[Qgis-developer] Breaking API and merge of threading branch
lists at linfiniti.com
Wed Aug 10 06:32:32 EDT 2011
On Wed, Aug 10, 2011 at 11:58 AM, Martin Dobias <wonder.sk at gmail.com> wrote:
> Hi there
> watching the recent developments in the master branch I see various
> nice improvements. And they have something in common: they do _not_
> break API compatibility :-)
> I would like to bring up this topic, because we should purge a lot of
> cruft and deprecated APIs on the road to 2.0. The reason why I start
> this is that I have been closely looking at the merge of threading
> branch to master. The changes in that branch are essentially of three
> 1. asynchronous map rendering (i.e. ability to browse the map while it
> is still being rendered)
> 2. thread-safe API for vector access
> 3. various speed optimizations
> Asynchronous map rendering is relatively simple to merge and requires
> only few API changes. But without new API for access to layers it is
> possible to produce crashes (e.g. map is still being rendered when you
> decide to start editing or do some analysis). The new API for vector
> access (using iterators) will require removal of the old API
> (select(), nextFeature()) in order to work well. And that means that
> virtually any functionality that uses layers' data has to be updated
> to new API. We need to break things in order to make threaded
> rendering work.
> The question is how to proceed and how to communicate that to users.
> The procedure could look like this:
> - set a date when we will start breaking API compatibility
> - create a tag in git that would be the last revision of master branch
> compatible with 1.x API
> - change plugin manager so that it allows only plugins that say qgis
> 2.0 is the minimal version for them to run
> - change plugin installer so that it works only with new plugin
> repository and shows plugins for qgis 2.0
> Is there anything else we should consider? The point is to make the
> transition as smooth as possible for those using qgis-master.
> And what are your opinions when should we start with the breakage...
> in a month? in a week? tomorrow? now? :-)
I think the fact that we are going to break API in 2.0 is well known
and discussed often in releases, blogs, qgis hackfests etc. and is a
general expectation when performing a major version numbering change.
I think your approach is good (tag last api compatibility etc). As far
as I am concerned we can go ahead with API breakage today. There are a
number of other gut wrenching changes I would like to make for 2.0:
- change theme to 'gis' theme by default
- get rid of kids theme
- rename default theme to 'legacy'
- remove all deprecated api calls
Many parts of our api have developed inconsistencies and I would like
to spend some time putting my 'rpl' binary to good use trying to clean
these up before we release. I think we should continue the practice of
marking new api additions / changes with a @note added in 2.0 or
@note replaces getLayerId in 2.0 etc. as far as possible so that
people can adapt to our changes easily.
Anyway +1 for API breakage from me - we will continue maintaining
1.7.x for now anyway for those looking for stability. The number of
things we can reasonable backport will shrink as we mess with the API
but I don't see the impact to be too big.
> Qgis-developer mailing list
> Qgis-developer at lists.osgeo.org
Tim Sutton - QGIS Project Steering Committee Member (Release Manager)
Please do not email me off-list with technical
support questions. Using the lists will gain
more exposure for your issues and the knowledge
surrounding your issue will be shared with all.
Visit http://linfiniti.com to find out about:
* QGIS programming and support services
* Mapserver and PostGIS based hosting plans
* FOSS Consulting Services
Irc: timlinux on #qgis at freenode.net
More information about the Qgis-developer