[p4] When do labels degrade Perforce performance?

[Robert Cowham]

Depending on how you use them, the new dynamic labels is a very nice
feature.

This effectively makes a label an alias for a changelist.


C:\bruno_ws>p4 label -o l1
# A Perforce Label Specification.
Label:  l1

Revision:       @800

View:
        //depot/Jam/MAIN/...

The above label is an alias for changelist 800 in the view, thus:

	p4 files //depot/Jam/MAIN/... at 800

Is the same as:

	p4 files @l1

In addition, with a spec depot, you can versioning of the labels.
 
The only problem with the above type of label is that it doesn't show up in
the output of the command:

	p4 labels //depot/Jam/MAIN/...

Since this does a scan of db.label which has no entries - a real shame.

The static labels cause db.label file to have an entry for every file in
every label. Obviously this can get big after a while. Depending on how you
use the labels will depend on what performance issues they can cause.

Robert

> -----Original Message-----
> From: perforce-user-bounces at perforce.com 
> [mailto:perforce-user-bounces at perforce.com] On Behalf Of Brian Jones
> Sent: 20 September 2007 15:40
> To: perforce-user at perforce.com
> Subject: [p4] When do labels degrade Perforce performance?
> 
> Does anyone have any quantitative information on how and when 
> labels impact Perforce performance?
> 
> Page 129 of the 2007.2 Admin manual states:
> 
> "Repeated references to large labels can be particularly 
> costly. Commands that refer to files using labels as 
> revisions will scan the whole label once for each file argument."
> 
> I just started working with Perforce this year. I have a lot 
> of experience with other SCM tools. I've searched the forums, 
> knowledge base, net, and user docs for information on how 
> labels impact Perforce performance but have not found much 
> satisfactory information. With non-changelist based SCM 
> systems it is a common practice to label every build. With 
> Perforce though, this is a bad idea. That being said, there 
> are still cases where labels are very helpful as they give a 
> human readable name to something in the absence of changelist 
> aliases. I would really like to better understand in real 
> terms how labels impact performance and where the performance 
> limits are in their use.
> 
> Brian Jones
> _______________________________________________
> perforce-user mailing list  -  perforce-user at perforce.com 
> http://maillist.perforce.com/mailman/listinfo/perforce-user
>