Replace elgg_view vars parameter with object + lazy views rendering (Trac #4922)

[elgg-gitbot]

Original ticket http://trac.elgg.org/ticket/4922 on 42891195-03-22 by trac user srokap, assigned to unknown.

Elgg version: 1.8

I was wondering about possibilities of optimizing elgg view system and I had few ideas such as

• replace copied for each vars array with object (implementing ArrayObject - we need compatibility here) to try to save amount of used memory and processing
• approach above allows to easily remove config, user, url and similar parameters from vars to make it much lighter (we can be still backward compatible)
• lazy and sequential output rendering (to allow streaming output) and render views only when used
• possibly some smart view-level caching policy based on parameters changes

I was experimenting slightly with possibility of using such object, and only problem I have now is situation when ElggViewVars object is used as parameter of array_merge function or similar. It works in case of casting values $vars = array_merge($default, (array)$vars). Ideal would be to migrate usage to sth $vars->mergeAsDefault($defaults);. I'll post my experimental branch in a minute.

[elgg-gitbot]

trac user srokap wrote on 42891220-06-26

Experimental working version #439

[elgg-gitbot]

trac user srokap wrote on 42891225-04-24

I'm not sure, but probably lazy loading of views could be done by returning object from elgg_view that has proper __toString() magic method that does actual rendering. Anybody has some experience with that method?

[elgg-gitbot]

trac user srokap wrote on 42891265-10-18

Correction, lazy views loading seems to work fine. Proof of concept included in PR.

[jdalsem]

i'm still interested in improving the views system especially if there are ideas about some fancy caching as views are probably becoming the biggest burden on the system, so if @Srokap still has some ideas lying around, maybe this ticket can be revived...

[ewinslow]

@jdalsem See my PR #7070 for my attempt at refreshing the views system implementation.

Feedback on how this might work with the improvements you want to make would be valuable.

[jdalsem]

closing this issue as there is no progress regarding this experimental approach