When I am creating a new database table, what factors should I take into account for selecting the primary key's data type?
Sorry to do that, but I found that the answers I gave to related questions (you can check this and this) could apply to this one. I reshaped them a little bit...
You will find many posts dealing with this issue, and each choice you'll make has its pros and cons. Arguments for these usually refer to relational database theory and database performance.
On this subject, my point is very simple: surrogate primary keys ALWAYS work, while Natural keys MIGHT NOT ALWAYS work one of these days, and this for multiple reasons: field too short, rules change, etc.
To this point, you've guessed here that I am basically a member of the uniqueIdentifier/surrogate primary key team, and even if I appreciate and understand arguments such as the ones presented here, I am still looking for the case where "natural" key is better than surrogate ...
In addition to this, one of the most important but always forgotten arguments in favor of this basic rule is related to code normalization and productivity:
each time I create a table, shall I lose time
identifying its primary key and its physical characteristics (type, size)
remembering these characteristics each time I want to refer to it in my code?
explaining my PK choice to other developers in the team?
My answer is no to all of these questions:
I have no time to lose trying to identify "the best Natural Primary Key" when the surrogate option gives me a bullet-proof solution.
I do not want to remember that the Primary Key of my Table_whatever is a 10 characters long string when I write the code.
I don't want to lose my time negotiating the Natural Key length: "well if You need 10 why don't you take 12 to be on the safe side?". This "on the safe side" argument really annoys me: If you want to stay on the safe side, it means that you are really not far from the unsafe side! Choose surrogate: it's bullet-proof!
So I've been working for the last five years with a very basic rule: each table (let's call it 'myTable') has its first field called 'id_MyTable' which is of uniqueIdentifier type. Even if this table supports a "many-to-many" relation, where a field combination offers a very acceptable Primary Key, I prefer to create this 'id_myManyToManyTable' field being a uniqueIdentifier, just to stick to the rule, and because, finally, it does not hurt.
The major advantage is that you don't have to care anymore about the use of Primary Key and/or Foreign Key within your code. Once you have the table name, you know the PK name and type. Once you know which links are implemented in your data model, you'll know the name of available foreign keys in the table.
And if you still want to have your "Natural Key" somewhere in your table, I advise you to build it following a standard model such as
Tbl_whatever
id_whatever, unique identifier, primary key
code_whatever, whateverTypeYouWant(whateverLengthYouEstimateTheRightOne), indexed
.....
Where id_ is the prefix for primary key, and code_ is used for "natural" indexed field. Some would argue that the code_ field should be set as unique. This is true, and it can be easily managed either through DDL or external code. Note that many "natural" keys are calculated (invoice numbers), so they are already generated through code
I am not sure that my rule is the best one. But it is a very efficient one! If everyone was applying it, we would for example avoid time lost answering to this kind of question!
If using a numeric key, make sure the datatype is giong to be large enough to hold the number of rows you might expect the table to grow to.
If using a guid, does the extra space needed to store the guid need to be considered? Will coding against guid PKs be a pain for developers or users of the application.
If using composite keys, are you sure that the combined columns will always be unique?
I don't really like what they teach in school, that is using a 'natural key' (for example ISBN on a bookdatabase) or even having a primary key made up off 2 or more fields. I would never do that. So here's my little advice:
Always have one dedicated column in every table for your primary key.
They all should have the same colomn name across all tables, i.e. "ID" or "GUID"
Use GUIDs when you can (if you don't need performance), otherwise incrementing INTs
EDIT:
Okay, I think I need to explain my choices a little bit.
Having a dedicated column namend the same across all table for you primary key, just makes your SQL-Statements a lot of easier to construct and easier for someone else (who might not be familiar with your database layout) easier to understand. Especially when you're doing lots of JOINS and things like that. You won't need to look up what's the primary key for a specific table, you already know, because it's the same everywhere.
GUIDs vs. INTs doesn't really matters that much most of the time. Unless you hit the performance cap of GUIDs or doing database merges, you won't have major issues with one or another. BUT there's a reason I prefer GUIDs. The global uniqueness of GUIDs might always come in handy some day. Maybe you don't see a need for it now, but things like, synchronizing parts of the database to a laptop / cell phone or even finding datarecords without needing to know which table they're in, are great examples of the advantages GUIDs can provide. An Integer only identifies a record within the context of one table, whereas a GUID identifies a record everywhere.
In most cases I use an identity int primary key, unless the scenario requires a lot of replication, in which case I may opt for a GUID.
I (almost) never used meaningful keys.
Unless you have an ultra-convenient natural key available, always use a synthetic (a.k.a. surrogate) key of a numeric type. Even if you do have a natural key available, you might want to consider using a synthetic key anyway and placing an additional unique index on your natural key. Consider what happened to higher-ed databases that used social security numbers as PKs when federal law changed, the costs of changing over to synthetic keys were enormous.
Also, I have to disagree with the practice of naming every primary key the same, e.g. "id". This makes queries harder to understand, not easier. Primary keys should be named after the table. For example employee.employee_id, affiliate.affiliate_id, user.user_id, and so on.
Do not use a floating point numeric type, since floating point numbers cannot be properly compared for equality.
Where do you generate it? Incrementing number's don't fit well for keys generated by the client.
Do you want a data-dependent or independent key (sometimes you could use an ID from business data, can't say if this is always useful or not)?
How well can this type be indexed by your DB?
I have used uniqueidentifiers (GUIDs) or incrementing integers so far.
Cheers
Matthias
Numbers that have meaning in the real world are usually a bad idea, because every so often the real world changes the rules about how those numbers are used, in particular to allow duplicates, and then you've got a real mess on your hands.
I'm partial to using an generated integer key. If you expect the database to grow very large, you can go with bigint.
Some people like to use guids. The pro there is that you can merge multiple instances of the database without altering any keys but the con is that performance can be affected.
For a "natural" key, whatever datatype suits the column(s). Artifical (surrogate) keys are usually integers.
It all depends.
a) Are you fine having unique sequential numeric numbers as your primary key? If yes, then selecting UniqueIdentifier as your primary key will suffice.
b) If your business demand is such that you need to have alpha numeric primary key, then you got to go for varchar or nvarchar.
These are the two options I could think of.
A great factor is how much data you're going to store. I work for a web analytics company, and we have LOADS of data. So a GUID primary key on our pageviews table would kill us, due to the size.
A rule of thumb: For high performance, you should be able to store your entire index in memory. Guids could easily break this!
Use natural keys when they can be trusted. Some sources of natural keys can't be trusted. Years ago, the Social Security Administration used to occasionally mess up an assign the same SSN to two different people. Theyv'e probably fixed that by now.
You can probably trust VINs for vehicles, and ISBNs for books (but not for pamphlets, which may not have an ISBN).
If you use natural keys, the natural key will determine the datatype.
If you can't trust any natural keys, create a synthetic key. I prefer integers for this purpose. Leave enough room for reasonable expansion.
I usually go with a GUID column primary key for all tables (rowguid in mssql). What could be natural keys I make unique constraints. A typical example would be a produkt identification number that the user have to make up and ensure that is unique. If I need a sequence, like in a invoice i build a table to keep a lastnumber and a stored procedure to ensure serialized access. Or a Sequence in Oracle :-) I hate the "social security number" sample for natural keys as that number will never be alway awailable in a registration process. Resulting in a need for a scheme to generate dummy numbers.
I usually always use an integer, but here's an interesting perspective.
https://blog.codinghorror.com/primary-keys-ids-versus-guids/
Whenever possible, try to use a primary key that is a natural key. For instance, if I had a table where I logged one record every day, the logdate would be a good primary key. Otherwise, if there is no natural key, just use int. If you think you will use more than 2 billion rows, use a bigint. Some people like to use GUIDs, which works well, as they are unique, and you will never run out of space. However, they are needlessly long, and hard to type in if you are just doing adhoc queries.
Related
I want to know if I can use human readable primary keys for a relatively small number of database objects, which will describe large metropolitan areas.
For example, using "washington_dc" as the pk for the Washington, DC metro area, or "nyc" for the New York City one.
Tons of objects will be foreign keyed to these metro area objects, and I'd like to be able to tell where a person or business is located just by looking at their database record.
I'm just worried because my gut tells me this might be a serious crime against good practices.
So, am I "allowed" to do this kind of thing?
Thanks!
It all depends on the application - natural primary keys make a good deal of sense on the surface, since they are human readable and don't require any joins when displaying data to end users.
However, natural primary keys tend to be larger than INT (or even BIGINT) suragate primary keys and there are very few domains where there isn't some danger of having a natural primary key change. To take your example, a city changing its name is not a terribly uncommon occurrence. When a city's name changes you are then left with either an update that needs to touch every instance of city as a foreign key or with a primary key that no longer reflects reality ("The data shows Leningrad, but it really is St. Petersburg.")
So in sum, natural primary keys:
Take up more disc space (most of the time)
Are more susceptible to change (in the majority of cases)
Are more human readable (as long as they don't change)
Whether #1 and #2 are sufficiently counteracted by #3 depends on what you are building and what its use is.
I think that this question
What are the design criteria for primary keys?
gives a really good overview of the tradeoffs you might be making. I think the answer given is the correct one, but its brevity belies some significant thinking you actually have to do to work out what's right for you.
(From that answer)
The criteria for consideration of a primary key are:
Uniqueness
Irreducibility (no subset of the key uniquely identifies a row in the table)
Simplicity (so that relational representation & manipulation can be simpler)
Stability (should not be altered frequently)
Familiarity (meaningful to the user)
For what it's worth, the small number of times I've had problems with scaling by choosing strings as the primary key is about the same as the number of time's I've had problems with redundant data using an autoincrement key. The problems that arise with autoincrement keys are worse, in my opinion, because you don't usually see them as soon.
A primary key must be unique and immutable, a human-readable string can be used as a PK so long as it meets both of those requirements.
In the example you've given, it sounds fine, given that cities don't change their names (and in the rare event they do then you can change the PK value with enough effort).
One of the main reasons you'd use numeric PKs instead of strings is performance (the other being to take advantage of automatically-incrementing IDs, see IDENTITY). If you anticipate more than a hundred queries per second on your textual PK then I would move to use int or bigint as a PK type. When you reach that level of database size and complexity you tend to stop using SSMS to edit table data directly and use your own tools, which would presumably perform a JOIN so you'd get the city name in the same resultset as the city's numeric PK.
you are allowed.
it is generally not the best practice.
numeric - auto incrementing keys are preferred. they are easily maintained and allow for coding of input forms and other interfaces where the user does not have to think up a new string as a key...
imagine: should it be washington, or washington_dc or dc or washingtondc.. etc.
I found this reading material on choosing a primary key.
Is there a guide / blog post on how to choose the primary key for a given table?
Should I use a auto-incremented/generated key, or should I base the primary key on the data being modeled (assuming it has a truly unique field)?
Should the primary key always be long for performance's sake, or can I take an external unique id as primary key, even if it's a string?
I believe that in practice using a natural key is rarely better than a surrogate key.
The following are the main disadvantages of using a natural key as the primary key:
You might have an incorrect key value, or you may simply want to rename a key value. To edit it, you would have to update all the tables that would be using it as a foreign key.
It is often difficult to have a truly unique natural key.
Natural keys are often strings. An index on an numeric field will be much more compact than one on a string field.
There is no hard rule on what the data type of the primary key should be. A numeric key normally performs better, but you could use a string, especially if the table is not big, and the tables that reference it are not big either.
A key is a set of attributes with two fundamental features: uniqueness and minimality. Minimality means the key has only the minimum number of attributes required to ensure uniqueness.
There are three criteria commonly applied as a guide to choosing a good key:
Familiarity - keys should be meaningful and familiar to the people who use them
Simplicity - keys should be as simple and concise as possible
Stability - key values should change infrequently
These are good guidelines but are not absolute requirements. In all cases functional requirements and the needs of data integrity should determine what keys to use.
I use surrogate keys, often referred to as non-sensical keys, made up of an autogenerated int/bigint datatype.
Here are some of the reasons I like using these keys.
When deleting several items from a list (such as old email) you can supply a comma separated list of integers instead of guids or natural keys
I find it makes writing your own cascade deletes easier
I think inner-joins are faster on integer fields
It can make learning a new system without documentation easier to understand.
Here are a couple of blog posts about primary keys:
http://www.mysqlperformanceblog.com/2006/10/03/long-primary-key-for-innodb-tables/
http://www.mysqlperformanceblog.com/2007/03/13/to-uuid-or-not-to-uuid/
I have worked with a lot of different data models in professional systems (mostly bank software) and there were different solutions. There was the GUID solution I have seen and it seemed not to have impacted the performances too much. I have seen the "number provided by a service as a system wide unique number". I have seen algorithms of providing something like a GUID "but shorter". I have seen also that the business key was used (like the account number) which is poor design and caused problems and I would not recommend it. I have seen the auto-incremented key for each table.
What did I like the most? The number provided by a service as a system wide number. It works well. And with a simple key translation table one can use a user key (like an account number) to find out what unique number and what sort of data object (not necessarily the table because the same unique key may apply to several tables if a data object is split up on different tables depending on its type).
So is there a blog or something? Well I have a book to recommend called "Data Modeling Essentials" by Graeme Simsion and Graham Witt. They might not suggest my preferred solution but they give many real live examples and show the different kind of solutions that are possible.
I always choose uuid as a primary key. In comparison to int/long key, there is a slight overhead, but there are a lot of benefits: you cannot run into type overflow, you can shard database later on without changing primary keys, you can integrate with other systems and be sure that your primary keys are always unique, uuid cannot be guessed etc.
As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.
Closed 12 years ago.
Almost every table in every database I've seen in my 7 years of development experience has an auto-incrementing primary key. Why is this? If I have a table of U.S. states where each state where each state must have a unique name, what's the use of an auto-incrementing primary key? Why not just use the state name as the primary key? Seems to me like an excuse to allow duplicates disguised as unique rows.
This seems plainly obvious to me, but then again, no one else seems to be arriving at and acting on the same logical conclusion as me, so I must assume there's a good chance I'm wrong.
Is there any real, practical reason we need to use auto-incrementing keys?
This question has been asked numerous times on SO and has been the subject of much debate over the years amongst (and between) developers and DBAs.
Let me start by saying that the premise of you question implies that one approach is universally superior to the other ... this is rarely the case in real life. Surrogate keys and natural keys both have their uses and challenges - and it's important to understand what they are. Whichever choice you make in your system, keep in mind there is benefit to consistency - it makes the data model easier to understand and easier to develop queries and applications for. I also want to say that I tend to prefer surrogate keys over natural keys for PKs ... but that doesn't mean that natural keys can't sometimes be useful in that role.
It is important to realize that surrogate and natural keys are NOT mutually exclusive - and in many cases they can complement each other. Keep in mind that a "key" for a database table is simply something that uniquely identifies a record (row). It's entirely possible for a single row to have multiple keys representing the different categories of constraints that make a record unique.
A primary key, on the other hand, is a particular unique key that the database will use to enforce referential integrity and to represent a foreign key in other tables. There can only be a single primary key for any table. The essential quality of a primary key is that it be 100% unique and non-NULL. A desirable quality of a primary key is that it be stable (unchanging). While mutable primary keys are possible - they cause many problems for database that are better avoided (cascading updates, RI failures, etc). If you do choose to use a surrogate primary key for your table(s) - you should also consider creating unique constraints to reflect the existence of any natural keys.
Surrogate keys are beneficial in cases where:
Natural keys are not stable (values may change over time)
Natural keys are large or unwieldy (multiple columns or long values)
Natural keys can change over time (columns added/removed over time)
By providing a short, stable, unique value for every row, we can reduce the size of the database, improve its performance, and reduce the volatility of dependent tables which store foreign keys. There's also the benefit of key polymorphism, which I'll get to later.
In some instances, using natural keys to express relationships between tables can be problematic. For instance, imagine you had a PERSON table whose natural key was {LAST_NAME, FIRST_NAME, SSN}. What happens if you have some other table GRANT_PROPOSAL in which you need to store a reference to a Proposer, Reviewer, Approver, and Authorizer. You now need 12 columns to express this information. You also need to come up with a naming convention of some kind to identify which columns belong to which kind of individual. But what if your PERSON table required 6, or 8, or 24 columns to for a natural key? This rapidly becomes unmanageable. Surrogate keys resolve such problems by divorcing the semantics (meaning) of a key from its use as an identifier.
Let's also take a look at the example you described in your question.
Should the 2-character abbreviation of a state be used as the primary key of that table.
On the surface, it looks like the abbreviation field meets the requirements of a good primary key. It's relatively short, it is easy to propagate as a foreign key, it looks stable. Unfortunately, you don't control the set of abbreviations ... the postal service does. And here's an interesting fact: in 1973 the USPS changed the abbreviation of Nebraska from NB to NE to minimize confusion with New Brunswick, Canada. The moral of the story is that natural keys are often outside of the control of the database ... and they can change over time. Even when you think they cannot. This problem is even more pronounced for more complicated data like people, or products, etc. As businesses evolve, the definitions for what makes such entities unique can change. And this can create significant problems for data modelers and application developers.
Earlier I mentioned that primary keys can support key polymorphism. What does that mean? Well, polymorphism is the ability of one type, A, to appear as and be used like another type, B. In databases, this concept refers to the ability to combine keys from different classes of entities into a single table. Let's look at an example. Imagine for a moment that you want have an audit trail in your system that identifies which entities were modified by which user on what date. It would be nice to create a table with the fields: {ENTITY_ID, USER_ID, EDIT_DATE}. Unfortunately, using natural keys, different entities have different keys. So now we need to create a separate linking table for each kind of entity ... and build our application in a manner where it understand the different kinds of entities and how their keys are shaped.
Don't get me wrong. I'm not advocating that surrogate keys should ALWAYS be used. In the real world never, ever, and always are a dangerous position to adopt. One of the biggest drawbacks of surrogate keys is that they can result in tables that have foreign keys consisting of lots of "meaningless" numbers. This can make it cumbersome to interpret the meaning of a record since you have to join or lookup records from other tables to get a complete picture. It also can make a distributed database deployment more complicated, as assigning unique incrementing numbers across servers isn't always possible (although most modern database like Oracle and SQLServer mitigate this via sequence replication).
No.
In most cases, having a surrogate INT IDENTITY key is an easy option: it can be guaranteed to be NOT NULL and 100% unique, something a lot of "natural" keys don't offer - names can change, so can SSN's and other items of information.
In the case of state abbreviations and names - if anything, I'd use the two-letter state abbreviation as a key.
A primary key must be:
unique (100% guaranteed! Not just "almost" unique)
NON NULL
A primary key should be:
stable if ever possible (not change - or at least not too frequently)
State two-letter codes definitely would offer this - that might be a candidate for a natural key. A key should also be small - an INT of 4 bytes is perfect, a two-letter CHAR(2) column just the same. I would not ever use a VARCHAR(100) field or something like that as a key - it's just too clunky, most likely will change all the time - not a good key candidate.
So while you don't have to have an auto-incrementing "artificial" (surrogate) primary key, it's often quite a good choice, since no naturally occuring data is really up to the task of being a primary key, and you want to avoid having huge primary keys with several columns - those are just too clunky and inefficient.
I think the use of the word "Primary", in the phrase "Primary" Key is in a real sense, misleading.
First, use the definition that a "key" is an attribute or set of attributes that must be unique within the table,
Then, having any key serves several often mutually inconsistent purposes.
Purpose 1. To use as joins conditions to one or many records in child tables which have a relationship to this parent table. (Explicitly or implicitly defining a Foreign Key in those child tables)
Purpose 2. (related) Ensuring that child records must have a parent record in the parent table (The child table FK must exist as Key in the parent table)
Purpose 3. To increase performance of queries that need to rapidly locate a specific record/row in the table.
Purpose 4. (Most Important from data consistency perspective!) To ensure data consistency by preventing duplicate rows which represent the same logical entity from being inserted itno the table. (This is often called a "natural" key, and should consist of table (entity) attributes which are relatively invariant.)
Clearly, any non-meaningfull, non-natural key (like a GUID or an auto-generated integer is totally incapable of satisfying Purpose 4.
But often, with many (most) tables, a totally natural key which can provide #4 will often consist of multiple attributes and be excessively wide, or so wide that using it for purposes #1, #2, or #3 will cause unacceptable performance consequencecs.
The answer is simple. Use both. Use a simple auto-Generating integral key for all Joins and FKs in other child tables, but ensure that every table that requires data consistency (very few tables don't) have an alternate natural unique key that will prevent inserts of inconsistent data rows... Plus, if you always have both, then all the objections against using a natural key (what if it changes? I have to change every place it is referenced as a FK) become moot, as you are not using it for that... You are only using it in the one table where it is a PK, to avoid inconsistent duplciate data...
The only time you can get away without both is for a completely stand alone table that participates in no relationships with other tables and has an obvious and reliable natural key.
In general, a numeric primary key will perform better than a string. You can additionaly create unique keys to prevent duplicates from creeping in. That way you get the assurance of no duplicates, but you also get the performance of numbers (vs. strings in your scenario).
In all likelyhood, the major databases have some performance optimizations for integer-based primary keys that are not present for string-based primary keys. But, that is only a reasonable guess.
Yes, in my opinion every table needs an auto incrementing integer key because it makes both JOINs and (especially) front-end programming much, much, much easier. Others feel differently, but this is over 20 years of experience speaking.
The single exception is small "code" or "lookup" tables in which I'm willing to substitute a short (4 or 5 character) TEXT code value. I do this because the I often use a lot of these in my databases and it allows me to present a meaningful display to the user without having to look up the description in the lookup table or JOIN it into a result set. Your example of a States table would fit in this category.
No, absolutely not.
Having a primary key which can't change is a good idea (UPDATE is legal for primary key columns, but in general potentially confusing and can create problems for child rows). But if your application has some other candidate which is more suitable than an auto-incrementing value, then you should probably use that instead.
Performance-wise, in general fewer columns are better, and particularly fewer indexes. If you have another column which has a unique index on it AND can never be changed by any business process, then it may be a suitable primary key.
Speaking from a MySQL (Innodb) perspective, it's also a good idea to use a "real" column as a primary key rather than an "artificial" one, as InnoDB always clusters the primary key and includes it in secondary indexes (that is how it finds the rows in them). This gives it potential to do useful optimisation with a primary key which it can't with any other unique index. MSSQL users often choose to cluster the primary key, but it can also cluster a different unique index.
EDIT:
But if it's a small database and you don't really care about performance or size too much, adding an unnecessary auto-increment column isn't that bad.
A non auto-incrementing value (e.g. UUID, or some other string generated according to your own algorithm) may be useful for distributed, sharded, or diverse systems where maintaining a consistent auto-incrementing ID is difficult (or impossible - think of a distributed system which continues to insert rows on both sides of a network partition).
I think there are two things that may explain the reason why auto-incrementing keys are sometimes used:
Space consideration; ok your state name doesn't amount to much, but the space it takes may add up. If you really want to store the state with its name as a primary key, then go ahead, but it will take more place. That may not be a problem in certain cases, and it sounds like a problem of olden days, but the habit is perhaps ingrained. And we programmers and DBA do love habits :D
Defensive consideration: i recently had the following problem; we have users in the database where the email is the key to all identification. Why not make the email the promary key? except suddenly border cases creep in where one guy must be there twice to have two different adresses, and nobody talked about it in the specs so the adress is not normalized, and there's this situation where two different emails must point to the same person and... After a while, you stop pulling your hairs out and add the damn integer id column
I'm not saying it's a bad habit, nor a good one; i'm sure good systems can be designed around reasonable primary keys, but these two points lead me to believe fear and habit are two among the culprits
It's a key component of relational databases. Having an integer relate to a state instead of having the whole state name saves a bunch of space in your database! Imagine you have a million records referencing your state table. Do you want to use 4 bytes for a number on each of those records or do you want to use a whole crapload of bytes for each state name?
Here are some practical considerations.
Most modern ORMs (rails, django, hibernate, etc.) work best when there is a single integer column as the primary key.
Additionally, having a standard naming convention (e.g. id as primary key and table_name_id for foreign keys) makes identifying keys easier.
I know this is subjective, but I'd like to know peoples opinions and hopefully some best practices that I can apply when designing sql server table structures.
I personally feel that keying a table on a fixed (max) length varchar is a no-no, because it means having to also propogate the same fixed length across any other tables that use this as a foreign key. Using an int, would avoid having to apply the same length across the board, which is bound to lead to human error, i.e. 1 table has varchar (10), and the other varchar (20).
This sounds like a nightmare to initially setup, plus means future maintaining of the tables is cumbersome too. For example, say the keyed varchar column suddenly becomes 12 chars instead of 10. You now have to go and update all the other tables, which could be a huge task years down the line.
Am I wrong? Have I missed something here? I'd like to know what others think of this and if sticking with int for primary keys is the best way to avoid maintainace nightmares.
When choosing the primary key usualy you also choose the clustered key. Them two are often confused, but you have to understand the difference.
Primary keys are logical business elements. The primary key is used by your application to identify an entity, and the discussion about primary keys is largely wether to use natural keys or surrogate key. The links go into much more detail, but the basic idea is that natural keys are derived from an existing entity property like ssn or phone number, while surrogate keys have no meaning whatsoever with regard to the business entity, like id or rowid and they are usually of type IDENTITY or some sort of uuid. My personal opinion is that surrogate keys are superior to natural keys, and the choice should be always identity values for local only applicaitons, guids for any sort of distributed data. A primary key never changes during the lifetime of the entity.
Clustered keys are the key that defines the physical storage of rows in the table. Most times they overlap with the primary key (the logical entity identifier), but that is not actually enforced nor required. When the two are different it means there is a non-clustered unique index on the table that implements the primary key. Clustered key values can actualy change during the lifetime of the row, resulting in the row being physically moved in the table to a new location. If you have to separate the primary key from the clustered key (and sometimes you do), choosing a good clustered key is significantly harder than choosing a primary key. There are two primary factors that drive your clustered key design:
The prevalent data access pattern.
The storage considerations.
Data Access Pattern. By this I understand the way the table is queried and updated. Remember that clustered keys determine the actual order of the rows in the table. For certain access patterns, some layouts make all the difference in the world in regard to query speed or to update concurency:
current vs. archive data. In many applications the data belonging to the current month is frequently accessed, while the one in the past is seldom accessed. In such cases the table design uses table partitioning by transaction date, often times using a sliding window algorithm. The current month partition is kept on filegroup located a hot fast disk, the archived old data is moved to filegroups hosted on cheaper but slower storage. Obviously in this case the clustered key (date) is not the primary key (transaction id). The separation of the two is driven by the scale requirements, as the query optimizer will be able to detect that the queries are only interested in the current partition and not even look at the historic ones.
FIFO queue style processing. In this case the table has two hot spots: the tail where inserts occur (enqueue), and the head where deletes occur (dequeue). The clustered key has to take this into account and organize the table as to physically separate the tail and head location on disk, in order to allow for concurency between enqueue and dequeue, eg. by using an enqueue order key. In pure queues this clustered key is the only key, since there is no primary key on the table (it contains messages, not entities). But most times the queue is not pure, it also acts as the storage for the entities, and the line between the queue and the table is blured. In this case there is also a primary key, which cannot be the clustered key: entities may be re-enqueued, thus changing the enqueue order clustered key value, but they cannot change the primary key value. Failure to see the separation is the primary reason why user table backed queues are so notoriously hard to get right and riddled with deadlocks: because the enqueue and dequeue occur interleaved trought the table, instead of localized at the tail and the head of the queue.
Correlated processing. When the application is well designed it will partition processing of correlated items between its worker threads. For instance a processor is designed to have 8 worker thread (say to match the 8 CPUs on the server) so the processors partition the data amongst themselves, eg. worker 1 picks up only accounts named A to E, worker 2 F to J etc. In such cases the table should be actually clustered by the account name (or by a composite key that has the leftmost position the first letter of account name), so that workers localize their queries and updates in the table. Such a table would have 8 distinct hot spots, around the area each worker concentrates at the moment, but the important thing is that they don't overlap (no blocking). This kind of design is prevalent on high throughput OLTP designs and in TPCC benchmark loads, where this kind of partitioning also reflects in the memory location of the pages loaded in the buffer pool (NUMA locality), but I digress.
Storage Considerations. The clustered key width has huge repercursions in the storage of the table. For one the key occupies space in every non-leaf page of the b-tree, so a large key will occupy more space. Second, and often more important, is that the clustered key is used as the lookup key by every non-clustred key, so every non-clustered key will have to store the full width of the clustered key for each row. This is what makes large clustered keys like varchar(256) and guids poor choices for clustered index keys.
Also the choice of the key has impact on the clustered index fragmentation, sometimes drastically affecting performance.
These two forces can sometimes be antagonistic, the data access pattern requiring a certain large clustered key which will cause storage problems. In such cases of course a balance is needed, but there is no magic formula. You measure and you test to get to the sweet spot.
So what do we make from all this? Always start with considering clustered key that is also the primary key of the form entity_id IDENTITY(1,1) NOT NULL. Separate the two and organize the table accordingly (eg. partition by date) when appropiate.
I would definitely recommend using an INT NOT NULL IDENTITY(1,1) field in each table as the
primary key.
With an IDENTITY field, you can let the database handle all the details of making sure it's really unique and all, and the INT datatype is just 4 bytes, and fixed, so it's easier and more suited to be used for the primary (and clustering) key in your table.
And you're right - INT is an INT is an INT - it will not change its size of anything, so you won't have to ever go recreate and/or update your foreign key relations.
Using a VARCHAR(10) or (20) just uses up too much space - 10 or 20 bytes instead of 4, and what a lot of folks don't know - the clustering key value will be repeated on every single index entry on every single non-clustered index on the table, so potentially, you're wasting a lot of space (not just on disk - that's cheap - but also in SQL Server's main memory). Also, since it's variable (might be 4, might be 20 chars) it's harder to SQL server to properly maintain a good index structure.
Marc
I'd agree that in general an INT (or identity) field type is the best choice in most "normal" database designs:
it requires no "algorithm" to generate the id/key/value
you have fast(er) joins and the optimizer can work a lot harder over ranges and such under the hood
you're following a defacto standard
That said, you also need to know your data. If you're going to blow through a signed 32-bit int, you need to think about unsigned. If you're going to blow through that, maybe 64-bit ints are what you want. Or maybe you need a UUID/hash to make syncing between database instances/shards easier.
Unfortunately, it depends and YMMV but I'd definitely use an int/identity unless you have a good reason not to.
Like you said, consistency is key. I personally use unsigned ints. You're not going to run out of them unless you are working with ludicrous amounts of data, and you can always know any key column needs to be that type and you never have to go looking for the right value for individual columns.
Based on going through this exercise countless times and then supporting the system with the results, there are some caveats to the blanket statement that INT is always better. In general, unless there is a reason, I would go along with that. However, in the trenches, here are some pros and cons.
INT
Use unless good reason not to do so.
GUID
Uniqueness - One example is the case where there is one way communication between remote pieces of the program and the side that needs to initiate is not the side with the database. In that case, setting a Guid on the remote side is safe where selecting an INT is not.
Uniqueness Again - A more far fetched scenario is a system where multiple customers are coexisting in separate databases and there is migration between customers like similar users using a suite of programs. If that user signs up for another program, their user record can be used there without conflict. Another scenario is if customers acquire entities from each other. If both are on the same system, they will often expect that migration to be easier. Essentially, any frequent migration between customers.
Hard to Use - Even an experienced programmer cannot remember a guid. When troubleshooting, it is often frustrating to have to copy and paste identifiers for queries, especially if the support is being done with a remote access tool. It is much easier to constantly refer to SELECT * FROM Xxx WHERE ID = 7 than SELECT * FROM Xxx WHERE ID = 'DF63F4BD-7DC1-4DEB-959B-4D19012A6306'
Indexing - using a clustered index for a guid field requires constant rearrangement of the data pages and is not as efficient to index as INTs or even short strings. It can kill performance - don't do it.
CHAR
Readability - Although conventional wisdom is that nobody should be in the database, the reality of systems is that people will have access - hopefully personnel from your organization. When those people are not savvy with join syntax, a normalized table with ints or guids is not clear without many other queries. The same normalized table with SOME string keys can be much more usable for troubleshooting. I tend to use this for the type of table where I supply the records at installation time so they do not vary. Things like StatusID on a major table is much more usable for support when the key is 'Closed' or 'Pending' than a digit. Using traditional keys in these areas can turn an easily resolved issue to something that requires developer assistance. Bottlenecks like that are bad even when caused by letting questionable personnel access to the database.
Constrain - Even if you use strings, keep them fixed length, which speeds indexing and add a constraint or foreign key to keep garbage out. Sometimes using this string can allow you to remove the lookup table and maintain the selection as a simple Enum in the code - it is still important to constrain the data going into this field.
For best performance, 99.999% of the time the primary key should be a single integer field.
Unless you require the primary key to be unique across multiple tables in a database or across multiple databases. I am assuming that you are asking about MS SQL-Server because that is how your question was tagged. In which case, consider using the GUID field instead. Though better than a varchar, the GUID field performance is not as good as an integer.
Use INT. Your points are all valid; I would prioritize as:
Ease of using SQL auto increment capabiity - why reinvent the wheel?
Managability - you don't want to have to change the key field.
Performance
Disk Space
1 & 2 require the developer's time/energy/effort. 3 & 4 you can throw hardware at.
If Joe Celko was on here, he would have some harsh words... ;-)
I want to point out that INTs as a hard and fast rule aren't always appropriate. Say you have a vehicle table with all types of cars trucks, etc. Now say you had a VehicleType table. If you wanted to get all trucks you might do this (with an INT identity seed):
SELECT V.Make, V.Model
FROM Vehicle as V
INNER JOIN VehicleType as VT
ON V.VehicleTypeID = VT.VehicleTypeID
WHERE VT.VehicleTypeName = 'Truck'
Now, with a Varchar PK on VehicleType:
SELECT Make, Model
FROM Vehicle
WHERE VehicleTypeName = 'Truck'
The code is a little cleaner and you avoid a join. Perhaps the join isn't the end of the world, but if you only have one tool in your toolbox, you're missing some opportunities for performance gains and cleaner schemas.
Just a thought. :-)
While INT is generally recommended, it really depends on your situation.
If you're concerned with maintainability, then other types are just as feasible. For example, you could use a Guid very effectively as a primary key. There's reasons for not doing this, but consistency is not one of them.
But yes, unless you have a good reason not to, an int is the simplest to use, and the least likely to cause you any problems.
With PostgreSQL I generally use the "Serial" or "BigSerial" 'data type' for generating primary keys. The values are auto incremented and I always find integers to be easy to work with. They are essentially equivalent to a MySQL integer field that is set to "auto_increment".
One should think hard about whether 32-bit range is enough for what you're doing. Twitter's status IDs were 32-bit INTs and they had trouble when they ran out.
Whether to use a BIGINT or a UUID/GUID in that situation is debatable and I'm not a hardcore database guy, but UUIDs can be stored in a fixed-length VARCHAR without worrying that you'll need to change the field size.
We have to keep in mind that the primary key of a table should not have "business logic" and it should be only an identity of the record it belongs. Following this simple rule an int and especially an identity int is a very good solution. By asking about varchar I guess that you mean using for example the "Full Name" as a key to the "people" table. But what if we want to change the name from "George Something" to "George A. Something" ? And what size will the field be ? If we change the size we have to change the size on all foreign tables too. So we should avoid logic on keys. Sometimes we can use the social ID (integer value) as key but I avoid that too. Now if a project has the prospects to scale up you should consider using Guids too (uniqueidentifier SQL type).
Keeping in mind that this is quite old a question, I still want to make the case for using varchar with surrogate keys fur future readers:
An environment with several replicated machines
Scenarios where it is required that the ID of a to be inserted row is known before it is actually inserted (i.e., the client assigns this ID, not the database)
I'm just wondering what the optimal solution is here.
Say I have a normalized database. The primary key of the whole system is a varchar. What I'm wondering is should I relate this varchar to an int for normalization or leave it? It's simpler to leave as a varchar, but it might be more optimal
For instance I can have
People
======================
name varchar(10)
DoB DateTime
Height int
Phone_Number
======================
name varchar(10)
number varchar(15)
Or I could have
People
======================
id int Identity
name varchar(10)
DoB DateTime
Height int
Phone_Number
======================
id int
number varchar(15)
Add several other one-to-many relationships of course.
What do you all think? Which is better and why?
I believe that the majority of people who have developed any significant sized real world database applications will tell you that surrogate keys are the only realistic solution.
I know the academic community will disagree but that is the difference between theoretical purity and practicality.
Any reasonable sized query that has to do joins between tables that use non-surrogate keys where some tables have composite primary keys quickly becomes unmaintainable.
Can you really use names as primary keys? Isn't there a high risk of several people with the same name?
If you really are so lucky that your name attribute can be used as primary key, then - by all means - use that. Often, though, you will have to make something up, like a customer_id, etc.
And finally: "NAME" is a reserved word in at least one DBMS, so consider using something else, e.g. fullname.
Using any kind of non-synthetic data (i.e. anything from the user, as opposed to generated by the application) as a PK is problematic; you have to worry about culture/localization differences, case sensitivity (and other issues depending on DB collation), can result in data problems if/when that user-entered data ever changes, etc.
Using non-user-generated data (Sequential GUIDs (or non-sequential if your DB doesn't support them or you don't care about page splits) or identity ints (if you don't need GUIDs)) is much easier and much safer.
Regarding duplicate data: I don't see how using non-synthetic keys protects you from that. You still have issues where the user enters "Bob Smith" instead of "Bob K. Smith" or "Smith, Bob" or "bob smith" etc. Duplication management is necessary (and pretty much identical) regardless of whether your key is synthetic or non-synthetic, and non-synthetic keys have a host of other potential issues that synthetic keys neatly avoid.
Many projects don't need to worry about that (tightly constrained collation choices avoid many of them, for example) but in general I prefer synthetic keys. This is not to say you can't be successful with organic keys, clearly you can, but for many projects they're not the better choice.
I think if your VARCHAR was larger you would notice you're duplicating quite a bit of data throughout the database. Whereas if you went with a numeric ID column, you're not duplicating nearly the same amount of data when adding foreign key columns to other tables.
Moreover, textual data is a royal pain in terms of comparisons, your life is much easier when you're doing WHERE id = user_id versus WHERE name LIKE inputname (or something similar).
If the "name" field really is appropriate as a primary key, then do it. The database will not get more normalized by creating a surrogate key in that case. You will get some duplicate strings for foreign keys, but that is not a normalization issue, since the FK constraint guarantrees integrity on strings just as it would on surrogate keys.
However you are not explaining what the "name" is. In practice it is very seldom that a string is appropriate as a primary key. If it is the name of a person, it wont work as a PK, since more than one person can have the same name, people can change names and so on.
One thing that others don't seem to have mentioned is that joins on int fields tend to perform better than joins on varchar fields.
And I definitely would always use a surrogate key over using names (of people or businesses) because they are never unique over time. In our database, for instance, we have 164 names with over 100 instances of the same name. This clearly shows the dangers of considering using name as a key field.
The original question is not one of normalization. If you have a normalized database, as you stated, then you do not need to change it for normalization reasons.
There are really two issues in your question. The first is whether ints or varchars a preferable for use as primary keys and foreign keys. The second is whether you can use the natural keys given in the problem definition, or whether you should generate a synthetic key (surrogate key) to take the place of the natural key.
ints are a little more concise than varchars, and a little more efficient for such things as index processing. But the difference is not overwhelming. You should probably not make your decision on this basis alone.
The question of whether the natural key provided really works as a natural key or not is much more significant. The problem of duplicates in a "name" column is not the only problem. There is also the problem of what happens when a person changes her name. This problem probably doesn't surface in the example you've given, but it does surface in lots of other database applications. An example would be the transcript over four years of all the courses taken by a student. A woman might get married and change her name in the course of four years, and now you're stuck.
You either have to leave the name unchanged, in which case it no longer agrees with the real world, or update it retroactively in all the courses the person took, which makes the database disagree with the printed rosters made at the time.
If you do decide on a synthetic key, you now have to decide whether or not the application is going to reveal the value of the synthetic key to the user community. That's another whole can of worms, and beyond the scope of this discussion.