On Saturday, 18th and Sunday 19th of February, a two days MapOSMatic hackfest took place in San Francisco, in the offices kindly provided by DotCloud. David Decotigny, Maxime Petazzoni, Jérôme Petazzoni and myself, Thomas Petazzoni, were the participants of this short session of hacking around MapOSMatic. The result of those two days of work is not yet visible on the production web site, but we hope to make it available soon.
Amongst the things that have been worked on:
- David has worked on understanding a Mapnik rendering bug that happens with “halo” text when the Cairo surface is scaled and rendered to a PDF file. This bug is currently very annoying for MapOSMatic because all text rendered with a “halo” in the background looks really ugly on PDF maps. We have reported this problem in August 2010 on Mapnik bugtracker (see this ticket, which has then be converted to Github to this ticket). After two days diving into the Mapnik source code, David found out that disabling the glyph cache used by Mapnik made the problem go away, but this workaround would presumably significantly affect the rendering performance, so it really is a workaround and not a solution. We hope to get some feedback from the Mapnik community about this, but since last month when David worked on the issue, we haven’t so far received any feedback.
- Jérôme has worked on the monitoring of our servers. He has set up a monitoring website at http://stats.maposmatic.org/r2d2/. Amongst classical metrics (CPU, disk, etc.), Jérôme has added one special metric: the lag that our OSM database has compared to the official OSM database. Many of our users have often been asking why they are not quickly seeing their OSM changes in MapOSMatic maps. The diagram is visible at http://stats.maposmatic.org/r2d2/browse/gcc10.fsffrance.org/replication/ and shows this replication lag in seconds. At the moment, we are about 1.2 million seconds behind the official OSM database, which means about 14 days of lag.
- Maxime has worked on the infrastructure as well, migrating the development version of the MapOSMatic installation to Mapnik 2, installing the MapQuest stylesheet, updating the OSM stylesheet, starting a new GIS database for importing the OSM data under the new schema expected by the latest OSM stylesheets.
- On my side, I worked on fixing various minor issues on the web frontend that are currently preventing us from putting in production the current development version.
All in all, in just two days, we made a bit of progress, but not a lot. None of us had touched the project for quite a while, so it took some time to get up to speed, and two days were not sufficient to bring major results. However, we knew that a second hackfest would happen in France a month after, so the work will definitely continue !