalised flat file is not "denormalised". It is chock full of duplication (nothing has been done to remove repeating groups and duplicate columns or to resolve dependencies) and Nulls, it is a performance hog in many ways, and prevents concurrency.
In order to be Denormalised, it has to first be Normalised, and then the Normalisation backed off a little for some good reason. Since it is not Normalised in the first place, it cannot be Denormalised. It is simply Unnormalised.
It cannot be said to be denormalised "for performance", because being a performance hog, it is the very antithesis of performance. Well, they need a justification for the lack of formalised design], and "for performance" is it. Even the smallest formal scrutiny exposed the misrepresentation (but very few people can provide, so it remains hidden, until they get an outsider to address, you guessed it, the massive performance problem).
Normalised structures perform far better than Unnormalised structures. More normalised structures (EAV/6NF) perform better than less normalised structures (3NF/5NF).
I am agreeing with the thrust of OMG Ponies, but not their labels and definitions
rather than saying 'don't "denormalise" unless you have to', I am saying, 'Normalise faithfully, period' and 'if there is a performance problem, you have not Normalised correctly'.
Wikipedia
The entries for Normal Forms and Normalisation offer definitions that are incorrect; they confuse the Normal Forms; they are lacking regarding the process of Normalisation; and they give equal weight to absurd or questionable NFs which have been debunked long ago. The result is, Wikipedia adds to an already confused and rarely understood subject. So don't waste your time.
However, in order to progress, without that reference posing a hindrance, let me say this.
The definition of 3NF is stable, and has not changed.
There is a lot of confusion of the NFs between 3NF and 5NF. The truth is that this is an area that progressed over the last 15 years; and many orgs, academics as well as vendors with their products with limitations, jumped to create a new "Normal Form" to validate their offerings. All serving commercial interests and academically unsound. 3NF in its original untampered state intended and guaranteed certain attributes.
The sum total is, 5NF is today, what 3NF was intended to be 15 years ago, and you can skip the commercial banter and the twelve or so "special" (commercial and pseudo-academic) NFs in-between, some of which are identified in Wikipedia, and even that in confusing terms.
Fifth Normal Form
Since you have been able to understand and implement the EAV in your post, you will have no problem understanding the following. Of course a true Relational Model is pre-requisite, strong keys, etc. Fifth Normal Form is, since we are skipping the Fourth:
Third Normal Form
which in simple definitive terms is, every non-key column in every table has a 1::1 relationship to the Primary Key of the table,
and to no other non-key columns
Zero data duplication (the result, if Normalisation is progressed diligently; not achieved by intelligence or experience alone, or by working toward it as a goal without the formal process)
no Update Anomalies (when you update a column somewhere, you do not have to update the same column located somewhere else; the column exists in one and only one place).
If you understand the above, 4NF, BCNF, and all the silly "NFs" can be dismissed, they are required for physicalised Record Filing Systems, as promoted by academics, quite foreign to the Relational Model (Codd).
Sixth Normal Form
The purpose is elimination of missing data (attribute columns), aka elimination of Nulls
This is the one true solution to the Null Problem (also called Handling Missing Values), and the result is a database without Nulls. (It can be done at 5NF with standards and Null substitutes but that is not optimal.) How you interpret and display the missing values is another story.
Technically, is not a true Normal Form, because it does not have 5NF as a pre-requisite, but it has a value
EAV vs Sixth Normal Form
All the databases I have written, except one, are pure 5NF. I have worked with (administered, fixed up, enhanced) a couple of EAV databases, and I have implemented many true 6NF databases. EAV is a loose implementation of 6NF, often done by people who do not have a good grasp on Normalisation and the NFs, but who can see the value in, and need the flexibility of, EAV. You are a perfect example.
The difference is this: because it is loose, and because implementers do not have a reference (6NF) to be faithful to, they only implement what they need, and they write it all in code; that ends up being an inconsistent model.
Whereas, a pure 6NF implementation does have a pure academic reference point, and thus it is usually tighter, and consistent. Typically this shows up in two visible elements:
6NF has a catalogue to contain metadata, and everything is defined in metadata, not code. EAV does not have one, everything is in code (implementers keep track of the objects and attributes). Obviously a catalogue eases the addition of columns, navigation, and allows utilities to be formed.
6NF when understood, provides the true solution to The Null Problem. EAV implementers, since they are absent the 6NF context, handle missing data in code, inconsistently, or worse, allow Nulls in the database. 6NF implementers disallow Nulls, and handle missing Data consistently and elegantly, without requiring code constructs (for Null handling; you still have to code for missing data of course).
Eg. For 6NF databases with a catalogue, I have a set of procs that will [re]generate the SQL required to perform all SELECTs, and I provide Views in 5NF for all users, so they do not need to know or understand the underlying 6NF structure. They are driven off the catalogue. Thus changes are easy and automated. EAV types do that manually, due to the absence of the catalogue.
Discussion
Now, we can start the discussion.
"Of course it can be more abstract if
value's are predefined (Example:
specialities could have their own
list)"
Sure. But do not get too "abstract". Maintain consistency and implement such lists in the same EAV (or 6NF) manner as other lists.
"If I take the abstract approach it
can be very flexible, but queries will
be more complex with a lot of joins.
But I don't know if this affects the
performance, executing these 'more
complex' queries."
Joins are pedestrian in relational databases. The problem is not the database, the problem is that SQL is cumbersome when handling joins, especially compound keys.
EAV and 6NF databases have more Joins, which just as pedestrian, no more, no less. If you have to code each SELECT manually, sure, the cumbersome gets really cumbersome.
The entire problem can be eliminated by (a) going with 6NF over EAV and (b) implementing a catalogue, from which you can (c) generate all the basic SQL. Eliminates an entire class of errors as well.
It is a common myth that Joins somehow have a cost. Totally false.
The join is implemented at compile time, there is nothing of substance to 'cost' CPU cycles.
The issue is the size of tables being joined, not the cost of the Join between those same tables.
Joining two tables with millions of rows each, on a correct PK⇢FK relation, each of which have the appropriate indices
(Unique on the parent [PK] side; Unique on the Child side [PK=parent FK + something]
is instantaneous
Where the Child index is not unique, but at least the leading columns are valid, it is slower; where there is no useful index, of course it is very slow.
None of it has to do with Join cost.
Where many rows are returned, the bottleneck will be the network and the disk layout; not the join processing.
Therefore you can get as "complex" as you like, there is no cost, SQL can handle it.
I would be interested to know what are
the up and downsides of both methods.
I can just imagine for myself, but I
don't have the experience to confirm
this.
5NF (or 3NF for those who have not made the progression) is the easiest and best, in terms of implementation; ease of use (developers as well as users); and maintenance.
The drawback is, every time you add a column, you have to change the database structure (table DDL). That is fine is some cases, but not in most cases, due to change control in place, quite onerous.
Second, you have to change existing code (code handling the new column does not count, because that is an imperative): where good standards are implemented, that is minimised; where they are absent, the scope is unpredictable.
EAV (which is what you have posted), allows columns to be added without DDL changes. That is the single reason people choose it. (code handling the new column does not count, because that is an imperative). If implemented well, it will not affect existing code; if not, it will.
But you need EAV-capable developers.
When EAV is implemented badly, it is abominable, a worse mess than 5NF done badly, but not any worse than Unnormalised which is what most databases out there are (misrepresented as "denormalised for performance").
Of course, it is even more important (than in 5NF/3NF) to hold a strong Transaction context, because the columns are far more distributed.
Likewise, it is essential to retain Declarative Referential Integrity: the messes I have seen were due in large part to the developers removing DRI because it became "too hard to maintain", the result was, as you can imagine, one mother of a data heap with duplicate 3NF/5NF rows and columns all over the place. And inconsistent Null handling.
There is no difference in performance, assuming that the server has been reasonably configured for the intended purpose. (Ok, there are specific optimisations that are possible only in 6NF, which are not possible in other NFs, but I think that is outside the scope of this thread.) And again, EAV done badly can cause unnecessary bottlenecks, no more so than Unnormalised.
Of course, if you go with EAV, I am recommending more formality; buy the full quid; go with 6NF; implement a catalogue; utilities to produce SQL; Views; handle Missing Data consistently; eliminate Nulls altogether. This reduces your vulnerability to the quality of your developers; they can forget about the EAV/6NF esoteric issues, use Views, and concentrate on the app logic.
In your question, you have presented at least two major issues at the same time. Those two issues are E-A-V and gen-spec.
First, let's talk about E-A-V. Your last table (object_id, field_id, value) is essentially an E-A-V. There is an upside to E-A-V and a downside to E-A-V. The upside is that the structure is so generic that it can accomodate almost any body of data describing almost any subject matter. That means that you can proceed to design and implementation with no data analysis and no understanding of the subject matter, and not worry about wrong assumptions. The down side is that at retrieval time, you have to do the data analysis that you skipped over before building the data base, in order to come up with queries that mean anything. This is much more serious than just retrieval efficiency. But you are also going to have terrible problems with retrieval efficiency. There are only two ways to learn about this pitfall: live through it or read about it from those who have. I recommend the reading.
Second, you have a gen-spec case. Your table (object_id, type_id) captures a gen-spec (generalization-specialization) pattern, along with the related tables. If I had to generalize between hotels and restaurants, I might call it something like "public accomodations" or "venues". But I'm not sure I understand your case, and you may be driving for something even more general than those two names suggest. After all, you've included "events" in your list, and an event is not a type of venue in my mind.
I've referred other people to readings on gen-spec and the relational model in previous responses.
When two tables are very similar, when should they be combined?
But I hesitate to send you off in the same direction, because it's not clear to me that you want to come up with a relational model of the data before building your database. A relational model of a body of data and an E-A-V model of the same data are almost totally at odds with each other. It seems to me you have to make that choice before you even explore how to express gen-spec in the relational model of data.
When you start to require a large number of different entities (or even before...), a nosql solution would be vastly simpler than either choice.
Just store each entity/record with the exact fields you require.
{
"id": 1,
"type":"Restaurant",
"name":"Messy Joe",
"address":"1 Main St.",
"tags":["asian","fusion","casual"]
}
The "abstract" approach is better known as "Normalization", looks like 3rd Normal Form (3NF).
The other one is called "Denormalized", and can be a valid performance option... when you've encountered speed issues using the Normalized approach, not before.
How do you have the listings represented in code? I'd guess Listing as a supertype, with Shop, Restuarant, etc. as subtypes?
Assuming so, this is a case of how to map subtypes to a relational database. There are generally three choices:
Option 1: single table per subtype,
with common attributes repeated in
each table (name, id, etc).
Option 2: single table for all objects (your single table approach)
Option 3: table for the supertype and one for each subtype
There's no universally correct solution. My preference is generally to start with option 3; it provides an intituitive structure to work with, is pretty well normalised and can easily be extended. It means a single join for retrieving each instance - but RDBMS are well optimised for doing joins so it doesn't really cause performance problems in practice.
Option 2 can be more performant for queries (no joins) but causes problems if other tables need to refer to all supertype instances (proliferation of foreign keys).
Option 1 appears at first sight to be the most performant, although 2 caveats: (1) It's not resilient to change. If you add a new subtype (and so different attributes) you'll need to change the table structure and migrate it. (2) It can be less efficient than it seems. Because the table population is sparse, some DBs don't store it particularly efficiently. As a consequence it can be less efficicent than option 1 - since the query engine can do joins faster than it can search bloated sparse table spaces.
Which to choose really comes down to knowing details of your problem. I'd suggest reading up a bit on the options: this article is a good place to start.
hth
Here's a simple example of an SQL table:
CREATE TABLE persons
(
id INTEGER,
name VARCHAR(255),
height DOUBLE
);
Since I haven't used SQL very much, I haven't yet learned to think in its terms. Effectively, my brain translates the above into this:
struct Person
{
int id;
string name;
double height;
Person(int id_, const char* name_, double height_)
:id(id_),name(name_),height(height_)
{}
};
Person persons[64];
Then, inserting some elements, in SQL:
INSERT INTO persons (id, name, height) VALUES (1234, 'Frank', 5.125);
INSERT INTO persons (id, name, height) VALUES (5678, 'Jesse', 6.333);
...and how I'm thinking of it:
persons[0] = Person(1234, "Frank", 5.125);
persons[1] = Person(5678, "Jesse", 6.333);
I've read that SQL can be thought of as two major parts: data manipulation and data definition. I'm more concerned about organizing my data in the first place, as opposed to querying and modifying it. There, the distinctions of SQL are immediately obvious. To me, it seems like the subtleties of how data can and should be structured in SQL is a more obscure topic. Where does the array-of-structs analogy I'm automatically drawing for myself break down?
To give a concrete example, let's say that I want each entry in my persons table (or each of my Person objects) to contain a field denoting the names of that person's children (actual fruit-of-your-loins children, not hierarchical data structure children). In reality, these would probably be cross-table references (or pointers to objects), but let's keep things simple and make this field contain zero or more names. In my C++ example, I'd modify the declaration like so:
vector<string> namesOfChildren;
...and do something like this:
persons[0].namesOfChildren.push_back("John");
persons[0].namesOfChildren.push_back("Jane");
But, from what I can tell, the typical usage of SQL doesn't mirror this approach. If I'm wrong and there's a simple, straightforward solution, great. If not, I'm sure a SQL novice like myself could benefit greatly from a little cogitation on the subject of how databases of SQL tables are meant to be used in contrast to bare, generic data structures.
To me, it seems like the subtleties of how data can and should be structured in SQL is a more obscure topic.
It's called "data(base) modeling" and is somewhere between engineering discipline and art (like much of the computer programming). If you are really interested in the topic, take a look at ERwin Methods Guide.
Where does the array-of-structs analogy I'm automatically drawing for myself break down?
At persistency, concurrency, consistency and scalability.
Persistency: The table is automatically saved to the permanent storage. It'll stay there and survive reboots (not that a real database server will reboot much) until you explicitly delete it or there is a catastrophic hardware failure. DBMSes have well-oiled backup procedures that should help in the latter case.
Concurrency: Tables are meant to be accessed and (need be) modified by many clients concurrently. Mechanisms such as locking and multi-version concurrency control are employed to ensure clients will not "step on each other's toes".
Consistency: You can define certain constraints (such as uniqueness, foreign keys or checks) and the DBMS will make sure they are never broken. Furthermore, this can often be done in a declarative manner, minimizing chance for errors. On top of that, everything you do in a database is transactional, so you reap the benefits of atomicity, consistency, isolation and durability (aka. "ACID"). In a nutshell, the database will defend itself from bad data.
Scalability: A well designed database schema can grow well beyond the confines of the available RAM, and still keep good performance, using techniques such as indexing, partitioning, clustering etc... Furthermore, SQL is declarative and set-based, which means that the DBMS has the latitude to pick the optimal "query execution plan" for the data at hand, auto-parallelize the query, cache the results in hope they will be reused etc... without changing the meaning of the query.
Your analogy to the array of structs is not bad ... for the beginning.
After this beginning the differences start in relation to organizing data.
Database people love their "Normal Forms" laws. We do not have these laws in C++ or similar programming languages. Organizing data in the tables according to these laws help database engines to do their magic (queries, joins) better, i.e keep databases compact and crunch millions of rows in fractions of a second, and allow multiple requests concurrently. They are not absolute laws: the 1NF (1st Normal Form) is followed in 99.9999% cases, but the bigger the number (2NF, 3NF, ...) the more often DB planners allow themselves to deviate from them.
Description of normal forms can be found for example here.
I will try to illustrate differences on your example.
In your example the fields of your struct correspond to the columns of the database table. Adding vector of the names as a new field of struct would correspond to adding comma separated list of the names into a new column of your table. This is a violation of the 1NF which demands that one cell is for one value - not for the list of values. To normalize your data you will need to have two arrays: one of Person structs, and another new of structs for Child. While in C++ we can use just pointers to link each child to its parent, in SQL we must use the mechanism of the key. You already added id field into Person struct, now we need to add ParentId field to Child struct so that database engine could find the Parent. ParentId column is called foreign key. Another approach to satisfy 1NF instead of creating the new table/struct for children is that we can switch to children-centric thinking and have just one table with a record per child which will include all the information about the parent of the child. Info about the parent obviously will be repeated in as many records as many children this parent has.
Note (this is also considered part of 1NF) that while in the array of structs we always know the order of the elements, in databases it is up to the engine in what order to store the records. It is just mathematical un-ordered set of records, the engine can resort it in internal storage for various optimizations as it likes. When you retrieve the records from the database with the SELECT statement, if you care about the order, you need to provide ORDER BY clause.
2NF is about removing repetitions from your records. Imagine you would have place of work related fields also as part of your Person struct. Imagine it would include Name of the company and company address. If many Persons in your dataset work in the same Company your would repeat address of the company in their records. Probably we wouldn't do these repetitions in C++ either, but nevertheless extracting these repetitions into a separate table would satisfy 2NF. Strictly speaking even if there is no repetitions and all your Persons work in different places, 2NF still requires to extract data about the workplaces into separate table because it requires that one table would represent one entity.
3NF is about removing transitive dependency and is considered kind of optional, so I will not describe it here. See link above.
Another feature of databases quite different from conventional programming of data structures in C++ is the database indexes. Simplifying, index is just a copy of a column (or columns) (i.e vertical slice) into a separate table where they are stored in an inherent for them order and each record in the index retains the reference to the whole record. So, in your example to create index by height you would create another array of 64 elems of the new
struct HeightIndexElem
{
double height;
Person* pFullRecord;
}
and sort them by height in this array. This will allow the DB engine to automatically optimize certain queries. The database engine itself decides when to use certain index. In C++ we usually create maps (Dictionaries in C#) to speed up finding element by certain characteristic but we must use these maps ourselves - no automatic aspect there.
There are major differences:-
SQL tables are persistent -- (English Tran: written to disk)
They are transactional -- (really written to disk)
They can be an arbitary size -- (Tables of a several hundred million rows are quite common)
They support relational algebra -- (Joins with other tables, filtering etc.)
Relational Algebra is provable -- For a given SELECT statement there is only one possible correct answer.
The biggest differences are that when you "UPDATE" and "COMMIT" you know your data is saved in the database and will be there until you decide to "DELETE" it. When you update a structure within an array its gone when the machine is switched off.
The other big difference is scale. The size of a modern DBMS is only limited by your hard disk budget.
[I really like farfareast's answer from an academic stand point, but I feel the need to add a more practically oriented answer too.]
SQL tables themselves are "bare, generic data structures" as you call C++'s structures. They are only different data structures: a table is always an array of (fixed size) structs and the only pointers you can use are foreign keys.
For example, when you are adding a vector<string> to your struct, you are already using pointers internally as strings are only a "fancy" way of writing char*. This would already require a second table in SQL (using a secondayr index column to keep the elements in order). Of course there are things like postgresql's arrays that can help in this specific case, but those are "only" shortcuts for similar hand-writeable constructs.
The real difference in data structure and algorithms comes from the fact that you can easily add declarations of index structures. Say you know you need to always access Persons in the order of their height. In C++ you'd use some kind of tree or sorted list to keep them in order. There is an STL container for that. The downside is, that when you need to access them in a different order (say by name), you'll have to add a second tree and duplicate the data or start using pointers to Persons. If you add a Person, you need to update all containers and so on. This becomes cumbersome and soon you'll be on the front page of The Daily WTF. SQL tables on the other side can have attached indices which automatically keep up with new and changed data. Of course, their maintenance also must be paid in performance, but the management of them is basically deciding which are required by your access patterns -- something needed in every case -- and defining them. In contrast to having to rewrite large parts of an application, this is a much more favorable situation.
What I'm doing
I am creating an SQL table that will provide the back-end storage mechanism for complex-typed objects. I am trying to determine how to accomplish this with the best performance. I need to be able to query on each individual simple type value of the complex type (e.g. the String value of a City in an Address complex type).
I was originally thinking that I could store the complex type values in one record as an XML, but now I am concerned about the search performance of this design. I need to be able to create variable schemas on the fly without changing anything about the database access layer.
Where I'm at now
Right now I am thinking to create the following tables.
TABLE: Schemas
COLUMN NAME DATA TYPE
SchemaId uniqueidentifier
Xsd xml //contains the schema for the document of the given complex type
DeserializeType varchar(200) //The Full Type name of the C# class to which the document deserializes.
TABLE: Documents
COLUMN NAME DATA TYPE
DocumentId uniqueidentifier
SchemaId uniqueidentifier
TABLE: Values //The DocumentId+ValueXPath function as a PK
COLUMN NAME DATA TYPE
DocumentId uniqueidentifier
ValueXPath varchar(250)
Value text
from these tables, when performing queries I would do a series of self-joins on the value table. When I want to get the entire object by the DocumentId, I would have a generic script for creating a view mimics a denormalized datatable of the complex-type.
What I want to know
I believe there are better ways to accomplish what I am trying to, but I am a little too ignorant about the relative performance benefits of different SQL techniques. Specifically I don't know the performance cost of:
1 - comparing the value of a text field versus of a varchar field.
2 - different kind of joins versus nested queries
3 - getting a view versus an xml document from the sql db
4 - doing some other things that I don't even know I don't know would be affecting my query but, I am experienced enough to know exist
I would appreciate any information or resources about these performance issues in sql as well as a recommendation for how to approach this general issue in a more efficient way.
For Example,
Here's an example of what I am currently planning on doing.
I have a C# class Address which looks like
public class Address{
string Line1 {get;set;}
string Line2 {get;set;}
string City {get;set;}
string State {get;set;}
string Zip {get;set;
}
An instance is constructed from new Address{Line1="17 Mulberry Street", Line2="Apt C", City="New York", State="NY", Zip="10001"}
its XML value would be look like.
<Address>
<Line1>17 Mulberry Street</Line1>
<Line2>Apt C</Line2>
<City>New York</City>
<State>NY</State>
<Zip>10001</Zip>
</Address>
Using the db-schema from above I would have a single record in the Schemas table with an XSD definition of the address xml schema. This instance would have a uniqueidentifier (PK of the Documents table) which is assigned to the SchemaId of the Address record in the Schemas table. There would then be five records in the Values table to represent this Address.
They would look like:
DocumentId ValueXPath Value
82415E8A-8D95-4bb3-9E5C-AA4365850C70 /Address/Line1 17 Mulberry Street
82415E8A-8D95-4bb3-9E5C-AA4365850C70 /Address/Line2 Apt C
82415E8A-8D95-4bb3-9E5C-AA4365850C70 /Address/City New York
82415E8A-8D95-4bb3-9E5C-AA4365850C70 /Address/State NY
82415E8A-8D95-4bb3-9E5C-AA4365850C70 /Address/Zip 10001
Just Added a Bounty...
My objective is to obtain the resources I need in order to give my application a data access layer that is fully searchable and has a data-schema generated from the application layer that does not require direct database configuration (i.e. creating a new SQL table) in order to add a new aggregate root to the domain model.
I am open to the possibility of using .NET compatible technologies other than SQL, but I will require that any such suggestions be adequately substantiated in order to be considered.
How about looking for a solution at the architectural level? I was also breaking my head on complex graphs and performance until I discovered CQRS.
[start evangelist mode]
You can go document-based or relational as storage. Even both! (Event Sourcing)
Nice separation of concerns: Read Model vs Write Model
Have your cake and eat it too!
Ok, there is an initial learning / technical curve to get over ;)
[end evangelist mode]
As you stated: "I need to be able to create variable schemas on the fly without changing anything about the database access layer." The key benefit is that your read model can be very fast since it's made for reading. If you add Event Sourcing to the mix, you can drop and rebuild your Read Model to whatever schema you want... even "online".
There are some nice opensource frameworks out there like nServiceBus which saves lots of time and technical challenges. All depends on how far you want to take these concepts what you're willing/can spend time on. You can even start with just basics if you follow Greg Young's approach. See the info in the links below.
See
CQRS Examples and Screencasts
CQRS Questions
Intro (Also see the video)
Somehow what you want sounds like a painful thing to do in SQL. Basically, you should treat the inside of a text field as opaque as when querying an SQL database. Text fields were not made for efficient queries.
If you just want to store serialized objects in a text field, that is fine. But do not try to build queries that look inside the text field to find objects.
Your idea sounds like you want to perform some joins, XML parsing, and XPath application to get to a value. This doesn't strike me as the most efficient thing to do.
So, my advise:
Either just store serialized objects in the db, and do nothing more than load them and perform all other operations in memory
Or, if you need to query complex data structures, you may really want to look into document stores/databases like CouchDB or MongoDB; you can also check Wikipedia on the subject. There are even databases specifically designed for storing XML, even though I personally don't like them very much.
Addendum, per your explanations above
Simply put, don't go over the top with this thing:
If you just want to persist C#/.NET objects, just use the XML Serialization already built into the framework, a single table and be done with it.
If you, for some reason, need to store complex XML, use a dedicated XML store
If you have a fixed database schema, but it is too complex for efficient queries, use a Document Store in memory where you keep a denormalized version of your data for faster queries (or just simplify your database schema)
If you don't really need a fixed schema, use just a Document Store, and forget about having any "schema definition" at all
As for your solution, yes, it could work somehow. As could a plain SQL schema if you set it up right. But for applying an XPath, you'll probably parse the whole XML document each time you access a record, which wouldn't be very efficient to begin with.
If you want to check out Document databases, there are .NET drivers for CouchDB and MongoDB. The eXist XML database offers a number of Web protocols, and you can probably create a client class easily with VisualStudio's point-and-shoot interface. Or just google for someone who already did.
I need to be able to create variable
schemas on the fly without changing
anything about the database access
layer.
You are re-implementing the RDBMS within an RDBMS. The DB can do this already - that is what the DDL statements like create table and create schema are for....
I suggest you look into "schemas" and SQL security. There is no reason with the correct security setup you cannot allow your users to create their own tables to store document attributes in, or even generate them automatically.
Edit:
Slightly longer answer, if you don't have full requirements immediately, I would store the data as XML data type, and query them using XPath queries. This will be OK for occasional queries over smallish numbers of rows (fewer than a few thousand, certainly).
Also, your RDBMS may support indexes over XML, which may be another way of solving your problem. CREATE XML INDEX in SqlServer 2008 for example.
However for frequent queries, you can use triggers or materialized views to create copies of relevant data in table format, so more intensive reports can be speeded up by querying the breakout tables.
I don't know your requirements, but if you are responsible for creating the reports/queries yourself, this may be an approach to use. If you need to enable users to create their own reports that's a bigger mountain to climb.
I guess what i am saying is "are you sure you need to do this and XML can't just do the job".
In part, it will depend of your DB Engine. You're using SQL Server, don't you?
Answering your topics:
1 - Comparing the value of a text field versus of a varchar field: if you're comparing two db fields, varchar fields are smarter. Nvarchar(max) stores data in unicode with 2*l+2 bytes, where "l" is the lengh. For performance issues, you will need consider how much larger tables will be, for selecting the best way to index (or not) your table fields. See the topic.
2 - Sometimes nested queries are easily created and executed, also serving as a way to reduce query time. But, depending of the complexity, would be better to use different kind of joins. The best way is try to do in both ways. Execute two or more times each query, for the DB engine "compiles" a query on first executing, then the subsequent are quite faster. Measure the times for different parameters and choose the best option.
"Sometimes you can rewrite a subquery to use JOIN and achieve better performance. The advantage of creating a JOIN is that you can evaluate tables in a different order from that defined by the query. The advantage of using a subquery is that it is frequently not necessary to scan all rows from the subquery to evaluate the subquery expression. For example, an EXISTS subquery can return TRUE upon seeing the first qualifying row." - link
3- There's no much information in this question, but if you will get the xml document directly from the table, would be a good idea insted a view. Again, it will depends of the view and the document.
4- Other issues is about the total records expected for your table; the indexing of the columns, in wich you need to consider sorting, joining, filtering, PK's and FK's. Each situation could demmand different aproaches. My sugestion is to invest some time reading about your database engine and queries functioning and relating to your system.
I hope I've helped.
Interesting question.
I think you may be asking the wrong question here. Broadly speaking, as long as you have a FULLTEXT index on your text field, queries will be fast. Much faster than varchar if you have to use wild cards, for instance.
However, if I were you, I'd concentrate on the actual queries you're going to be running. Do you need boolean operators? Wildcards? Numerical comparisons? That's where I think you will encounter the real performance worries.
I would imagine you would need queries like:
"find all addresses in the states of New York, New Jersey and Pennsylvania"
"find all addresses between house numbers 1 and 100 on Mulberry Street"
"find all addresses where the zipcode is missing, and the city is New York"
At a high level, the solution you propose is to store your XML somewhere, and then de-normalize that XML into name/value pairs for querying.
Name/value pairs have a long and proud history, but become unwieldy in complex query situations, because you're not using the built-in optimizations and concepts of the relational database model.
Some refinements I'd recommend is to look at the domain model, and at least see if you can factor out separate data types into the "value" column; you might end up with "textValue", "moneyValue", "integerValue" and "dateValue". In the example you give, you might factor "address 1" into "housenumber" (as an integer) and "streetname".
Having said all this - I don't think there's a better solution other than completely changing tack to a document-focused database.
I was wondering if you have a website with a dozen different types of listings (Shops, Restaurants, Clubs, Hotels, Events) that require different fields, is there a benefit of creating a table with columns defined like so
Example Shop:
shop_id | name | X | Y | city | district | area | metro | station | address | phone | email | website | opening_hours
Or a more abstract approach similar to this:
object_id | name
---------------
1 | Messy Joe's
2 | Bate's Motel
type_id | name
---------------
1 | hotel
2 | restaurant
object_id | type_id
---------------
1 | 2
2 | 1
field_id | name | field_type
---------------
1 | address | text
2 | opening_hours | date
3 | speciality | text
type_id | field_id
---------------
1 | 1
1 | 2
2 | 1
2 | 3
object_id | field_id | value
1 | 1 | 1st street....
1 | 3 | English Cuisine
Of course it can be more abstract if value's are predefined (Example: specialties could have their own list)
If I take the abstract approach it can be very flexible, but queries will be more complex with a lot of joins.
But I don't know if this affects the performance, executing these 'more complex' queries.
I would be interested to know what are the up and downsides of both methods. I can just imagine for myself, but I don't have the experience to confirm this.
Certain issues need to be clarified and resolved before we can enter into a reasonable discussion.
Pre-requisite Resolution
Labels
In a profession that demands precision, it is important that we use precise labels, to avoid confusion, and so that we can communicate without having to use long-winded descriptions and qualifiers.
What you have posted as FixedTables, is Unnormalised. Fair enough, it may be an attempt at Third Normal form, but in fact it is a flat file, Unnormalised (not "denormalised). What you have posted as AbstractTables is, to be precise, Entity-Attribute-Value, which is almost, but not quite, Sixth Normal form, and is therefore more Normalised than 3NF. Assuming it is done correctly, of course.
The Unnormalised flat file is not "denormalised". It is chock full of duplication (nothing has been done to remove repeating groups and duplicate columns or to resolve dependencies) and Nulls, it is a performance hog in many ways, and prevents concurrency.
In order to be Denormalised, it has to first be Normalised, and then the Normalisation backed off a little for some good reason. Since it is not Normalised in the first place, it cannot be Denormalised. It is simply Unnormalised.
It cannot be said to be denormalised "for performance", because being a performance hog, it is the very antithesis of performance. Well, they need a justification for the lack of formalised design], and "for performance" is it. Even the smallest formal scrutiny exposed the misrepresentation (but very few people can provide, so it remains hidden, until they get an outsider to address, you guessed it, the massive performance problem).
Normalised structures perform far better than Unnormalised structures. More normalised structures (EAV/6NF) perform better than less normalised structures (3NF/5NF).
I am agreeing with the thrust of OMG Ponies, but not their labels and definitions
rather than saying 'don't "denormalise" unless you have to', I am saying, 'Normalise faithfully, period' and 'if there is a performance problem, you have not Normalised correctly'.
Wikipedia
The entries for Normal Forms and Normalisation offer definitions that are incorrect; they confuse the Normal Forms; they are lacking regarding the process of Normalisation; and they give equal weight to absurd or questionable NFs which have been debunked long ago. The result is, Wikipedia adds to an already confused and rarely understood subject. So don't waste your time.
However, in order to progress, without that reference posing a hindrance, let me say this.
The definition of 3NF is stable, and has not changed.
There is a lot of confusion of the NFs between 3NF and 5NF. The truth is that this is an area that progressed over the last 15 years; and many orgs, academics as well as vendors with their products with limitations, jumped to create a new "Normal Form" to validate their offerings. All serving commercial interests and academically unsound. 3NF in its original untampered state intended and guaranteed certain attributes.
The sum total is, 5NF is today, what 3NF was intended to be 15 years ago, and you can skip the commercial banter and the twelve or so "special" (commercial and pseudo-academic) NFs in-between, some of which are identified in Wikipedia, and even that in confusing terms.
Fifth Normal Form
Since you have been able to understand and implement the EAV in your post, you will have no problem understanding the following. Of course a true Relational Model is pre-requisite, strong keys, etc. Fifth Normal Form is, since we are skipping the Fourth:
Third Normal Form
which in simple definitive terms is, every non-key column in every table has a 1::1 relationship to the Primary Key of the table,
and to no other non-key columns
Zero data duplication (the result, if Normalisation is progressed diligently; not achieved by intelligence or experience alone, or by working toward it as a goal without the formal process)
no Update Anomalies (when you update a column somewhere, you do not have to update the same column located somewhere else; the column exists in one and only one place).
If you understand the above, 4NF, BCNF, and all the silly "NFs" can be dismissed, they are required for physicalised Record Filing Systems, as promoted by academics, quite foreign to the Relational Model (Codd).
Sixth Normal Form
The purpose is elimination of missing data (attribute columns), aka elimination of Nulls
This is the one true solution to the Null Problem (also called Handling Missing Values), and the result is a database without Nulls. (It can be done at 5NF with standards and Null substitutes but that is not optimal.) How you interpret and display the missing values is another story.
Technically, is not a true Normal Form, because it does not have 5NF as a pre-requisite, but it has a value
EAV vs Sixth Normal Form
All the databases I have written, except one, are pure 5NF. I have worked with (administered, fixed up, enhanced) a couple of EAV databases, and I have implemented many true 6NF databases. EAV is a loose implementation of 6NF, often done by people who do not have a good grasp on Normalisation and the NFs, but who can see the value in, and need the flexibility of, EAV. You are a perfect example.
The difference is this: because it is loose, and because implementers do not have a reference (6NF) to be faithful to, they only implement what they need, and they write it all in code; that ends up being an inconsistent model.
Whereas, a pure 6NF implementation does have a pure academic reference point, and thus it is usually tighter, and consistent. Typically this shows up in two visible elements:
6NF has a catalogue to contain metadata, and everything is defined in metadata, not code. EAV does not have one, everything is in code (implementers keep track of the objects and attributes). Obviously a catalogue eases the addition of columns, navigation, and allows utilities to be formed.
6NF when understood, provides the true solution to The Null Problem. EAV implementers, since they are absent the 6NF context, handle missing data in code, inconsistently, or worse, allow Nulls in the database. 6NF implementers disallow Nulls, and handle missing Data consistently and elegantly, without requiring code constructs (for Null handling; you still have to code for missing data of course).
Eg. For 6NF databases with a catalogue, I have a set of procs that will [re]generate the SQL required to perform all SELECTs, and I provide Views in 5NF for all users, so they do not need to know or understand the underlying 6NF structure. They are driven off the catalogue. Thus changes are easy and automated. EAV types do that manually, due to the absence of the catalogue.
Discussion
Now, we can start the discussion.
"Of course it can be more abstract if
value's are predefined (Example:
specialities could have their own
list)"
Sure. But do not get too "abstract". Maintain consistency and implement such lists in the same EAV (or 6NF) manner as other lists.
"If I take the abstract approach it
can be very flexible, but queries will
be more complex with a lot of joins.
But I don't know if this affects the
performance, executing these 'more
complex' queries."
Joins are pedestrian in relational databases. The problem is not the database, the problem is that SQL is cumbersome when handling joins, especially compound keys.
EAV and 6NF databases have more Joins, which just as pedestrian, no more, no less. If you have to code each SELECT manually, sure, the cumbersome gets really cumbersome.
The entire problem can be eliminated by (a) going with 6NF over EAV and (b) implementing a catalogue, from which you can (c) generate all the basic SQL. Eliminates an entire class of errors as well.
It is a common myth that Joins somehow have a cost. Totally false.
The join is implemented at compile time, there is nothing of substance to 'cost' CPU cycles.
The issue is the size of tables being joined, not the cost of the Join between those same tables.
Joining two tables with millions of rows each, on a correct PK⇢FK relation, each of which have the appropriate indices
(Unique on the parent [PK] side; Unique on the Child side [PK=parent FK + something]
is instantaneous
Where the Child index is not unique, but at least the leading columns are valid, it is slower; where there is no useful index, of course it is very slow.
None of it has to do with Join cost.
Where many rows are returned, the bottleneck will be the network and the disk layout; not the join processing.
Therefore you can get as "complex" as you like, there is no cost, SQL can handle it.
I would be interested to know what are
the up and downsides of both methods.
I can just imagine for myself, but I
don't have the experience to confirm
this.
5NF (or 3NF for those who have not made the progression) is the easiest and best, in terms of implementation; ease of use (developers as well as users); and maintenance.
The drawback is, every time you add a column, you have to change the database structure (table DDL). That is fine is some cases, but not in most cases, due to change control in place, quite onerous.
Second, you have to change existing code (code handling the new column does not count, because that is an imperative): where good standards are implemented, that is minimised; where they are absent, the scope is unpredictable.
EAV (which is what you have posted), allows columns to be added without DDL changes. That is the single reason people choose it. (code handling the new column does not count, because that is an imperative). If implemented well, it will not affect existing code; if not, it will.
But you need EAV-capable developers.
When EAV is implemented badly, it is abominable, a worse mess than 5NF done badly, but not any worse than Unnormalised which is what most databases out there are (misrepresented as "denormalised for performance").
Of course, it is even more important (than in 5NF/3NF) to hold a strong Transaction context, because the columns are far more distributed.
Likewise, it is essential to retain Declarative Referential Integrity: the messes I have seen were due in large part to the developers removing DRI because it became "too hard to maintain", the result was, as you can imagine, one mother of a data heap with duplicate 3NF/5NF rows and columns all over the place. And inconsistent Null handling.
There is no difference in performance, assuming that the server has been reasonably configured for the intended purpose. (Ok, there are specific optimisations that are possible only in 6NF, which are not possible in other NFs, but I think that is outside the scope of this thread.) And again, EAV done badly can cause unnecessary bottlenecks, no more so than Unnormalised.
Of course, if you go with EAV, I am recommending more formality; buy the full quid; go with 6NF; implement a catalogue; utilities to produce SQL; Views; handle Missing Data consistently; eliminate Nulls altogether. This reduces your vulnerability to the quality of your developers; they can forget about the EAV/6NF esoteric issues, use Views, and concentrate on the app logic.
In your question, you have presented at least two major issues at the same time. Those two issues are E-A-V and gen-spec.
First, let's talk about E-A-V. Your last table (object_id, field_id, value) is essentially an E-A-V. There is an upside to E-A-V and a downside to E-A-V. The upside is that the structure is so generic that it can accomodate almost any body of data describing almost any subject matter. That means that you can proceed to design and implementation with no data analysis and no understanding of the subject matter, and not worry about wrong assumptions. The down side is that at retrieval time, you have to do the data analysis that you skipped over before building the data base, in order to come up with queries that mean anything. This is much more serious than just retrieval efficiency. But you are also going to have terrible problems with retrieval efficiency. There are only two ways to learn about this pitfall: live through it or read about it from those who have. I recommend the reading.
Second, you have a gen-spec case. Your table (object_id, type_id) captures a gen-spec (generalization-specialization) pattern, along with the related tables. If I had to generalize between hotels and restaurants, I might call it something like "public accomodations" or "venues". But I'm not sure I understand your case, and you may be driving for something even more general than those two names suggest. After all, you've included "events" in your list, and an event is not a type of venue in my mind.
I've referred other people to readings on gen-spec and the relational model in previous responses.
When two tables are very similar, when should they be combined?
But I hesitate to send you off in the same direction, because it's not clear to me that you want to come up with a relational model of the data before building your database. A relational model of a body of data and an E-A-V model of the same data are almost totally at odds with each other. It seems to me you have to make that choice before you even explore how to express gen-spec in the relational model of data.
When you start to require a large number of different entities (or even before...), a nosql solution would be vastly simpler than either choice.
Just store each entity/record with the exact fields you require.
{
"id": 1,
"type":"Restaurant",
"name":"Messy Joe",
"address":"1 Main St.",
"tags":["asian","fusion","casual"]
}
The "abstract" approach is better known as "Normalization", looks like 3rd Normal Form (3NF).
The other one is called "Denormalized", and can be a valid performance option... when you've encountered speed issues using the Normalized approach, not before.
How do you have the listings represented in code? I'd guess Listing as a supertype, with Shop, Restuarant, etc. as subtypes?
Assuming so, this is a case of how to map subtypes to a relational database. There are generally three choices:
Option 1: single table per subtype,
with common attributes repeated in
each table (name, id, etc).
Option 2: single table for all objects (your single table approach)
Option 3: table for the supertype and one for each subtype
There's no universally correct solution. My preference is generally to start with option 3; it provides an intituitive structure to work with, is pretty well normalised and can easily be extended. It means a single join for retrieving each instance - but RDBMS are well optimised for doing joins so it doesn't really cause performance problems in practice.
Option 2 can be more performant for queries (no joins) but causes problems if other tables need to refer to all supertype instances (proliferation of foreign keys).
Option 1 appears at first sight to be the most performant, although 2 caveats: (1) It's not resilient to change. If you add a new subtype (and so different attributes) you'll need to change the table structure and migrate it. (2) It can be less efficient than it seems. Because the table population is sparse, some DBs don't store it particularly efficiently. As a consequence it can be less efficicent than option 1 - since the query engine can do joins faster than it can search bloated sparse table spaces.
Which to choose really comes down to knowing details of your problem. I'd suggest reading up a bit on the options: this article is a good place to start.
hth
My employer, a small office supply company, is switching suppliers and I am looking through their electronic content to come up with a robust database schema; our previous schema was pretty much just thrown together without any thought at all, and it's pretty much led to an unbearable data model with corrupt, inconsistent information.
The new supplier's data is much better than the old one's, but their data is what I would call hypernormalized. For example, their product category structure has 5 levels: Master Department, Department, Class, Subclass, Product Block. In addition the product block content has the long description, search terms and image names for products (the idea is that a product block contains a product and all variations - e.g. a particular pen might come in black, blue or red ink; all of these items are essentially the same thing, so they apply to a single product block). In the data I've been given, this is expressed as the products table (I say "table" but it's a flat file with the data) having a reference to the product block's unique ID.
I am trying to come up with a robust schema to accommodate the data I'm provided with, since I'll need to load it relatively soon, and the data they've given me doesn't seem to match the type of data they provide for demonstration on their sample website (http://www.iteminfo.com). In any event, I'm not looking to reuse their presentation structure so it's a moot point, but I was browsing the site to get some ideas of how to structure things.
What I'm unsure of is whether or not I should keep the data in this format, or for example consolidate Master/Department/Class/Subclass into a single "Categories" table, using a self-referencing relationship, and link that to a product block (product block should be kept separate as it's not a "category" as such, but a group of related products for a given category). Currently, the product blocks table references the subclass table, so this would change to "category_id" if I consolidate them together.
I am probably going to be creating an e-commerce storefront making use of this data with Ruby on Rails (or that's my plan, at any rate) so I'm trying to avoid getting snagged later on or having a bloated application - maybe I'm giving it too much thought but I'd rather be safe than sorry; our previous data was a real mess and cost the company tens of thousands of dollars in lost sales due to inconsistent and inaccurate data. Also I am going to break from the Rails conventions a little by making sure that my database is robust and enforces constraints (I plan on doing it at the application level, too), so that's something I need to consider as well.
How would you tackle a situation like this? Keep in mind that I have the data to be loaded already in flat files that mimic a table structure (I have documentation saying which columns are which and what references are set up); I'm trying to decide if I should keep them as normalized as they currently are, or if I should look to consolidate; I need to be aware of how each method will affect the way I program the site using Rails since if I do consolidate, there will be essentially 4 "levels" of categories in a single table, but that definitely seems more manageable than separate tables for each level, since apart from Subclass (which directly links to product blocks) they don't do anything except show the next level of category under them. I'm always a loss for the "best" way to handle data like this - I know of the saying "Normalize until it hurts, then denormalize until it works" but I've never really had to implement it until now.
I would prefer the "hypernormalized" approach over a denormal data model. The self referencing table you mentioned might reduce the number of tables down and simplify life in some ways, but in general this type of relationship can be tricky to deal with. Hierarchical queries become a pain, as does mapping an object model to this (if you decide to go that route).
A couple of extra joins is not going to hurt and will keep the application more maintainable. Unless performance degrades due to the excessive number of joins, I would opt to leave things like they are. As an added bonus if any of these levels of tables needed additional functionality added, you will not run into issues because you merged them all into the self referencing table.
I totally disagree with the criticisms about self-referencing table structures for parent-child hierarchies. The linked list structure makes UI and business layer programming easier and more maintainable in most cases, since linked lists and trees are the natural way to represent this data in languages that the UI and business layers would typically be implemented in.
The criticism about the difficulty of maintaining data integrity constraints on these structures is perfectly valid, though the simple solution is to use a closure table that hosts the harder check constraints. The closure table is easily maintained with triggers.
The tradeoff is a little extra complexity in the DB (closure table and triggers) for a lot less complexity in UI and business layer code.
If I understand correctly, you want to take their separate tables and turn them into a hierarchy that's kept in a single table with a self-referencing FK.
This is generally a more flexible approach (for example, if you want to add a fifth level), BUT SQL and relational data models don't tend to work well with linked lists like this, even with new syntax like MS SQL Servers CTEs. Admittedly, CTEs make it much better though.
It can be difficult and costly to enforce things, like that a product must always be on the fourth level of the hierarchy, etc.
If you do decide to do it this way, then definitely check out Joe Celko's SQL for Smarties, which I believe has a section or two on modeling and working with hierarchies in SQL or better yet get his book that is devoted to the subject (Joe Celko's Trees and Hierarchies in SQL for Smarties).
Normalization implies data integrity, that is: each normal form reduces the number of situations where you data is inconsistent.
As a rule, denormalization has a goal of faster querying, but leads to increased space, increased DML time, and, last but not least, increased efforts to make data consistent.
One usually writes code faster (writes faster, not the code faster) and the code is less prone to errors if the data is normalized.
Self referencing tables almost always turn out to be much worse to query and perform worse than normalized tables. Don't do it. It may look to you to be more elegant, but it is not and is a very poor database design technique. Personally the structure you described sounds just fine to me not hypernormalized. A properly normalized database (with foreign key constraints as well as default values, triggers (if needed for complex rules) and data validation constraints) is also far likelier to have consistent and accurate data. I agree about having the database enforce the rules, likely this is part of why the last application had bad data because the rules were not enforced in the proper place and people were able to easily get around them. Not that the application shouldn't check as well (no point even sending an invalid date for instance for the datbase to fail on insert). Since youa redesigning, I would put more time and effort into designing the necessary constraints and choosing the correct data types (do not store dates as string data for instance), than in trying to make the perfectly ordinary normalized structure look more elegant.
I would bring it in as close to their model as possible (and if at all possible, I would get files which match their schema - not a flattened version). If you bring the data directly into your model, what happens if data they send starts to break assumptions in the transformation to your internal application's model?
Better to bring their data in, run sanity checks and check that assumptions are not violated. Then if you do have an application-specific model, transform it into that for optimal use by your application.
Don't denormalize. Trying to acheive a good schema design by denormalizing is like trying to get to San Francisco by driving away from New York. It doesn't tell you which way to go.
In your situation, you want to figure out what a normalized schema would like. You can base that largely on the source schema, but you need to learn what the functional dependencies (FD) in the data are. Neither the source schema nor the flattened files are guaranteed to reveal all the FDs to you.
Once you know what a normalized schema would look like, you now need to figure out how to design a schema that meets your needs. It that schema is somewhat less than fully normalized, so be it. But be prepared for difficulties in programming the transformation between the data in the flattened files and the data in your desgined schema.
You said that previous schemas at your company cost millions due to inconsistency and inaccuracy. The more normalized your schema is, the more protected you are from internal inconsistency. This leaves you free to be more vigilant about inaccuracy. Consistent data that's consistently wrong can be as misleading as inconsistent data.
is your storefront (or whatever it is you're building, not quite clear on that) always going to be using data from this supplier? might you ever change suppliers or add additional different suppliers?
if so, design a general schema that meets your needs, and map the vendor data to it. Personally I'd rather suffer the (incredibly minor) 'pain' of a self-referencing Category (hierarchical) table than maintain four (apparently semi-useless) levels of Category variants and then next year find out they've added a 5th, or introduced a product line with only three...
For me, the real question is: what fits the model better?
It's like comparing a Tuple and a List.
Tuples are a fixed size and are heterogeneous -- they are "hypernormalized".
Lists are an arbitrarty size and are homogeneous.
I use a Tuple when I need a Tuple and a List when I need a list; they fundamentally server different purposes.
In this case, since the product structure is already well defined (and I assume not likely to change) then I would stick with the "Tuple approach". The real power/use of a List (or recursive table pattern) is when you need it to expand to an arbitrary depth, such as for a BOM or a genealogy tree.
I use both approaches in some of my database depending upon the need. However, there is also the "hidden cost" of a recursive pattern which is that not all ORMs (not sure about AR) support it well. Many modern DBs have support for "join-throughs" (Oracle), hierarchy IDs (SQL Server) or other recursive patterns. Another approach is to use a set-based hierarchy (which generally relies on triggers/maintenance). In any case, if the ORM used does not support recursive queries well, then there may be the extra "cost" of using the to the DB features directly -- either in terms of manual query/view generation or management such as triggers. If you don't use a funky ORM, or simply use a logic separator such as iBatis, then this issue may not even apply.
As far as performance, on new Oracle or SQL Server (and likely others) RDBMS, it ought to be very comparable so that would be the least of my worries: but check out the solutions available for your RDBMS and portability concerns.
Everybody who recommends you not to have a hierarchy introduced in the database, considering just the option of having a self-referenced table. This is not the only way to model the hierarchy in the database.
You may use a different approach, that provides you with easier and faster querying without using recursive queries.
Let's say you have a big set of nodes (categories) in your hierarchy:
Set1 = (Node1 Node2 Node3...)
Any node in this set can also be another set by itself, that contains other nodes or nested sets:
Node1=(Node2 Node3=(Node4 Node5=(Node6) Node7))
Now, how we can model that? Let's have each node to have two attributes, that set the boundaries of the nodes it contains:
Node = { Id: int, Min: int, Max: int }
To model our hierarchy, we just assign those min/max values accordingly:
Node1 = { Id = 1, Min = 1, Max = 10 }
Node2 = { Id = 2, Min = 2, Max = 2 }
Node3 = { Id = 3, Min = 3, Max = 9 }
Node4 = { Id = 4, Min = 4, Max = 4 }
Node5 = { Id = 5, Min = 5, Max = 7 }
Node6 = { Id = 6, Min = 6, Max = 6 }
Node7 = { Id = 7, Min = 8, Max = 8 }
Now, to query all nodes under the Set/Node5:
select n.* from Nodes as n, Nodes as s
where s.Id = 5 and s.Min < n.Min and n.Max < s.Max
The only resource-consuming operation would be if you want to insert a new node, or move some node within the hierarchy, as many records will be affected, but this is fine, as the hierarchy itself does not change very often.