Selection, as if it were text. - Design and Architecture Discussion Boards

Quite often, I have had to work (read fix here) with other people's database developments where they implement separate insert/update methods in the business layer.

A lot of the time, this doesn't actually make sense, and it is nicer just to provide a Save method where the save itself works out whether or not it needs to insert or update the record. As an example, consider the following Stored Procedure:

CREATE PROCEDURE [MyUser].[SaveUser]
@ID INT OUT,
@UserName NVARCHAR(30)
AS

SET NOCOUNT ON

IF @ID = 0
BEGIN
INSERT INTO UserDetails(ID, Name) VALUES (@ID, @UserName)
SET @ID = SCOPE_IDENTITY
END
ELSE
BEGIN
UPDATE UserDetails SET Name = @UserName WHERE ID = @ID
END

This means that the datalayer just needs to call the SaveUser procedure and the SP will work out whether it needs to do a save or an insert. Now, there are some arguments as to whether or not this is the responsibility of the datalayer, but look at how this simplifies procedure development. Instead of having Insert/Update/Delete, you end up having Save/Delete, thus simplifying your datalayer. The downside here is that you have to check the @ID field first, which is a (minor) performance hit that you may want to forego.

Alternatively, you could have the datalayer implement the same logic and determine whether or not this is an insert by checking the ID field beforehand.

public void Save()
{
  if (_id == 0)
  {
    // This is an insert.
  }
  else
  {
    // This is an update.
  }
}

All I am asking, is that you remove the decision as to whether to insert or update from the business layer. Please... Smile | :)

the last thing I want to see is some pasty-faced geek with skin so pale that it's almost translucent trying to bump parts with a partner - John Simmons / outlaw programmer

Deja View - the feeling that you've seen this post before.

Pete O`Hanlon wrote:
Alternatively, you could have the datalayer implement the same logic and determine whether or not this is an insert by checking the ID field beforehand.

Or Alternatively you could let the data adapter do its job and make exactly the same decision for you.

Notice that the Data Adapter has only 2 main methods:

Fill - Executes your Select
Update - Executes your Insert,Update or Delete depending on the current state of the row being saved.

Sorry to say I 100% disagree with your oppinion that proper CRUD stored procedures are not a good thing.

---

How to get answers to your questions[^]

J4amieC wrote:
Or Alternatively you could let the data adapter do its job and make exactly the same decision for you.

Notice that the Data Adapter has only 2 main methods:

Fill - Executes your Select
Update - Executes your Insert,Update or Delete depending on the current state of the row being saved.

This only works if you are using a Data Adapter. In most cases, I have been dealing with more traditional business objects that are mapped into the database. For instance, having a Customer class and a Customers collection with the Customers collection being responsible for returning a number of populated Customer classes.

J4amieC wrote:
Sorry to say I 100% disagree with your oppinion that proper CRUD stored procedures are not a good thing.

Where in my original post did I state that CRUD procedures are not a good thing? I present a couple of alternatives for saving data - one which uses a composite INSERT/UPDATE procedure, and the other which uses C# code to accomplish the same thing. What I did state was that it shouldn't be the responsibility of the business layer to work out whether to do an insert or an update. My last line was the important one:

All I am asking, is that you remove the decision as to whether to insert or update from the business layer. Please...

the last thing I want to see is some pasty-faced geek with skin so pale that it's almost translucent trying to bump parts with a partner - John Simmons / outlaw programmer

Deja View - the feeling that you've seen this post before.

Pete O`Hanlon wrote:
All I am asking, is that you remove the decision as to whether to insert or update from the business layer.

I agree and infact when I do use SPs I try and do this, makes the client code simpler and that's more likely to change so why make more work?

I have no idea what I just said. But my intentions were sincere.

Ed.Poore wrote:
when I do use SPs I try and do this, makes the client code simpler and that's more likely to change so why make more work?

And that's the best approach that you can take.

the last thing I want to see is some pasty-faced geek with skin so pale that it's almost translucent trying to bump parts with a partner - John Simmons / outlaw programmer

Deja View - the feeling that you've seen this post before.

(As is always the case) if the situation requires it, use it.

I have no idea what I just said. But my intentions were sincere.

I've worked on projects that supported this approach in the past.

I learned (the hard way) that I had to modify the business object to ensure that it had a valid ID.

So, I had to track state of the object. When loaded from the database, I set a hidden variable.

If the object was created from the database load, then an internal flag would hold that fact. If the object was not instantiated from the database, then the flag would never be set.

So maintaining state in the object allowed me to implement this technique.

However, it is much easier to have CRUD (Create, Read, Update and Delete) implemented on any single object representing a single record from the database. Which should map to the corresponding (Insert, Select, Update, Delete) procedures.

~ CodeDoctor ~

I am attempting to create a more efficient way to select data(like text). Before I used an index to the data that was stored in an array. I would have a separate collection or array that held the indexes of the data for another array. Each index in the "selection array" was selected. I think I have made a more efficient way to do that using two longs per selection, it does not matter how large the selection is, it will always be two longs. You can select multiple items. I'm using this in a application that can select data just like in notepad or a hex editor.

void SelectData(long start, long end)
{
  _selected.Add(new Range(start, end));
  if (_selected.Count > 1)
    RemoveRangeRedundancies();
}

void RemoveRangeRedundancies()
{
  foreach (Range rr in _selected)
  {
    foreach (Range r in _selected)
    {
      if (rr == r) continue;

      if (rr.MinValue >= r.MinValue && rr.MaxValue <= r.MaxValue)  //selection completely inside an already selected range of data.
      {
        _selected.Remove(rr);
        continue;
      }

      if (rr.MinValue <= r.MinValue && rr.MaxValue >= r.MaxValue) //selection bigger on both or one rr.MaxValue
      {
        _selected.Remove(r);
        continue;
      }

      //selection bigger or connected to the left (join)
      if (rr.MinValue < r.MinValue && (rr.MaxValue >= r.MinValue - 1) && rr.MaxValue <= r.MaxValue)
      {
        Range newRange = new Range(rr.MinValue, r.MaxValue);
        _selected.Remove(rr);
        _selected.Remove(r);
        _selected.Add(newRange);
        continue;
      }

      //selection bigger or connected to the right (join)
      if (rr.MinValue >= r.MinValue && (rr.MinValue <= r.MaxValue + 1) && rr.MaxValue > r.MaxValue)
      {
        Range newRange = new Range(r.MinValue, rr.MaxValue);
        _selected.Remove(rr);
        _selected.Remove(r);
        _selected.Add(newRange);
        continue;
      }
    }
  }
}

Do you think that is a good design?

█▒▒▒▒▒██▒█▒██
█▒█████▒▒▒▒▒█
█▒██████▒█▒██
█▒█████▒▒▒▒▒█
█▒▒▒▒▒██▒█▒██

Captain See Sharp wrote:
it does not matter how large the selection is, it will always be two longs

Two ints would be just fine unless you anticipate working with buffers >1.99GB.

I haven't analyzed your code in depth, but it looks sufficient provided you don't have a large number of selection ranges (and in most cases you wouldn't). But if you think it would be wise to plan for that, then a double loop is not a good idea (O(N^2) performance overhead - i.e. the code in the inner loop is executed N*N times, where N is the number of ranges).

Performance-wise, it would be best to do the following (provided that you want the selection ranges to always stay non-overlapping). When adding a new range:

Do a binary search to find the insertion location for the range.
Look at the range that is currently at the insertion location, and the one right before it. If the new change overlaps either of these, combine the new range with them.
Insert the range at the insertion location (calculated via the binary search).

This will take significantly less time, and you have the added bonus of always having your selection range collection arranged in sorted order. That is what I did with text change ranges in a textbox I wrote for a client, where text changes are stored up and the text layout/flow is updated in a batched layout pass.

Thanks for you input.

I'm building a new version of my hex editor control just to let you know.

J. Dunlap wrote:
Two ints would be just fine unless you anticipate working with buffers >1.99GB.

In the last version of my control I used ints. It was suffecient however in my new version I want it to be able to work with extremely large files. For large files I'm going to make it so it will not load the entire file so it wont use too much memory. If the program selects data that is not loaded into memory it will still need to know the index of the data so I will use longs.

A binary search would be a good idea and I may do that, How would I sort the ranges? Could I just add the MaxValue, MinValue, and Range together and sort by that sum?

█▒▒▒▒▒██▒█▒██
█▒█████▒▒▒▒▒█
█▒██████▒█▒██
█▒█████▒▒▒▒▒█
█▒▒▒▒▒██▒█▒██

Captain See Sharp wrote:
Thanks for you input.

You're welcome. Smile | :)

I'm hoping to make this forum grow and bring some people together who can share design ideas and expertise.

Captain See Sharp wrote:
I'm building a new version of my hex editor control just to let you know.

Is there going to be an article about it? I was going to write one myself but have never had the time. I wanted to write a tool that would make it easier to deconstruct binary file formats.

Captain See Sharp wrote:
It was suffecient however in my new version I want it to be able to work with extremely large files. For large files I'm going to make it so it will not load the entire file so it wont use too much memory.

Nice!

Captain See Sharp wrote:
A binary search would be a good idea and I may do that, How would I sort the ranges? Could I just add the MaxValue, MinValue, and Range together and sort by that sum?

Use the MinValue. If you add them together, then you will get indeterminate sort results and you won't be able to just check for overlap with the 2 ranges before and at the insertion location, which defeats the purpose. Implement IComparable<T> on your Range class like this:

int IComparable<Range>CompareTo(Range other)
{
   if(MinValue>other.MinValue)
      return 1;
   if(MinValue<other.MinValue)
     return -1;
   return 0;
}

Then all you have to do is insert them in the correct sorted location as you add them. (I assume you don't modify the ranges after they are created, other than to combine them when they overlap? If you do you will need to update the sort order on modification as well.) You can use binary search, as I mentioned, to find the insertion location, assuming the structure is a List<Range>:

int insertIdx=ranges.BinarySearch(newRange);
if(insertIdx<0) //if there isn't an existing item matching the MinValue in the collection already...
   insertIdx=~insertIdx; //bitwise complement of the insertion index was returned, so complement it again.
//insertIdx is now the location to insert at

Or you can do it this way (Fast custom binary search with comparison built in, and no bitwise complement needed):

int GetInsertIndex(IList<Range> ranges, long minValue)
{
    int left = 0;
    int right = (0 + ranges.Count) - 1;
    while (left <= right)
    {
        int idx = left + ((right - left) >> 1);
        long listVal = ranges[idx].MinValue;
        if (listVal == minValue)
            return idx;
        if (minValue < listVal)
            right = idx - 1;
        else
            left = idx + 1;
    }
    return left;       
}

Then here is an adapted version of my code to insert ranges:

int insertIdx=GetInsertIndex(ranges,range.MinValue);
//if the range at the insertion location overlaps the range being added...
if (insertIdx < ranges.Count && ranges[insertIdx].MinValue <= range.MaxValue)
    ranges[insertIdx] = CombineRanges(range, ranges[insertIdx]); //combine them
else
    ranges.Insert(insertIdx, range); //insert the range at the insertion location

//if the range before the insertion location overlaps...
if (insertIdx > 0 && ranges[insertIdx - 1].MaxValue >= range.MinValue)
{
    //combine the newly inserted range with the one before it
    ranges[insertIdx - 1] = CombineRanges(range, ranges[insertIdx - 1]);
    //remove the newly inserted range because it has been merged
    ranges.RemoveAt(insertIdx);
}

...

private Range CombineRanges(Range change1, Range change2)
{
    return new Range(
        Math.Min(change1.MinValue, change2.MinValue),
        Math.Max(change1.MaxValue, change2.MaxValue)
        );
}

Excellent, I see how its all going to fit together. When I get the index to insert the new Range will be it like this?

I get 2 as the place to insert and when I insert it will go between 1 and 2?

     ..<.<.<.<.<.<.<.<------ 
0  1 . 2  3
     .
     ^
  new rage
     ^
0 1  2  3  4

J. Dunlap wrote:
Is there going to be an article about it?

Yeah, I had an article written for version 1.0. Version 1.0 was slow and limited. If you wanted to do copy and paste operations or undo/redo operations you had to type a lot of code to work with the control. I made a hex editor to go along with it and I realized that it wasn't as good as I want it to be. Its still good but not good enough. It had too many loops and and arrays, it was inefficient. The very core of the control is called the Data class and it does the hard work, its the smallest class in the program but all other code is wrapped around it and its the most performance critical part. Thats what all this is going into, the core. Then I can build the public API and UI part of the control which is much easier especially since I have an extensive API (from 1.0) that only needs slight additions and modifications.

Implementing binary search and sorting seems pretty simple. Hopefully it will stay that way when I'm actually implementing it. Thanks for your help. Smile | :)

█▒▒▒▒▒██▒█▒██
█▒█████▒▒▒▒▒█
█▒██████▒█▒██
█▒█████▒▒▒▒▒█
█▒▒▒▒▒██▒█▒██

I think I got it. I just want to share the code in case you or someone else finds that its flawed or something.

UPDATE: I am currently testing it and it seems to work as planned. I think this should work for about any control that selects sequential data such as a text box or whatever you can think of.

public void SelectData(long start, long end)
{
  //if the range of data being selected does not join with any other selections
  //or there are no selections then this will remain true
  bool addNewRange = true;
  byte inc = 0;
  int insertIndex = _selected.BinarySearch(new Range(start, 0)); //search for the appropriate index to insert a
                                                                 //new selection range
  if (insertIndex < 0)  //if there is not already data selected at the start index then
      insertIndex = ~insertIndex; //figure out where the selection range should be placed

  if (_selected.Count > 0)  //if there is no selected data then checking for range redundancies is not needed
  {
    //if there is only 1 selection range and the new selection needs to be placed at index 1 then
    if (_selected.Count == insertIndex)
    {
      insertIndex--;
      inc = 1;
    }

    if (end >= _selected[insertIndex].MinValue - 1)  //check for reducdancies (overlapping)
    {
      if (start < _selected[insertIndex].MinValue)
      {
        _selected[insertIndex] = new Range(start, _selected[insertIndex].MaxValue);
        addNewRange = false;
      }

      if (end > _selected[insertIndex].MaxValue && start < _selected[insertIndex].MaxValue)
      {
        addNewRange = false;

        _selected[insertIndex] = new Range(_selected[insertIndex].MinValue, end);

        //if the new selection range is longer to the right of the current overlapping range then see how far it goes
        //and remove redundancies
        for (int i = insertIndex + 1; i < _selected.Count; i++)
        {
          if (end < _selected[i].MinValue - 1) break;

          if (end >= _selected[i].MinValue - 1 && end <= _selected[i].MaxValue)
          {
            _selected[insertIndex] = new Range(_selected[insertIndex].MinValue, _selected[i].MaxValue);
            _selected.RemoveAt(i);
            i--;
            continue;
          }
          //if end is greater than i's maxvalue then just remove it because it completely overlaps it
          if (end > _selected[i].MaxValue)
          {
            _selected.RemoveAt(i);
            i--;
          }
        }
      }
    }

    //this block checks to see if the new selection overlaps the selection to the left
    if (insertIndex > 0)
    {
      if (start <= _selected[insertIndex - 1].MaxValue + 1)
      {
        if (!addNewRange)
          end = _selected[insertIndex].MaxValue;

        addNewRange = false;
        _selected[insertIndex - 1] = new Range(_selected[insertIndex - 1].MinValue, end);
      }
    }
  }

  if(addNewRange)
    _selected.Insert(insertIndex + inc, new Range(start, end));

}

-- modified at 1:38 Thursday 4th January, 2007
-- modified at 21:34 Thursday 4th January, 2007

-- modified at 22:11 Thursday 4th January, 2007

█▒▒▒▒▒██▒█▒██
█▒█████▒▒▒▒▒█
█▒██████▒█▒██
█▒█████▒▒▒▒▒█
█▒▒▒▒▒██▒█▒██

I know.. I know... Place your ones wherever

ss

Brad
Australian

-CAUTION-
The previous statement may contain traces of PHP, and by reading this statement you negate the right to vote me down.

I go away for a bit and I come back to find that the powers-that-be have granted my request! Big Grin | :-D

Thanks Chris! I will post some things here when I get back into the swing of coding after Christmas.

Umm, excluding zero and one that is. Smile | :)

Sheesh, the math one isn't used much except by me. Smile | :)

Where's that geek icon when you need it?

"Marge, don't discourage the boy! Weasling out of things is important to learn. It's what separates us from the animals! Except the weasel." - Homer Simpson

Web - Blog - RSS - Math - LinkedIn - BM

two & three are suffering about your discrimination. Though not perfects, they think to be square too...

If the Lord God Almighty had consulted me before embarking upon the Creation, I would have recommended something simpler.
-- Alfonso the Wise, 13th Century King of Castile.

Bassam Abdul-Baki wrote:
he math one isn't used much except by me

I visit the Math/Algorithms forum a bit, but not a whole bunch going on there. I've wondered about putting math quiz questions in there to help keep people's math skills sharp...

Some people have a memory and an attention span, you should try them out one day. - Jeremy Falcon

As long as the maths questions are not too difficult. Remember not every programmer is a maths genius, and many don't have a need to be.

modified 1-Aug-19 21:02pm.

Richard A. Abbott wrote:
Remember not every programmer is a maths genius, and many don't have a need to be.

I figured maybe something of different challenge levels. I could only go as far as calculus or linear algebra. Any partial differential equations questions would get me :->

Some people have a memory and an attention span, you should try them out one day. - Jeremy Falcon

Try Sudoku or logic puzzles. Those may make for an interesting challenge. Big Grin | :-D

"Marge, don't discourage the boy! Weasling out of things is important to learn. It's what separates us from the animals! Except the weasel." - Homer Simpson

Web - Blog - RSS - Math - LinkedIn - BM

Bassam Abdul-Baki wrote:
Try Sudoku or logic puzzles.

Yep, they are a lot of fun. I got a Kakuro[^] puzzle book for Christmas and it's pretty addictive Smile | :)

Some people have a memory and an attention span, you should try them out one day. - Jeremy Falcon

Played that once. Make's Sudoku look like child's play. There's an interesting Sudoku[^] client here that you can try. Check the messages for the executable.

"Marge, don't discourage the boy! Weasling out of things is important to learn. It's what separates us from the animals! Except the weasel." - Homer Simpson

Web - Blog - RSS - Math - LinkedIn - BM

Bassam Abdul-Baki wrote:
There's an interesting Sudoku[^] client here that you can try.

I'll look at it closer later today. Looks really cool Smile | :)

Maybe you might like this http://www.essex.ac.uk/maths/dept/fun/Sudoku1.pdf[^]

modified 1-Aug-19 21:02pm.