Threads Best Practice

[Artificial_Intelligence]

Hi, all delphi fans , me again.

Which is best when working with threads?
1: put all the code in the body of the thread
2:have the thread make function calls on itself MyThead->some_function

Option 2 presumes that we have sub classed the TThread class

As it stands functions are still the best way to illiminate code complexity , but do we want to do that?

[Blunty Killer]

It boils down to the purpose of the thread that will determine what method is best used. The purpose of the thread would decide whether or not to call its code in the execute method or in procedures of the class.

For example, a thread that is created every 5 minutes to check file I/O status can be achieved by just executing that code in the Execute method of the thread. Now, if you're interacting with the main thread and need to update visual controls, you will have to create a descendant and use the Synchronize method which will require a procedure of the class.

The purpose of the thread dictates how you go about designing it.

[Artificial_Intelligence]

My thoughts were more along the lines of performance/accuracy when working with databases and having to execute multiple queries against one of more data-stores. at present i am working on a archive system (don't ask why)
that has to check for the existence of records coming from the src and going to the destination. If there are records in the destination coming in i would have to log the records to another database and then delete from where they are now then import sounds..mixed i know...

this has me wondering about how the thread will person if it has to make too many procedure/function calls or should it simple write the entire thing in the execute statment

[Blunty Killer]

Does any of the databases in question support triggers?

[Artificial_Intelligence]

Sure , the servers are MSSQL2000. What did you have in mind?

[Blunty Killer]

The checking of existing records and saving them before overwriting is a candidate for an audit trails setup. Setup triggers to save that data to either shadow tables or another database dump with a timestamp, process id and possibly a user id; then overwrite that data. This would move that function to the database and save you the headache of having to code threads for it.

Databases are very efficient at auditing themselves. Now as far as batch updates are concerned, those can be done right in the Execute method of the thread, so as long as you are not using a dataset connected to a widget.

[Artificial_Intelligence]

The checking of existing records and saving them before overwriting is a candidate for an audit trails setup

Can you provide a sample of the above or explain it some more...how will i trigger the the (trigger) to make the copy of a record that i choose?

[Blunty Killer]

Here is a trigger function (PostgreSQL) that records changes made to one of the tables used in an app I'm developing. It monitors changes in benefits to employees.

CREATE or REPLACE FUNCTION "audit"."audit_public_allowances_func"()
RETURNS "pg_catalog"."trigger" AS 
$BODY$
BEGIN
				INSERT INTO audit.public_allowances_history
					SELECT NOW(), TG_OP, OLD.*;
				RETURN NULL;
			END;
$BODY$
LANGUAGE 'plpgsql' VOLATILE;

Whatever change is made to this table is first recorded in its audit counterpart along the with operation (INSERT, UPDATE, DELETE) and the value being changed and the time changed. That way I have a complete history of all values that was ever recorded in my table and when those values were changed.

[Artificial_Intelligence]

Ho hell my brain must be frozen. off course triggers...............lol......sigh. thanks man.......i forgot about those..........will do ...........

[Blunty Killer]

Good, so now you can use your thread to populate the tables. Let databases monitor themselves.