Using Row Hashing for Incremental Integrations: Difference between revisions
Jkuehlthau (talk | contribs) |
Jkuehlthau (talk | contribs) (→C#) |
||
(One intermediate revision by the same user not shown) | |||
Line 24: | Line 24: | ||
I had an integration that pulled data from a single table into Accounts, Contacts, and Opportunities. Each of these target tables used different fields from the source. I didn't want to process every record as it changed, just the ones I cared about. So, I created VBScript before and after save procedures to save the last rowhash and compare that to the new rowhash. If they are different, then I process the record. | I had an integration that pulled data from a single table into Accounts, Contacts, and Opportunities. Each of these target tables used different fields from the source. I didn't want to process every record as it changed, just the ones I cared about. So, I created VBScript before and after save procedures to save the last rowhash and compare that to the new rowhash. If they are different, then I process the record. | ||
==VB Script== | |||
'''Before Save'''<br> | '''Before Save'''<br> | ||
Repeat Each Row | Repeat Each Row | ||
Line 82: | Line 83: | ||
End Sub | End Sub | ||
</source> | </source> | ||
Latest revision as of 19:49, 13 July 2018
Hashing The Whole Record
Typically when setting up an ongoing integration, we only want to pull data from your origin which has changed since the last time Starfish ran. This is accomplished using LastRunDate and applying it as a filter against your origin query. See Retrieve Records Modified After Last Run DateTime. This reduces the set of data returned and greatly speeds up syncronizations.
However if your source system does not contain a "Last Modified Date" field this becomes tricky. One option we have in Starfish ETL is store and compare a hash of the record. This still involves reading all data from the origin, but only passing records to the destination that have been changed. Use the following steps to set this up.
1. Enable UseDatabaseXref - See Engine Web.Config Settings.
2. Initialize an Xref list for both read and write, in our example "contact-hash". Ensure to check the append checkbox on the write side. See Cross-reference (Xref).
3. Perform the comparison. Create a new Exec Before VBScript Procedure, call it "Compare Hash". This code will compare the origin's row hash to that stored in the xref. If it matches, the row is skipped (the data hasn't changed). If it does not match the record has been changed. If there is no match, this is a new record. In this case we store the new hash and allow the job execution to continue - you should set up an Update/Insert stage to process the record. In this example, ContactID is the primary key field from your origin table.
Sub VBScriptProcedure If "@@ORG:#RowHashCode@@" = XrefRead("contact-hash","@@ORG:ContactID@@") Then Starfish.GotoNextRow Else XrefWrite "contact-hash", "@@ORG:ContactID@@", "@@ORG:#RowHashCode@@" End If End Sub
Please note the first time you run the job, all records will be considered new because no hashes will exist in the Xref. Subsequent runs will then use the stored hashes.
Hashing Partial Records
I had an integration that pulled data from a single table into Accounts, Contacts, and Opportunities. Each of these target tables used different fields from the source. I didn't want to process every record as it changed, just the ones I cared about. So, I created VBScript before and after save procedures to save the last rowhash and compare that to the new rowhash. If they are different, then I process the record.
VB Script
Before Save
Repeat Each Row
Sub VBScriptProcedure Dim UTF8 Set UTF8 = CreateObject("System.Text.UTF8Encoding") stringToUTFBytes = UTF8.GetBytes_4("FIELDS TO BE TRACKED") Dim MD5 set MD5 = CreateObject("System.Security.Cryptography.MD5CryptoServiceProvider") MD5.Initialize() 'Note you MUST use computehash_2 to get the correct version of this method, and the bytes MUST be double wrapped in brackets to ensure they get passed in correctly. md5hashBytes = MD5.ComputeHash_2( (stringToUTFBytes) ) Dim BytesToBase64 With CreateObject("MSXML2.DomDocument").CreateElement("b64") .dataType = "bin.base64" .nodeTypedValue = md5hashBytes BytesToBase64 = .Text End With 'LogMessage BytesToBase64 Dim oldHash oldHash = XrefRead("SalesQuoteToAccountHash","@@ORG:quote_id@@") 'LogMessage oldHash If oldHash = BytesToBase64 Then GoToNextRow End If End Sub
After Operation
Run Each Row
Sub VBScriptProcedure If Not PreviewMode Then Dim UTF8 Set UTF8 = CreateObject("System.Text.UTF8Encoding") stringToUTFBytes = UTF8.GetBytes_4("@@ORG:quote_id@@@@ORG:customer@@@@ORG:billing_address@@") Dim MD5 set MD5 = CreateObject("System.Security.Cryptography.MD5CryptoServiceProvider") MD5.Initialize() 'Note you MUST use computehash_2 to get the correct version of this method, and the bytes MUST be double wrapped in brackets to ensure they get passed in correctly. md5hashBytes = MD5.ComputeHash_2( (stringToUTFBytes) ) Dim BytesToBase64 With CreateObject("MSXML2.DomDocument").CreateElement("b64") .dataType = "bin.base64" .nodeTypedValue = md5hashBytes BytesToBase64 = .Text End With 'LogMessage BytesToBase64 XrefWrite "SalesQuoteToAccountHash","@@ORG:quote_id@@",BytesToBase64 End If End Sub