Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming > #5989

Should I exclude non-changed fields from update in DB?

Show all headers | View raw

I am writing a lightweight ORM for SQL in Perl.

I am now developing a kind of an active record.

This active record uses lazy-loading to facilitate performance.

As a consequence from lazy-loading, at update() statement it may be 
sometimes known or sometimes unknown the old value of a field which is going 
to be updated.

My question: If the old value of a field is known and moreover the update() 
is going to change it's value to the same as the old one (that is not going 
to change it at all), then should I exclude this field from UPDATE SQL 
statement?

I have the following thoughts on this:

To exclude it from UPDATE statement may improve performance of the server a 
little (I assume Perl comparison of two variables is faster than writing to 
the server.)

To exclude it from UPDATE statement my in principle break relational 
validity in the case of a race condition. (Consider toy example when a table 
t has INT fields a and b and it always must be b=a+1. In this case when we 
run `UPDATE t SET a=?` instead of `UPDATE t SET a=?, b=a+1` under assumption 
that b is already one above a, this in principle could break relational 
validity.) Note that our company uses MyISAM, not a transactional DB. (I 
know that this is wrong.)

I have no other thoughts on this topic. So I am interested in your advice.

-- 
Victor Porton - http://portonvictor.org

Back to comp.programming | Previous | Next — Next in thread | Find similar | Unroll thread

Thread

Should I exclude non-changed fields from update in DB? Victor Porton <porton@narod.ru> - 2015-11-26 23:48 +0200
  Re: Should I exclude non-changed fields from update in DB? Richard Heathfield <rjh@cpax.org.uk> - 2015-11-26 22:13 +0000

csiph-web