Re: [Nss-mysql-devel] getent

[Guillaume Morin]

Dans un message du 13 May à 19:33, Jelmer Vernooij écrivait :
> It seems that first, a list with users is retrieved using 
> SELECT user.uid FROM user (or alike) and then a query to get the data 
> of a single user /for each user/. I have about 2000 people in my MySQL
> table and one can imagine that this slows programs like 'finger' down 
> noticably. 
> 
> Are there any reasons to do SELECT user.uid FROM user and then per
> user queries? In my nss_mysql module, I just do 
> 'SELECT user.uid,user.username,user.gecos,user.shell,etc FROM user',
> which seems to work pretty fine (and faster).

That is mainly for historical reasons and to reuse the fill_struct
functions.

You may have notice that we proceed differently with groups, we cache
all data in setgrent. This has been implemented this way because a lot
of apps do frequent getgrents. Of course, it is slow for apps which only
does one.

Implementing what you suggest is a good idea. But that implies a
redesign, fill_struct funtiocs will have to be split to a function which
does the query and one that actually fills the struct.

Thanks for your remarks. Regards,

-- 
Guillaume Morin <address@hidden>

      A friend in need is a friend indeed. A friend who bleeds is better.
         My friend confessed, she passed the test. We will never sever.
                                   (Placebo)