SpatialTau is my weekly newsletter that goes out every Wednesday. The archive shows up in my blog a month after the newsletter is published. If you’d like to subscribe, please do so here.
Tilting at the Shapefile
Now I’m sure if I went back to my blog and searched for how many times I’ve tried to kill off the shapefile even I would be surprised at how many times I’ve blogged about it. Thus it seems about for the second newsletter I’ve ever written to focus on the “Shapefile Problem”
So what exactly is this problem? I mean what is so bad about a well supported, somewhat open file format? I’ve told this story before but it never hurts to repeat. My dad was borrowing my laptop a couple years ago and commented about all these DBF files all over my desktop. He wondered why on earth would I have a format that he used in the late 80’s and outgrew because of it’s limitations. Well I proceeded to explain to him the shapefile and how it worked and he just laughed. That’s right, my 72 year old dad laughs at us wankers and our shapefile. The DBF is only half the problem with the shapefile. It doesn’t understand topology, only handles simple features (ever try and draw a curve in a shapefile?), puny 2GB file size limitation and not to mention you can’t combine points, polygons and lines in one file (hence every shapefile name has the word point, line or poly in it).
Oh and it’s anywhere between 3 and 15ish file types/extensions. Sure 3 are required but the rest just clutter up your folders. I love the *.shp.xml one especially because clearly they thought so much about how to render metadata. If I had a penny for every time someone emailed me just the *.shp file without the other two I’d be a rich man. Heck just the other day I got the *.shp and *.dbf but not the *.shx. Just typing the sentence makes me cringe.
- The File Geodatabase (FGDB): Esri’s default format for their tools. It’a spatial database in a folder format. The less mentioned about the Personal Geodatabase, the better. But unlike most companies in the past 5 years, it isn’t built on SQLite, but Esri proprietary geodatabase format. There isn’t anything inherently wrong with Esri taking this path but it means you’re stuck using their software or their APIs to access the file format. To me this severely limits the FGDB to me an interchange file format and I think that is perfectly fine with Esri as they don’t really care too much if the FGDB doesn’t work with other’s software. I’d link to an Esri page that describes the FGDB but there isn’t one. It’s a secret proprietary format that even Esri doesn’t want to tell you about.
- SpatiaLite: SpatiaLite has everything going for it. It’s a spatial extension to SQLite which means at its core it’s open. It’s OGC Simple Features compliant. It is relatively well supported by GIS software (even Esri technically can support it with the help of Safe Software). Plus it supports all those complex features that the shapefile can’t. Heck OGC even chose it as the reference implementation for the GeoPackage (assuming people still care about that). Heck supports rasters too! But honestly, SpatiaLite was released in 2008 and hasn’t really made a dent into the market. I can’t ever remember downloading or being sent a SpatiaLite file. I’m guessing you can’t either. I mean we all want a format that is similar to PostGIS and easily transferable (one file). On paper that’s SpatiaLite. But I think we have to chalk this up as Esri not supporting the format and it is relegated to niche use.
- GML/KML: Ron Lake probably loves I grouped these together but honestly they’re so similar in basic structure I’ve really just left them together. My company uses KML quite a bit to share georeferenced photos. That’s about it, pretty low use. There is a ton of KML out there but it is mostly points. There might be a ton of GML out there but I’m not Ron Lake. KML is nice in the sense it has visualization included in the spec (you can make a line yellow) but it isn’t enough to get excited about. It’s an OGC standard but as with SpatiaLite that doesn’t really seem to matter in the real world. Don’t even try and use a different projection. They have their use in specific cases but the limits of the formats means you’ll never see it being an interchange format. Plus XML? Oh and feel free to email me how GML is powerful because it supports OGC Simple features, I’ll still include it with KML.
- GeoJSON: It’s an open standard, so open in fact that OGC isn’t involved. That’s a huge plus because mostly standards organizations do is make complex file formats for simple uses. That’s not what GeoJSON is. It can be many types of projections, it can be points, polygons and lines (with variations of many), it supports topology with the TopoJSON format and it’s JSON so it’s human readable. But alas it isn’t supported by Esri so we run into the same problem as SpatiaLite. BUT, Esri has shown interest in GeoJSON so there is hope that it will be well supported soon. As with the shapefile/KML and unlike SpatiaLite it won’t support curves and other complex geometry or rasters and never will. Thus it is not well suited as a shapefile replacement.
- Well Known Text (WKT): This comes out of the OGC and is used by software such as PostGIS for storage. WKT supports lots of geometric objects (curves!) and TINs. I’ve never been limited by WKT for vector files (you can almost feel where the end of this is going though) and many spatial databases from PostGIS and Oracle to SpatiaLite and SQL Server use the WKB (Well Known Binary) equivalent to store information. But alas, we still don’t support rasters. It’s a vector format for vector data. SpatiaLite and the File Geodatabase both support rasters.
There are many other formats but I think these are the only ones that really have any traction. I could list formats such as GeoTIFF and say you could use that for rasters but you are limited to 4GB of data. The vector guy in me wants to just say the heck with it all and use GeoJSON and WKT to solve the problem but given I’m still writing about this subject in December 2014 neither is a good solution. We’re left with one simple truth…
The shapefile will outlive us all. Unless Esri stops supporting it with their software at the same time as QGIS, Autodesk, etc it will continue to be the format that everyone uses. In 2014 I’d wager 80% of all production geospatial data (I’m sandbagging here, probably this number is 95%) is stuck in the shapefile format where it resides comfortably. Personally I’m a big fan of GeoJSON but I’ve started to get back into WKT lately and love the complex geometry support. If there is one thing I’ve learned in the past 20 years of “professional GIS” I’ve done, the shapefile is king.