Downloading XLSX sometimes has large column gaps
longhotsummer opened this issue · 8 comments
Downloading some wide XLSX data sometimes produces large empty columns. This seems to be an issue with the gdal library we're using to build the XLSX.
It looks like this was fixed in GDAL 1.11 at some point a while back, I'm not sure why we're still seeing it.
I can confirm it still happens with GDAL 1.11.1, perhaps 1.11.2 fixed it.
I think we should explore upgrading to GDAL 2.x, taking into account common GDAL versions on Ubuntu (14.04 and 16.04).
Ubuntu 14.04 LTS (Trusty Tar) gets 1.10 by default, which is actually too old for Wazimap in any case, and we have to pull the 1.11 version from a separate repo.
Ubuntu 16.04 LTS has 1.11.3 which should be suitable.
So sticking with >1.11.1 and <2.0 seems reasonable for now.
The buildpack we use to pull in this dependency in production (dokku, heroku) is fixed at 1.11.1 for the heroku/dokku ceder-14 stack: https://github.com/cyberdelia/heroku-geo-buildpack
$ curl http://cyberdelia-geo-buildpack.s3.amazonaws.com/cedar-14/manifest.gdal
1.11.1
Another alternative is to write our own Excel files, that's pretty straight forward.
We may be able to use the Apt buildpack from heroku to pull in exactly the gdal we want. https://elements.heroku.com/buildpacks/heroku/heroku-buildpack-apt