nULL OR nOT?
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
Was about to say if you could have a valid customer without an address then...but it still doesn't make sense to not just use NULL. If you can actually have a customer without an address then this would be better indicated with a separate 'NoAddress' bool/bit field, as its a specific bit of data. As for the price thing thats just confusing as hell...-1 would be slightly better (as at least a reader would be forced to think 'what the hell?' for a sec), but 0 in amount field that doesn't mean '0 amount of' is just nuts.
-
Was about to say if you could have a valid customer without an address then...but it still doesn't make sense to not just use NULL. If you can actually have a customer without an address then this would be better indicated with a separate 'NoAddress' bool/bit field, as its a specific bit of data. As for the price thing thats just confusing as hell...-1 would be slightly better (as at least a reader would be forced to think 'what the hell?' for a sec), but 0 in amount field that doesn't mean '0 amount of' is just nuts.
What annoys me about the customer thing is that, rather than creating an address record with empty columns (which is daft but forgivable) they created a single record with a zero Id and refer all customers with unknown addresses to that single record - so the data tells you "All of these customers live at the same address - which happens to have empty strings as the details" Which is just poor modelling.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
_Maxxx_ wrote:
I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item.
I'd like to address this example. If '0' represents that the item hasn't been sold then the database design problem isn't a NULL problem. The problem at that point is assigning more than one meaning to a database field - something the NULL designation wouldn't remedy. In a normalized database that record would have a corresponding record in another table that would contain the details of the sale. The presence of that record would indicate the item was sold whereas the lack of a sale record would indicate that it wasn't sold. Overloading the sale price field with more than one meaning is bad design.
_Maxxx_ wrote:
What's wrong with a null value which tells you explicitly "The customer has no address"
Questions: 1: How does the address appear to the end user in the UI? Do you change the UI based on whether or not the address is NULL or merely empty strings? 2: What happens in the UI if the address is partially filled in? If the first line is supplied but the second line is NULL does the second line show up as a different color or something? On a line per line basis how do I tell, in the UI, if the line is NULL or merely empty? 3: What if the customer had an address but lost it? Can the program user set the address fields back to NULL to indicate that the customer no longer has an address? 4: What if a user accidently enters an address but then removes it? Does that mean, for database purposes, that the account had an address but now doesn't have one or does it mean that the account never had an address? What would NULL mean for those values? As per my initial example, I think giving a field additional meaning is a rat's nest of problems. The NULL thing may have some uses, but it's an academic curiosity as far as I'm concerned.
-
_Maxxx_ wrote:
I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item.
I'd like to address this example. If '0' represents that the item hasn't been sold then the database design problem isn't a NULL problem. The problem at that point is assigning more than one meaning to a database field - something the NULL designation wouldn't remedy. In a normalized database that record would have a corresponding record in another table that would contain the details of the sale. The presence of that record would indicate the item was sold whereas the lack of a sale record would indicate that it wasn't sold. Overloading the sale price field with more than one meaning is bad design.
_Maxxx_ wrote:
What's wrong with a null value which tells you explicitly "The customer has no address"
Questions: 1: How does the address appear to the end user in the UI? Do you change the UI based on whether or not the address is NULL or merely empty strings? 2: What happens in the UI if the address is partially filled in? If the first line is supplied but the second line is NULL does the second line show up as a different color or something? On a line per line basis how do I tell, in the UI, if the line is NULL or merely empty? 3: What if the customer had an address but lost it? Can the program user set the address fields back to NULL to indicate that the customer no longer has an address? 4: What if a user accidently enters an address but then removes it? Does that mean, for database purposes, that the account had an address but now doesn't have one or does it mean that the account never had an address? What would NULL mean for those values? As per my initial example, I think giving a field additional meaning is a rat's nest of problems. The NULL thing may have some uses, but it's an academic curiosity as far as I'm concerned.
MehGerbil wrote:
The problem at that point is assigning more than one meaning to a database field - something the NULL designation wouldn't remedy.
It would remedy it perfectly. In your example I think you are suggesting that the database wold have a table containing just the sale price, while another table would hold details of the sold item (or, actually, the item Id) and the sale to which it belongs, etc. As the sale price is an intrinsic part of the line item in most cases I'd suggest that the 'invoice line' table would include the sale price as a column, and would not hold a reference to another table.
MehGerbil wrote:
: How does the address appear to the end user in the UI? Do you change the UI based on whether or not the address is NULL or merely empty strings?
That shouldn't make any difference to the Data design. The UI is a representation that should be able to change independently. If the users want to see blanks on the screen, or big red starts, or the words "address unknown" , or whatever, that should not affect our DB design in any way.
MehGerbil wrote:
: What happens in the UI if the address is partially filled in? If the first line is supplied but the second line is NULL does the second line show up as a different color or something? On a line per line basis how do I tell, in the UI, if the line is NULL or merely empty?
Again, you're asking UI questions - this shouldn't change our data design.
MehGerbil wrote:
3: What if the customer had an address but lost it? Can the program user set the address fields back to NULL to indicate that the customer no longer has an address?
That's a business decision - can an address be deleted once it's added? Again, no effect on data design
MehGerbil wrote:
4: What if a user accidently enters an address but then removes it? Does that mean, for database purposes, that the account had an address but now doesn't have one or does it mean that the account never had an address? What would NULL mean for those values?
It depends whether the requirement for history tracking is there or not. If the system needs to know that there used to be an address and now there isn't then we need to keep history. If we need to keep history then ab address hist
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
The databases in question used VB6 for the front-end? (Zero is common for extremely low values of null in VB6). Primary reason that I don't like to use 0 for a key value and start from one, don't want to break anything. Now, as far as not null goes, I prefer the fourth normal form in cases you have described. I will leave the understanding as an exercise for the reader.
Need custom software developed? I do custom programming based primarily on MS tools with an emphasis on C# development and consulting. "And they, since they Were not the one dead, turned to their affairs" -- Robert Frost "All users always want Excel" --Ennis Lynch
-
MehGerbil wrote:
The problem at that point is assigning more than one meaning to a database field - something the NULL designation wouldn't remedy.
It would remedy it perfectly. In your example I think you are suggesting that the database wold have a table containing just the sale price, while another table would hold details of the sold item (or, actually, the item Id) and the sale to which it belongs, etc. As the sale price is an intrinsic part of the line item in most cases I'd suggest that the 'invoice line' table would include the sale price as a column, and would not hold a reference to another table.
MehGerbil wrote:
: How does the address appear to the end user in the UI? Do you change the UI based on whether or not the address is NULL or merely empty strings?
That shouldn't make any difference to the Data design. The UI is a representation that should be able to change independently. If the users want to see blanks on the screen, or big red starts, or the words "address unknown" , or whatever, that should not affect our DB design in any way.
MehGerbil wrote:
: What happens in the UI if the address is partially filled in? If the first line is supplied but the second line is NULL does the second line show up as a different color or something? On a line per line basis how do I tell, in the UI, if the line is NULL or merely empty?
Again, you're asking UI questions - this shouldn't change our data design.
MehGerbil wrote:
3: What if the customer had an address but lost it? Can the program user set the address fields back to NULL to indicate that the customer no longer has an address?
That's a business decision - can an address be deleted once it's added? Again, no effect on data design
MehGerbil wrote:
4: What if a user accidently enters an address but then removes it? Does that mean, for database purposes, that the account had an address but now doesn't have one or does it mean that the account never had an address? What would NULL mean for those values?
It depends whether the requirement for history tracking is there or not. If the system needs to know that there used to be an address and now there isn't then we need to keep history. If we need to keep history then ab address hist
_Maxxx_ wrote:
It would remedy it perfectly. In your example I think you are suggesting that the database wold have a table containing just the sale price, while another table would hold details of the sold item (or, actually, the item Id) and the sale to which it belongs, etc.
As the sale price is an intrinsic part of the line item in most cases I'd suggest that the 'invoice line' table would include the sale price as a column, and would not hold a reference to another table.Actually, I was taking a larger view of the issue than what you've stated here. An item could be sold, but it could also be donated, destroyed, lost, transferred, or disassembled into parts. My other table would be expressing how the item was 'released', means by which it was 'released' and other details. The item could be sold for a negative value, perhaps part of a package deal. NULL indicating 'not sold' seems rather limiting. Additionally, 'price' in my sphere may include two different amounts, two different payment methods, and perhaps two different payees which are summed to get the final total price. I'd have a table of 'transactions' that make up the 'sold price' and so 'sold price' wouldn't be on that item table. A fundamental rule of database design is that fields aren't overloaded with multiple meanings. When NULL for sale price means there is no sale price that is fine. When NULL for sale price means the item hasn't been sold that is a bad thing.
_Maxxx_ wrote:
Again, you're asking UI questions - this shouldn't change our data design.
I'm using the UI to illustrate a point about the data design. If NULL is going to mean something then that needs to be expressed to the users via the UI. If the users don't care about the academic difference between 'never had a value' and 'currently has no value' then I have to wonder if the database design is being made to conform to a textbook or if it's being designed to meet the user's needs.
_Maxxx_ wrote:
That's a business decision - can an address be deleted once it's added? Again, no effect on data design
I pointed out the issues that NULL can raise in the UI to illustrate that decisions made in the database design impact the application throughout and if NULL is going to mean a specific thing on a field then the developer has to have a way to communicate that to the use
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
_Maxxx_ wrote:
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database.
So at best they are ignorant.
_Maxxx_ wrote:
The argument I get is that "the customer does have an address, but we just don't know what it is yet."
The database is a persistent store - not a god. It "models" the data objects. And obviously if they want "We don't know what it is" then that is exactly what they should put in the database.
_Maxxx_ wrote:
but am willing to change my view if anyone can give a sound reason for any of this garbage.
One reason of course is that they do not in fact understand what null is. It confuses them. So a "sound" reason would be that you don't want to confuse them.
-
Mhm... depends (the typical answer of an engineer :))... in my main application, there are null values, due the fact, that I can query especially these dataset by searching for null values. I normally try to avoid null values to prevent any exceptions, but IMHO the rigorous obedience "null XOR not null" isn't suitable for all solutions...
_Maxxx_ wrote:
The argument I get is that "the customer does have an address, but we just don't know what it is yet."
Blocking -> "So what's the default value in case there isn't an address, and is it consistently used?" end of my two cents...
(yes|no|maybe)*
s_mon wrote:
I normally try to avoid null values to prevent any exceptions
The fallacy in that argument though is that there is in fact some value there which must be dealt with in a special way regardless. So you avoid a crash to exchange it for a harder to figure out data bug.
-
I am very restrictive allowing null values in database columns, but I don't try to replace the null values by some other kind of convention once I have loaded data rows into data objects. Not having to check for null in the program code is a small benefit which I gladly take, but that is not the main reason for this. I have seen databases where every column allowed nulls, including those which served as primary key without such a constraint being put on them. The importer, if it had a bad day, liked to insert rows into the database which consisted mostly of nulls. All this made the database one big source of useless data and errors. Validating a data row in which each and every value can be null already is a pain if it's done in one method and a code horror if it's done all over the application. Allowing nulls sparingly in the database helped a lot in getting this application in better shape: - Validation of data objects became easier (obviously). - The importer could not import junk anymore. Logging and solving the errors in the importer then led to a more useful database and the application had to deal with less error scenarios. - While they are no replacement for proper validation, constraints and restrictive use of nulls in the database provide another safeguard and the resulting errors give me the information I need to keep even bad applications running.
Sent from my BatComputer via HAL 9000 and M5
CDP1802 wrote:
- The importer could not import junk anymore
You explanation fails to explain how this was prevented. Certainly one can import 5 spaces or 5 X's into a string field or a date of June 1, 1099 and that wouldn't cause problems in the application? Conversely if one write a import tool/process that validates the data, all of the data, then this of course insures that all sorts of invalid data cannot be imported.
-
The correct answer is that, usually, it doesn't really matter that much. For the properties of some entities it is necessary to distinguish between 'unset' and the default value. This can be achieved with a nullable data type/field or by using some sentinel value. The selection of that sentinel value is easy to select in some cases and difficult in others. In some systems using a well chosen sentinel value that is within the range of the data type used can save a lot of space because it doesnt require another bool to indicate null/non null. In others, and I would argue that almost any database would fall into this category, using an existing facility (dot net's Nullable<> or a nullable database field) is preferable as you can reasonably expect others familiar with the product / technology to instantly understand what it is and why it's there.
_Josh_ wrote:
In some systems using a well chosen sentinel value that is within the range of the data type used can save a lot of space because it doesnt require another bool to indicate null/non null.
Perhaps, but it would require there are no other impacts on the system. For example that non-null fields do not always require space in the paged record and that indexes are not impacted in the same way. And it is only applicable for a table where there is a LOT of data. And thus it would be an optimization for one table requiring it for all tables is pointless.
-
_Maxxx_ wrote:
It would remedy it perfectly. In your example I think you are suggesting that the database wold have a table containing just the sale price, while another table would hold details of the sold item (or, actually, the item Id) and the sale to which it belongs, etc.
As the sale price is an intrinsic part of the line item in most cases I'd suggest that the 'invoice line' table would include the sale price as a column, and would not hold a reference to another table.Actually, I was taking a larger view of the issue than what you've stated here. An item could be sold, but it could also be donated, destroyed, lost, transferred, or disassembled into parts. My other table would be expressing how the item was 'released', means by which it was 'released' and other details. The item could be sold for a negative value, perhaps part of a package deal. NULL indicating 'not sold' seems rather limiting. Additionally, 'price' in my sphere may include two different amounts, two different payment methods, and perhaps two different payees which are summed to get the final total price. I'd have a table of 'transactions' that make up the 'sold price' and so 'sold price' wouldn't be on that item table. A fundamental rule of database design is that fields aren't overloaded with multiple meanings. When NULL for sale price means there is no sale price that is fine. When NULL for sale price means the item hasn't been sold that is a bad thing.
_Maxxx_ wrote:
Again, you're asking UI questions - this shouldn't change our data design.
I'm using the UI to illustrate a point about the data design. If NULL is going to mean something then that needs to be expressed to the users via the UI. If the users don't care about the academic difference between 'never had a value' and 'currently has no value' then I have to wonder if the database design is being made to conform to a textbook or if it's being designed to meet the user's needs.
_Maxxx_ wrote:
That's a business decision - can an address be deleted once it's added? Again, no effect on data design
I pointed out the issues that NULL can raise in the UI to illustrate that decisions made in the database design impact the application throughout and if NULL is going to mean a specific thing on a field then the developer has to have a way to communicate that to the use
MehGerbil wrote:
An item could be sold, but it could also be donated, destroyed, lost, transferred, or disassembled into parts.
I see what you're getting at - and I completely agree - it depends on requirements.
MehGerbil wrote:
When NULL for sale price means there is no sale price that is fine.
When NULL for sale price means the item hasn't been sold that is a bad thing.In the example I was thinking of, which was auction based, there is a 'reason not sold' which could be passed-in, destroyed etc. But there is no status for 'not sold because it hasn't been offered yet' - and a record was created in the Db with a sale price of zero and (effectively) a status of 'sold' and, in fact, a BuyerId of 0 (and a buyer record pre-exists with an Id of zero). So, as you can gather, the design sucks out of the box!
MehGerbil wrote:
I'm using the UI to illustrate a point about the data design.
If NULL is going to mean something then that needs to be expressed to the users via the UI.Again - I see what you mean. And agree. In the case above, having a Null buyer Id is more meaningful (IMHO) to show there is no buyer than having a buyer record with a description of something like 'No Buyer'. And, assuming there's no buyer, there is no sale price so it should be null and not zero. So Null isn't really representing anything other than the lack of a sale price - while the Null buyer shows there is no buyer. I guess my gripe is that people seem to sometimes avoid using Nulls because of their laziness (they can't be bothered to differentiate client-side so add a record representing no buyer, or whatever, to simplify their SQL - but really they end up with code that tests for magic Ids instead of Null.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
-
The databases in question used VB6 for the front-end? (Zero is common for extremely low values of null in VB6). Primary reason that I don't like to use 0 for a key value and start from one, don't want to break anything. Now, as far as not null goes, I prefer the fourth normal form in cases you have described. I will leave the understanding as an exercise for the reader.
Need custom software developed? I do custom programming based primarily on MS tools with an emphasis on C# development and consulting. "And they, since they Were not the one dead, turned to their affairs" -- Robert Frost "All users always want Excel" --Ennis Lynch
Ennis Ray Lynch, Jr. wrote:
The databases in question used VB6 for the front-end?
Interestingly in this case the OLD platform was VB6 - and teh new Db was designed by those old VB programmers - so you're right, that could well be the reason.
Ennis Ray Lynch, Jr. wrote:
I prefer the fourth normal form in cases you have described
our database is, unfortunately, not normal in any way ;) It falls short of 3NF in many areas already!
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
-
_Maxxx_ wrote:
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database.
So at best they are ignorant.
_Maxxx_ wrote:
The argument I get is that "the customer does have an address, but we just don't know what it is yet."
The database is a persistent store - not a god. It "models" the data objects. And obviously if they want "We don't know what it is" then that is exactly what they should put in the database.
_Maxxx_ wrote:
but am willing to change my view if anyone can give a sound reason for any of this garbage.
One reason of course is that they do not in fact understand what null is. It confuses them. So a "sound" reason would be that you don't want to confuse them.
jschell wrote:
One reason of course is that they do not in fact understand what null is. It confuses them. So a "sound" reason would be that you don't want to confuse them.
NOOOOOooooo! That is certainly not a sound reason for using bad practices - it is sound reason for education!
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
I like nulls; I don't like "magic values". Disallowing nulls is not a solution to the problem.
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
Is it possible they have simply read or heard that nulls are bad and so eradicate them with defaut values without understanding the database principles behind the statement? (I regularly use nulls in database design, normalising every last one away is rarely a worth it). Greg
-
d@nish wrote:
So, when exactly should I drink coffee?
At 8:50 - just before your scrum meeting. (I generally drink more tea than coffee just to reduce my caffeine and calorie intake - I drink earl gray tea, black nothing added) You do have a scrum meeting with yourself, don't you?
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
Darjeeling, black, then attend to correspondence. The coffee comes mid morning when the code starts happening. Lucky enough to have an espresso machine at the office, so it's usually a cappuccino for me :cool:
-- For a moment, nothing happened. Then, after a second or so, nothing continued to happen --
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
Yes, your example is FUD tactic for wrong practice. NULL is standard, my counter questions for those programmers is: "How do you handle three state logic and do you enforce "standards of yours" across all of your projects?"
Mislim, dakle jeo sam.
-
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
The main reason for doing this, especially on FK's, is that you don't have to worry about what kind of join you're doing. If you have nulls, you need to usually use an outer join rather than a full join or an inner join. Otherwise, it's sort of a semantic difference to use 0 instead of a null, but that can also be really dangerous. Null means something which is different from zero, or any other value.
_Maxxx_ wrote:
but am willing to change my view if anyone can give a sound reason for any of this garbage.
My reason regarding the joins should be moot - any decent SQL generator should know to use the correct join when dealing with nullable FK's. On the other hand, people are lazy. Not a good enough reason to adopt their policies, IMO. Marc
Testers Wanted!
Latest Article: User Authentication on Ruby on Rails - the definitive how to
My Blog -
I've come across a practice in more than one company where one or more develpers insit on not using null values in the database. So, for example, if a customer address is unknown, the AddressId on the Customer table will have a value of (for example) zero - and the Address table will be pre-populated with a record with an Id of zero and (usually) empty string or zero values in the other columns. I've also come across the situation where, rather than having a null value in a 'sold price' column, the value of zero represents an unsold item. Putting aside my feelings (that these are weak-minded imbeciles who shouldn't be allowed near a database) do any of you subscribe to this theory? What's wrong with a null value which tells you explicitly "The customer has no address" or "This item has no price" The argument I get is that "the customer does have an address, but we just don't know what it is yet." I put it down to FUD factors - but am willing to change my view if anyone can give a sound reason for any of this garbage.
MVVM # - I did it My Way ___________________________________________ Man, you're a god. - walterhevedeich 26/05/2011 .\\axxx (That's an 'M')
NULL is not indexed use NULL in arithmetic, string, logical and others operations is incorrect its use is the result of a incorrect designing database.