Report back on the QGIS hackfest in Vienna, November 2009
First a big 'thank you'
So I've just arrived back in Zurich en route to Johannesburg after a thrilling and tiring week hackfesting in Vienna. I'm going to try to summarise some of the major activities and outcomes from the hackfest and will ask others to fill in what I forget in the comments.
First off I would like to say a big thank you to Werner Macho and Dr. Reinfried Mansberger. Dr. Mansberger (Institute of Surveying, Remote Sensing and Land Information (IVFL), Vienna) kindly allowed for the use of facilities (meeting rooms, kitchen, LCD projector, internet access). The facilities were excellent, comfortable and the perfect fit for the twenty or so attendees at the hackfest. Werner Macho organised everything and provided for a hassle free and extremely pleasant hackfest where everything 'just worked'. As an additional bonus his girlfriend Verena baked up a storm keeping hungry hackers fed with an endless stream of delicious cakes! It should be noted that in three and a half days of hackfesting, the hackfest team consumed more coffee than is normally used in four months at the faculty! Dr. Mansberger kindly ensured that the bill for all that coffee was covered, a gesture which is heartily appreciated by the QGIS team who would otherwise been a bunch of zombies and not managed to hack into the wee hours of the morning!
Secondly let me thank the other attendees. I can honestly say that I would be hard pressed to find a nicer bunch of people to spend 3 days locked up in the same room with. Each person brought a little bit of magic to the meeting making the whole event a real pleasure. Above and beyond that, the attendees were extremely dedicated and enthusiastic QGIS and FOSSGIS users and the conversations I had with the various people I got to spend time with were extremely stimulating. It was also really great having virtual participation in the IRC channel from those who couldnt be there in person - the input from IRC was a great motivator and a good way to keep us on our toes and pick up any issues to commits to the code base.
Lastly, I would like to say a huge thank you for the people and companies who so kindly donated towards our hackfest funding drive. Personally I am self employed and needed to travel half way around the world to attend the hackfest so attending it means a week of no income for me before I even consider the travel costs. Others are in the same boat and the money that was donated was put to really good use in covering the travel expenses of a number of attendees. Not having to worry about travel and accommodation costs on top of losing a week of work time made it a lot easier for me to attend. Also some of the attendees had their presence at the hackfest paid for by their employers so thank you to those employers who made this possible. I do hope you will consider sponsoring our project or making a donation so that we can continue to hold these six monthly get togethers and keep up the momentum of the project.
Summarising the hackfest
With those thanks out of the way I would like to summarise the hackfest a little for you eager folks out there dying to know what we got up to. The hackfest is actually an 'unmeeting' in that we arrive with no real agenda except to 'make QGIS better' and then organise ourselves on the fly. This works well because it allows us to be dynamic and make the best use of the skills of the people that arrive at the hackfest. That said we did have some broad topic areas on the hackfest wiki page that we planned to cover.
Activities at the hackfest could broadly be broken up into:
As you can see from my list, by far the most emphasised activity was hands-on work on the code base. Presentations were spread out across the event so that we did not get 'presentation overload' and had a chance to digest the contents.
Presentations
Martin Dobias gave the first presentation where he demonstrated the symbology-ng branch and his work on an improved labelling engine. The symbology-ng (next generation) branch is a complete overhaul of the existing symbology infrastructure in order to address some limitations of the current architecture and provide a capability to do high quality cartography within QGIS. These limitations include:
The symbology-ng implementation also provides a simpler interface for symbol selection and definition.
Martin's presentation spawned a huge amount of discussion, the outcome of which was that we all voted for Martin to merge his work into trunk as soon as possible. While the symbology-ng implementation still has outstanding issues, it can run alongside the existing symbology implementation and will be turned off by default. This will allow early adopters a chance to play with the technology from within the mainstream QGIS builds. It will also provide a platform for better bug reporting and feedback. We will aim for symbology-ng to be the default symbology rendering engine in the future QGIS 2.0 release. Here are a couple of screenshots of what you can expect when his work is merged into trunk. Note that all of the images below can be clicked on for a larger view.
Martin followed up his symbology-ng presentation with one on a new labelling engine based on PAL. This provides a way to address various deficiencies within the existing QGIS labelling implementation by providing collision detection, label priorities and optimal label placement capabilities. The new labelling engine is also able to make labels follow lines. Martin showed us the labelling engine calculating label placement in various situations and it was indeed impressive. As with the symbology-ng we voted to include the new labelling engine in trunk at Martin's earliest convenience since it runs as a plugin and will be disabled by default so is a low risk inclusion into the code base. Some screenies of the labelling plugin follow:
Massimo Di Stefano presented the work he has been involved in on the Google Summer of Code project they did to integrate QGIS and ossimPlanet. The integration work uses TCP/IP to update viewpoint position and send raster data over to the ossimPlanet application - where it is viewed on a 3D globe. The globe syncronises to the current QGIS view extents and additional tools provide fine tuned options for postioning the ossimPlanet view. We were all really impressed by their work. Some discussion followed with the general concensus being that we wouldnt be able to ship this as a core part of QGIS due to the number of dependencies it adds and the fact that it does not yet meet ease of use requirements. It is however a great step forward towards providing 3D visualisation capabilities within QGIS and we will work with the developers to make their work available as an add on to QGIS (not part of the standard installation) once the code matures and can be easily packaged.
Alexander Bruy presented work on two projects he has been involved in. In the first they have been building a high quality sample dataset for Ukranian and Russian open source GIS users to use as the basis for tutorials and learning GIS. The dataset has Russian place names etc. and so is more familiar to users than the traditionally English based datasets common in GIS tutorials. The second part of his presentations covered a plugin they have written for recording protected areas and the species that occurr in them. This work will be a great contribution towards and eventual open source EcoGIS implementation (see discussion section below). It was really gratifying to see Alex's presentation and the great use they are putting QGIS to in the Ukraine. Alex is not the only Russian QGIS hacker - other members of the GIS-Lab group have provided us with some great improvements to QGIS especially in the Image Registration tool.
Martin gave a final presentation where he showed off the work they did with the gismapper project. The project involved building tools to map infrastructure using a car fitted with GPS, spherical camera, compass and a computer fitted with a touch screen and running Ubuntu with QGIS. I posted an earlier video about which I am including again here for your convenience.
I know the video quality isn't brilliant but hopefully it gives you a feel for this excellent project (which is all open source by the way!). The presentation prompted some discussion about the project and which aspects of it make sense to include into QGIS core. The basic outcome of that discussion is that we will aim to include the GPS communication code and support for a moving GPS cursor on the map canvas into core, and evaluate what else makes sense to incorporate on a case-by-case basis.
Discussions
We had a number of great round table discussions covering a range of topics such as:
I am going to touch on the user interface related stuff briefly here and invite you to discuss other items in more detail on the QGIS mailing lists. Note that I did provide quite a bit of detail of the individual things people were working on over at the QGIS Blog
The main thing I worked on during the hackfest was a user interface consistency review. On the first day of the hackfest we formed a small team consisting of myself, Milena Nowotarska (Poland), Werner Macho (Austria), and Anita Graser (Austria). The goal of our work is to visit each and every dialog and ensure that it meets our Human Interface Guidelines (HIG). We started discussing the existing HIG and planning what things we were going to cover during the hackfest and beyond to the release of QGIS 2.0. This can be summarised as:
Needless to say this is quite a tall order but we made a good dent by doing an initial pass through the dialogs - you should already be able to see quite a lot of the improvements that we have made in the nightly builds of QGIS or in your local svn checkout. The topic of icon usage within QGIS received quite a bit of attention. The GIS icon theme is great in that it is visually consistent but the feedback I received from most users was that they still preferred the 'old' default theme. We discussed why this may be the case, with the concensus being that it was down to:
We would really like to have a consistent icon theme in QGIS 2.0 so Anita has started a revision of the icons to try to address some of these issues. We will use this posting as a starting point for a discussion with Robert Szczepanek (the original authoer of the GIS theme) and try to work with him to pass our changes upstream where he is agreeable. You can see an example of the original icon theme from Robert (first image below) and the updates from Anita (second image below).
I spent quite a lot of time exploring how to improve the user friendliness of the map composer and other dialogs where there are many widgets and options. Below you can see the outcome of some of this work.
Firstly you can see that the widgets are now arranged vertically - removing horizontal widget arrangment means that the form part of the composer (to the right of the split) can occupy a lot less space and degrades gracefully as it narrows. The previous implementation shrank the widgets into non-existence as it got too narrow. You will see that where widgets have labels (e.g. Title), the label is placed above the widget. This allows the label to be longer in translated versions of QGIS without causing the related widget ot disappear into oblivion. Experimentally we are also going to be placing lables into widgets as seen in the spinboxes in the above screenie since it saves vertical space and allows more options to be viewed on the screen at one go.
You will also notice that we are making use of a toolbox (areas marked 'General', 'Legend Items', and 'Item Options') in the above screenie. This allows us to pack a lot more information into a small space and easily access forms in each section. In the next screenie below, you can see what happens when another toolbox item is clicked.
These kind of user interface cleanups will be taking place throughout the application with the goal of completing them for 2.0. - I hope they are a step in the right direction for everyone.
No discussion of the QGIS user interface would be complete without a lengthy and heated debate about splash screens - and of course we had one! Suffice to say that the QGIS developers are getting a little tired of planetary moons as the theme for our releases and come version 2.0 we will be switching to a new theme for our splashscreens....you will have to wait a little to find out what that will be...
Planning for QGIS 2.0
We also had a discussion about the roadmap to version 2.0. QGIS 2.0 will break API compatibility with version 1.0 and thus presents an opportunity for us to overhaul various aspects of the API that we feel are not working well or could otherwise be improved. This will include changes needed for symbology-ng, labelling, better geometry classes and other yet to be determined changes. We will aim for a release of QGIS 1.4 by December 2009. Following that we will do one more feature focussed release, planned for around March or April 2010. After 1.5, we will do only bug fix 1.5.x releases and experimental builds of trunk. 1.5 will be the last API compatible general release and the next official release will be QGIS 2.0 which will require most plugin and third party application writers to update their work to the new API. The experimental releases leading up to version 2.0 will be numbered 1.9.x.
QGIS 2.0 will be released 'when its ready' but hopefully some time in the latter half of 2010.
Planning for hackfest #3
We discussed options for our next hackfest - we will aim to hold it in the first half of 2010 in Piza, Italy. Paolo Cavallini will be the local organiser / liason.
General thoughts on hackfest planning
That's it for now
I'm wrapping up this hackfest for summary, although I have only scratched the surface of what we did - hopefully other attendees will also blog about their activities or leave comments about what they were up to. Also don't forget to check out the developer profiles link I mentioned above. Now I'm off to catch up on 4 nights of missed sleep....
