7

I would like to create a RasterLayer from GeoTIFF data provided by a server. I'll query the server for this data using a httr::GET call (the data is provided on-demand, so in the application there won't be a url ending in .tif but a query url).

After writing the result of this call to disk as a GeoTIFF file it's easy enough to create the RasterLayer from the resulting GeoTIFF file on disk:

library(httr)
library(raster)

url <- 'http://download.osgeo.org/geotiff/samples/gdal_eg/cea.tif'

geotiff_file <- tempfile(fileext='.tif')
httr::GET(url,httr::write_disk(path=geotiff_file))
my_raster <- raster(geotiff_file)
my_raster

However, I would like to skip the write to disk part and create the raster straight from the in-memory server response.

response <- httr::GET(url,httr::write_memory())
response

The content of the response is a raw string which I would need to interpret as geoTIFF data.

str(httr::content(response))

However, I can only find raster or rgdal functions to read from a file. Any suggestions on translating this raw string to a raster?

Thanks!

WillemMaetens
  • 73
  • 7

1 Answers1

7

GDAL has some cool virtual file system driver, one of which is /vsicurl that

allows on-the-fly random reading of files available through HTTP/FTP web protocols, without prior download of the entire file. It requires GDAL to be built against libcurl.

Since the raster package builds on rgdal you can simply do this:

library(raster)

r <- raster('/vsicurl/http://download.osgeo.org/geotiff/samples/gdal_eg/cea.tif')

plot(r)

enter image description here

Val
  • 6,585
  • 5
  • 22
  • 52
  • This does not work for me on windows. Perhaps because the windows rgdal was not built with libcurl? – Robert Hijmans Sep 07 '18 at 19:06
  • @RobertHijmans Haven't tested it on windows - works on Linux. Could be either an issue with libcurl or the gdal version. Do the other virtual file systems work for you? – Val Sep 07 '18 at 21:06
  • Works on Linux, but on Windows the raster() line gives an error: Error in .rasterObjectFromFile(x, band = band, objecttype = "RasterLayer", : Cannot create a RasterLayer object from this file. (file does not exist) – WillemMaetens Sep 12 '18 at 06:21
  • I tried with a path to a file on disk, but that gives the same error. Not sure how to check whether other virtual file systems work. – WillemMaetens Sep 12 '18 at 06:31
  • @WillemMaetens You could try `/vsizip' which can read images from a `.zip` archive on disk. If that works, it's probably an issue with libcurl. Another option would be trying python GDAL, to see if it's an issue with `rgdal` – Val Sep 12 '18 at 06:42
  • 1
    I tried a couple of things on Windows. First off, gdal.Open("/vsicurl/http://download.osgeo.org/geotiff/samples/gdal_eg/cea.tif") does work in python. In R, gdalinfo() from package gdalUtils (https://www.rdocumentation.org/packages/gdalUtils/versions/2.0.1.14/topics/gdalinfo) on the url works: gdalUtils::gdalinfo('/vsicurl/http://download.osgeo.org/geotiff/samples/gdal_eg/cea.tif'). Similarly /vsizip on a zip file works with gdalUtils::gdalinfo(), but not with raster(). Seems to point to how the paths are handled by the raster function, strange thing is that it's specific for windows. – WillemMaetens Sep 12 '18 at 08:41
  • @WillemMaetens nicely tested. Looks like the issue is with `rgdal`, either with the gdal version or with the libcurl functionality. – Val Sep 12 '18 at 09:07
  • A workaround on windows is to build a vrt 1st (QGIS has this same issue) ``` library(gdalUtils) gdalbuildvrt('/vsicurl/http://download.osgeo.org/geotiff/samples/gdal_eg/cea.tif','cea.vrt') r <- raster('cea.vrt') ``` – wildintellect Sep 21 '18 at 22:25