The Quest for Performance: deduplicate EMF, IQBase and Rete input

eclipse-viatra / org.eclipse.viatra

Main components of the VIATRA framework

https://eclipse.dev/viatra

Eclipse Public License 2.0

0 stars 1 forks source link

The Quest for Performance: deduplicate EMF, IQBase and Rete input #4

Open eclipse-viatra-bot opened 6 months ago

eclipse-viatra-bot commented 6 months ago

| --- | --- | | Bugzilla Link | 398762 | | Status | NEW | | Importance | P3 normal | | Reported | Jan 22, 2013 10:51 EDT | | Modified | Nov 03, 2017 06:16 EDT | | Version | oldinquery | | Depends on | 526801, 521206 | | Reporter | Gabor Bergmann |

Description

Cloned from: 325: The Quest for Performance: deduplicate EMF, IQBase and Rete input\ http://github.com/ujhelyiz/EMF-IncQuery/issues/issue/325

This is a continuation of 398761.

Rete input nodes and their indexers should be thin wrappers that directly query IQBase
(low prio) IQBase should perhaps consider avoiding the storage of certain EMF data, such as indexing relations that have eOpposites for reverse navigation

Remark: implementing both options at once would conflict with https://bugs.eclipse.org/bugs/show_bug.cgi?id=398721

eclipse-viatra-bot commented 6 months ago

By Gabor Bergmann on Mar 27, 2013 15:35

(slightly related: #404488 )

eclipse-viatra-bot commented 6 months ago

By Istvan Rath on May 20, 2013 04:35

I'm setting version to "unspecified" as we currently do not (firmly) know when these future ideas will be addressed.

eclipse-viatra-bot commented 6 months ago

By Istvan Rath on Apr 24, 2014 04:13

I'm expecting this to be addressed for 0.9.

eclipse-viatra-bot commented 6 months ago

By Gabor Bergmann on Jun 02, 2015 10:15

A few leftover performance-related TODOs from bug 456815 fit here nicely:

NavigationHelper is currently grossly inefficient for some of the new query runtime context functionality – some indexes should be changed
ExternalInputNode/ExternalFilterNode should offer specialized indexers directly backed by the context – if possible
We should consider introducing control for when Rete update propagation should happen (e.g. at the end of the coalescing period during updates?), as opposed to simply sending the update tuples to the input nodes

eclipse-viatra-bot commented 6 months ago

By Gabor Bergmann on Feb 01, 2016 10:22

A further idea:

In non-wildcard mode, when the first root(s) are added to the freshly created IQ base index, it should delay actually traversing the model until the backends have had their chance to add the first observed types, so that the first actual model traversal is already valuable.

eclipse-viatra-bot commented 6 months ago

By Abel Hegedus on Apr 19, 2016 05:24

Updating to correct milestone.