|
Nagy Vilmos wrote: A customer is a person or a company, A company also has a person as a contact.
What about putting the Persons class as a nested class in the Company class in that case?
Complicated?
|
|
|
|
|
Not a good idea.
The attributes of a person - name, date of birth, sex, etc - are independent of the role they perform. A decent representation of a person will be reusable if you ignore how it is used.
Similarly, the concept of a company is independent of the problem domain. A company has employees but, and this is true in real life as much as in a program, the employee does not have a company.
Panic, Chaos, Destruction. My work here is done.
Drink. Get drunk. Fall over - P O'H
OK, I will win to day or my name isn't Ethel Crudacre! - DD Ethel Crudacre
I cannot live by bread alone. Bacon and ketchup are needed as well. - Trollslayer
Have a bit more patience with newbies. Of course some of them act dumb - they're often *students*, for heaven's sake - Terry Pratchett
|
|
|
|
|
I still missed the big point but if you do get around to writing the article, It would be interesting. I guess your talking about some kind of interface that both person and customer implement.
Anyway I've just spent the last month discovering values such as the "delivery contact name" being stored in the weather_sunny field.  So I'm a bit paranoid about such things at the moment.
"You get that on the big jobs."
|
|
|
|
|
This is a long story which I will try to make as short as possible, thanks for your patience in advance 
I am working on a program that has evolved from a web application to a Windows Forms client, then was ported to a WPF client and now has become a XNA client with a homemade UI. Porting the application from Windows Forms to WPF and now or own UI was possible because it was based on a (modified) MVP pattern. It was just a matter of porting the MVP framework to the new UI and then redesigning the views. The presenters or other logic remained the same.
You may have noted the (modified) above. This modification deals with the application's workflow and and controlling the application's workspace. The main workspace is nothing else than an empty window. The application can then subdivide it as needed and assign the subworkspaces eventually to presenters which display views in them. Seen from the perspective of the UI, workspaces are windows(!), panels and all kinds of layout controls.
Now, while working on my own UI, I have stumbled over a question which I already swept under the rug in the initial Windows Forms implementation and kept under the rug in the WPF implementation. Now I cannot put it off any longer, but I also sit between two chairs.
My UI has a hierarchy of controls, very similar to what you are accustomed to from Windows Forms or WPF. With our MVP being an inegral part of the UI, windows (or forms, if you like) are in the 'workspace' branch of the class hierarchy. Traditionally UIs don't have any concept of workspaces and treat windows just like any other control. This leaves me with those choices:
1) Windows are workspaces. To stay consistent with the established architecture, the correct procedure will be to open an empty window and assign it to a presenter as workspace. The presenter will then show its view inside this window. The option to open a simple dialog on the fly, like a MessageBox, will be lost.
2) Windows are controls. This will stand in opposition to the application's main workspace, which must be a window. Also, this would be absolutely against the existing architechture. Dialogs would always have to be opened like controls and would be outside the MVP.
Of course I tend to 1), but I don't like departing from the accustomed way of opening a window or dialog. Seeing a MessageBox (or any other 'common dialog') as yet another control of the view is practical at times. Simply implementing both versions looks to me like bad class design and would also introduce significant redundancy. The classes would be a few ancestors remote and with single inheritance this could not be resolved without redundancy. The redundancy may even get worse if I also take modal windows and modeless windows into account.
So what do you say? Have I overlooked something? Which path should I follow?
And from the clouds a mighty voice spoke:
"Smile and be happy, for it could come worse!"
And I smiled and was happy
And it came worse.
|
|
|
|
|
This solution to a real world problem just seems too easy to be foolproof, so I thought I would see if anyone can show a problem with it. Say I have the same item in two sizes, one cost $2.00, and the other $2.25. The summary tables used for reporting have only fields for the total quantity sold: 137, and the total cost: $286.50. The customer would like a monthly report/export showing quantity by price. The first approach was pulling the inforation directly from the transaction tables which contain millions of records. Though this approach provided the seperation, the performance hit made me search for a better solution. It made sense to me that only one correct combination of (x * 2.00) + (y * 2.25) could equal 286.50. So, how to get the correct combination? Examples in vbscript for simplicity/proof of concept.
totalQuantity = 137
wantedResult = 286.5
premiumPrice = 2.25
standardPrice = 2.00
for x = totalQuantity to 0 step -1
y = totalQuantity - x
if (x * premiumPrice) + (y * standardPrice) = wantedResult then exit for
next
This approach, even it the raw form, is much, much quicker than querying the transaction tables. The following revision finds the correct answer in 6 iterations. When I tested it for an extreme case of 1 million total (999,999 of standard price and 1 premium) it found the answer in 25 guesses.
totalQuantity = 137
wantedResult = 286.5
premiumPrice = 2.25
standardPrice = 2.00
success = 1
currentSplit = Int(totalQuantity / 2)
a = currentSplit
b = totalQuantity - a
counter = 0
currentresult = (a * premiumPrice) + (b * standardPrice)
counter = counter + 1
Do While currentresult <> wantedResult
if currentSplit > 1 then
currentSplit = Int(currentSplit / 2)
end if
if counter > totalQuantity then
msgbox "result not found"
success = -1
exit do
end if
If currentresult < wantedResult Then
a = a + currentSplit
Else
a = a - currentSplit
End If
counter = counter + 1
b = totalQuantity - a
currentresult = (a * premiumPrice) + (b * standardPrice)
Loop
if success = 1 then
MsgBox a & " at " & premiumPrice & " and " & b & " at " & standardPrice & " in " & counter & " guesses"
end if
The point is, this function may keep me (or the customer) from having to ask the database vendor for a change to their summary tables. I am not asking for code testing, only opinions on the soundness of the concept. As far as I know, there will only ever be two price points for this single item and the highest total quantity would likely be around 5000 for a reporting range.
"Go forth into the source" - Neal Morse
|
|
|
|
|
Why not just use a system of linear equations?
"The clue train passed his station without stopping." - John Simmons / outlaw programmer
"Real programmers just throw a bunch of 1s and 0s at the computer to see what sticks" - Pete O'Hanlon
"Not only do you continue to babble nonsense, you can't even correctly remember the nonsense you babbled just minutes ago." - Rob Graham
|
|
|
|
|
The algebra is a bit rusty...I gave up on an elegant one equation solution, and found it easier to go for a brute force attack. Thanks to your suggestion, I tried to form the equation again, realizing that it should be formed with just one variable instead of two and ended up with:
2a + 2.25(137 - a) = 286.5 resulting in:
2a + (308.25 - 2.25a) = 286.5 resulting in:
2a - 2.25a = -21.75 resulting in:
-.25a = -21.75 resulting in:
a = 87 which is correct!
Now, getting this into code:
totalQuantity = 137
totalCost = 286.50
a = 0
b = 0
standardPrice = 2.00
premiumPrice = 2.25
a = (totalCost - (premiumPrice * totalQuantity)) / (standardPrice - premiumPrice)
b = totalQuantity - a
It's a bit alarming how quickly I had given up on getting the solution correctly in favor of something way too complicated!...and looking at it now, I realize this is probably a 5th grade math problem! (hanging head)
"Go forth into the source" - Neal Morse
|
|
|
|
|
Excellent! That is the same answer I found as well.
kmoorevs wrote: I realize this is probably a 5th grade math problem!
In this day and age, in the fifth grade they do begin to introduce this type of problem. My kids were faced with the same type of problem just before junior high school. At the pace they are trying to teach kids math now, it is not of surprise seeing high school sophomores or juniors in calculus courses.
"The clue train passed his station without stopping." - John Simmons / outlaw programmer
"Real programmers just throw a bunch of 1s and 0s at the computer to see what sticks" - Pete O'Hanlon
"Not only do you continue to babble nonsense, you can't even correctly remember the nonsense you babbled just minutes ago." - Rob Graham
modified 18-Oct-11 11:48am.
|
|
|
|
|
Paul Conrad wrote: At the pace they are trying to teach kids math now, it is not of surprise seeing high school sophomores or juniors in calculus courses.
It's just too bad they don't make the kids learn what they're teaching. Looking at why my younger siblings are learning, the schools are making kids more reliant on calculators and computers and less capable of doing the problems by hand.
|
|
|
|
|
Bert Mitton wrote: the schools are making kids more reliant on calculators and computers and less capable of doing the problems by hand
That is sad. When I did teach high school math a while back, I tried to get the students to be less reliant on calculators, and it did work with a lot of the kids to regain confidence that they could solve the problems without a calculator.
In college, my Physics professor did not allow calculators at all. He wanted to see how we derived our solution, and could have cared less about punching a bunch of buttons on a calculator to come up with an answer.
"The clue train passed his station without stopping." - John Simmons / outlaw programmer
"Real programmers just throw a bunch of 1s and 0s at the computer to see what sticks" - Pete O'Hanlon
"Not only do you continue to babble nonsense, you can't even correctly remember the nonsense you babbled just minutes ago." - Rob Graham
|
|
|
|
|
kmoorevs wrote: only opinions on the soundness of the concept.
I seriously doubt that is going to work.
Price1: 2.00
Price2: 2.25
Total price: 18.00
Solution1: 9 items at price1
Solution2: 8 items at price2
Scenario 2
Total price: 36.00
Solution1: 18 items at price1
Solution2: 16 items at price2
Solution3: 9 items at price1 and 8 items at price2
kmoorevs wrote: The first approach was pulling the inforation directly from the transaction
tables which contain millions of records.
No indexes?
What about creating a daily summary table locally. Your app pulls daily in the background. You use those local data stores to build your monthly report.
|
|
|
|
|
the total number of items is also known, so there are two linear equations and two unknown values, yielding as it should zero, one or an infinite number of solutions depending on the values.
|
|
|
|
|
Not two unknown values...that's what confused me at first until I realized that the second unkonwn can be expressed as a relationship to the first unknown, in this case, (137 - a), thus giving the equation:
(2.00 * a) + 2.25 * (137 - a) = 386.5
this works out to a = 87, so b = 50.
the equation in code from an earlier post is this:
totalQuantity = 137
totalCost = 286.50
a = 0
b = 0
standardPrice = 2.00
premiumPrice = 2.25
a = (totalCost - (premiumPrice * totalQuantity)) / (standardPrice - premiumPrice)
b = totalQuantity - a
This is really starting to remind of aa algebra problem with apples, oranges, and farmer Brown!
"Go forth into the source" - Neal Morse
|
|
|
|
|
there really is no need for you to explain me an elementary math problem, you are the one who had a problem with two unknown values (count1 and count2) and two linear equations (totalCount and totalPrice), not me. And from your reply, it is clear to me you now know how to get to a solution without really understanding the general methodology.
Maybe you should now try and figure out the counts of tree types of objects (apples, oranges, pears) with known prices (1,2,3) and weights (2,3,5), when the total item count (247), total price (472) and total weight (801) are known.
|
|
|
|
|
Sorry Luc, I certainly did not mean to offend. I admit, reposting the solution was a bit overboard. While getting the solution through whatever means was my original intent, getting a solution based on pure math was my goal, understanding the methodology would be a bonus. It's not often that I have to solve these kinds of problems, but it has been refreshing.
I could not resist the challenge, though I must admit it through brute force. I really would like to see the correct mathematical approach to this.
Apples: 104 : Oranges: 61 : Pears: 82
"Go forth into the source" - Neal Morse
|
|
|
|
|
No problem.
And your result is correct; you should have done it by hand though in order to learn something.
There basically are two approaches (although they do in essence the same thing):
1. Manually, write down the three (or N) equations. Then replace one by a linear combination of (some of) the others in such a way that one variable disappears; if you do that for all but one of the original equations, always getting rid of the same variable, you end up with n-1 equations in n-1 unknown values (and one of the original equations); you solve that in the same way, so iteratively you get at a single equation, a single unknown.
2. Theoretically, and progrsammatically, you would formulate the set of N equations using a N*N matrix (in the example the coefficients would be 1,1,1, 1,2,3, 2,3,5) and a size-N array with the known values. Then solve that with one of the known algorithms, Gauss-Seidel is one way, LU-decomposition another (it basically does what the manual approach does, with the added benefit that the intermediate steps get recorded in the L- and U-triangle you obtain. Using those, you could quickly solve any similar problem having the same matrix coefficients but different known values.
|
|
|
|
|
As we all know passwords0 should be encrypted (using an approriate one way algorithm) before they are stored anywhere and when we need to validate a second value against it we should use the same algo to encypt the second value and compare that.
My question is where in the architectural model do you believe is the best place to perform this encyption?
a) UI Tier
b) Business Tier
c) Data Tier
0This is applicable to any sensitive data
|
|
|
|
|
I have a custom password text box which hashes the password as soon as the user has entered it right in the UI layer, and I pass the hashed password through the application. This ensures that no one can see the original password at all.
When I was a coder, we worked on algorithms. Today, we memorize APIs for countless libraries — those libraries have the algorithms - Eric Allman
|
|
|
|
|
This is good - for a website. But what if I have a client that connects over a webservice? Then the control and the entire code for hashing or encrypting are in the client's assembly and can be examined by a potential attacker.
And from the clouds a mighty voice spoke:
"Smile and be happy, for it could come worse!"
And I smiled and was happy
And it came worse.
|
|
|
|
|
I don't do websites. I mainly do WPF Desktop Apps , and the hashing is done inside a custom MarkUpExtension. My apps do connect to the DataBase via WCF services, but the client apps are obfuscated and compiled assemblies. Yes, they could probably be easily reverse engineered, and they could perhaps find the hashing algorithms used, but still I don't think it possible to actually reverse any particular password that has been hashed, and this way there is no where in the assembly where the password is not hashed. Personally I don't see any other way of doing it, that is safer. If you have any better ideas, I will love to hear them.
When I was a coder, we worked on algorithms. Today, we memorize APIs for countless libraries — those libraries have the algorithms - Eric Allman
|
|
|
|
|
I also prefer clients, that's why I also had to give this some thought. Or, if it has to be a website, I still do the administrative part with a client. But you are right, there is no better solution and securing the webservice is much more important. I would suggest using https and, if you are really paranoid, additional encryption with the help of SOAP extensions.
And from the clouds a mighty voice spoke:
"Smile and be happy, for it could come worse!"
And I smiled and was happy
And it came worse.
|
|
|
|
|
CDP1802 wrote: and securing the webservice is much more important
You hit the nail on the head there.
When I was a coder, we worked on algorithms. Today, we memorize APIs for countless libraries — those libraries have the algorithms - Eric Allman
|
|
|
|
|
If you look at my question at the top of the page, it is about my private little monster. It spreads over two clients and four webservices and now even has its own XNA UI
And from the clouds a mighty voice spoke:
"Smile and be happy, for it could come worse!"
And I smiled and was happy
And it came worse.
|
|
|
|
|
CDP1802 wrote: monster
Indeed 
When I was a coder, we worked on algorithms. Today, we memorize APIs for countless libraries — those libraries have the algorithms - Eric Allman
|
|
|
|
|
It's my version of a model train
The goal is to see how far I can go when building it and where my limits are. The fun stuff which my bosses would never allow me to do.
And from the clouds a mighty voice spoke:
"Smile and be happy, for it could come worse!"
And I smiled and was happy
And it came worse.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
