Storing Hash Codes

[Brian]

I know this is the wrong way to do it, but maybe someone can tell me the
right way to do it...

I have two different databases that I need to synchronize. The database
doesn't have keys exactly, but it does provide a rowid function.

So, I am storing a dictionary of <string, long>, string being the rowid
and long being a sum of the hash code for each column in the row.

With this dictionary, it is very easy to determine if a row is new, has
changed, or is already in sync with the target system:

* If I have a rowid that doesn't exist in the dictionary, then I know
it is a new row and I can call my insert code.

* If I have a rowid that exists in the dictionary, and the sum of hash
codes is the same as in my dictionary, then I know that the row has not
changed and there is no action to take.

* If I have a rowid that exists in the dictionary, and the sum of hash
codes is not the same as my dictionary, then I know the row of has
changed, and I need to call my update code.

-----

So, my question is, with this situation, how can I get the same results,
but not store GetHashCode(). I need some kind of "stable" hash
calculation. And it needs to take all kinds of objects; string, int,
decimal, double, etc.

I started to think I could just use a hash code on string and call
ToString() on each column, but this looks just as dangerous as trusting
GetHashCode() not to change.
Thanks for any tips...
Brian

[Jon Skeet [C# MVP]]

Brian <[e-mail] > wrote:

<snip>

So, my question is, with this situation, how can I get the same results,
but not store GetHashCode(). I need some kind of "stable" hash
calculation. And it needs to take all kinds of objects; string, int,
decimal, double, etc.

Use an MD5 hash (or something similar). You'll need to convert each
type into a stable binary representation, but then you can add each
field to the hash as you go.

--
Jon Skeet - <sk***@pobox.co m>
http://www.pobox.com/~skeet Blog: http://www.msmvps.com/jon.skeet
If replying to the group, please do not mail me too

[Dale]

Let me just add to Jon's reply. Look at the System.Security .Cryptography
classes.
--
Dale Preston
MCAD C#
MCSE, MCDBA
"Brian" wrote:

I know this is the wrong way to do it, but maybe someone can tell me the
right way to do it...

I have two different databases that I need to synchronize. The database
doesn't have keys exactly, but it does provide a rowid function.

So, I am storing a dictionary of <string, long>, string being the rowid
and long being a sum of the hash code for each column in the row.

With this dictionary, it is very easy to determine if a row is new, has
changed, or is already in sync with the target system:

* If I have a rowid that doesn't exist in the dictionary, then I know
it is a new row and I can call my insert code.

* If I have a rowid that exists in the dictionary, and the sum of hash
codes is the same as in my dictionary, then I know that the row has not
changed and there is no action to take.

* If I have a rowid that exists in the dictionary, and the sum of hash
codes is not the same as my dictionary, then I know the row of has
changed, and I need to call my update code.

-----

So, my question is, with this situation, how can I get the same results,
but not store GetHashCode(). I need some kind of "stable" hash
calculation. And it needs to take all kinds of objects; string, int,
decimal, double, etc.

I started to think I could just use a hash code on string and call
ToString() on each column, but this looks just as dangerous as trusting
GetHashCode() not to change.
Thanks for any tips...
Brian

[Dale]

And to add to my own reply, here's a sample from an application I wrote a few
years ago:

public static byte[] HashBytes512(st ring pw, bool UE)
{
byte[] pwBytes;
if (UE)
{
pwBytes = Encoding.Unicod e.GetBytes(pw);
}
else
{
pwBytes = Encoding.ASCII. GetBytes(pw);
}
SHA512Managed sha = new SHA512Managed() ;
byte[] hashBytes = sha.ComputeHash (pwBytes);
return hashBytes;
}

--
Dale Preston
MCAD C#
MCSE, MCDBA
"Dale" wrote:

Let me just add to Jon's reply. Look at the System.Security .Cryptography
classes.
--
Dale Preston
MCAD C#
MCSE, MCDBA
"Brian" wrote:

I know this is the wrong way to do it, but maybe someone can tell me the
right way to do it...

I have two different databases that I need to synchronize. The database
doesn't have keys exactly, but it does provide a rowid function.

So, I am storing a dictionary of <string, long>, string being the rowid
and long being a sum of the hash code for each column in the row.

With this dictionary, it is very easy to determine if a row is new, has
changed, or is already in sync with the target system:

* If I have a rowid that doesn't exist in the dictionary, then I know
it is a new row and I can call my insert code.

* If I have a rowid that exists in the dictionary, and the sum of hash
codes is the same as in my dictionary, then I know that the row has not
changed and there is no action to take.

* If I have a rowid that exists in the dictionary, and the sum of hash
codes is not the same as my dictionary, then I know the row of has
changed, and I need to call my update code.

-----

So, my question is, with this situation, how can I get the same results,
but not store GetHashCode(). I need some kind of "stable" hash
calculation. And it needs to take all kinds of objects; string, int,
decimal, double, etc.

I started to think I could just use a hash code on string and call
ToString() on each column, but this looks just as dangerous as trusting
GetHashCode() not to change.
Thanks for any tips...
Brian