SQLite insert techique for unique key

[kylomas]

SQLite Experts,

What is a common table insert practice, test for the record and insert if it does not exist or insert and test for duplicate return code?

This seems like a "it depends..." kind of situation.

Searches on Google show it done both ways.

kylomas

[MvGulik]

Don't forget the SQlite help itself.

http://www.sqlite.org/lang.html

Seems that REPLACE might be the one your looking for.

(also take a look at ON CONFLICT)

[jchd]

It all depends on your schema and its underlying rules. Logical use of either a (hidden) rowid or an integer primary key can make a huge difference, whether you handle the id yourself or leave SQLite create one for you.

Never forget that INSERT OR REPLACE will first delete one or more existing row(s) then insert one anew. If you have FK on the deleted row(s) it may not be what you need!

AFAICT there is no definitive "rule", every situation calls for its own logic.

[kylomas]

Thank you, gentelmen.

@iEvKI3gv9Wrkd41u - I have read that doc and others and use it for reference. I have a local copy of it. I wish that it had examplem, however. Do you mind if I ask you how you came up with that screen name and what it means?

@jhcd - Yes, I thought a much. Unfortunately I do not yet understand enough about DB management to make any intelligent decisions regarding DB organization. I am planning the schema (is that the word that describes the DB organization?) now. It seems like a trivial event, but I am tickled pink that I actually have a DB loaded. I read much of what you posted and learned enough to get something going. I have a mountain of questions but this forum is hardly the place for DBA type discussions. I'll keep digging till I get it.

I used the transaction batching technique that you recently advised on another post. The difference on my computer is at least 300% (probably due to minimized I/O activity).

kylomas

[MvGulik]

Do you mind if I ask you how you came up with that screen name and what it means?

I did not and it means nothing. Its just a random generated (password) string. (Its my actual login name on this forum)

... kinda boring, but thats the way it is ...

Edited March 14, 2011 by iEvKI3gv9Wrkd41u

[jchd]

You'll certainly benefit from freely digging into the SQLite mailing list archive which is the premium mine of information both theoritical and practical.

[Xenobiologist]

Hi,

normally, you do know whether the key already exists. (Programmatical logic)

Guess what: If you don't know then there are several ways to get it done.

1 ) Check for key (via SQL) e.g. by return code

a ) if it exsis, then update

b ) if not, then insert

2 ) Firstly make a delete then an insert

3 ) Using a special command (depends on the used database)

E.g. in DB2 >= V9 you can try using Merge.

You can also create your own function which provides the logic/code.

Mega

Edited March 14, 2011 by Xenobiologist

[jchd]

Yes, that or much more simply, declare a ON CONFLICT IGNORE on the unique column(s) and forget about applicative code complexity.

[Zedna]

Just execute

Select Count(*) from table where id = value

before your INSERT statement

This way you can see if it already exists in the table.