Introducing osmextract

This vignette provides an introduction to using the package, building on the README which covers installation and our motivations for creating it.

Loading the package generates important messages about the license associated with OSM data.

#> Data (c) OpenStreetMap contributors, ODbL 1.0.
#> Check the package website,, for more details.

The first thing to say is: do not ignore this message! The Open Street Map (OSM) extracts are stored by external providers such as Geofabrik, Bbbike, or There are important legal considerations that you should be aware of before using OSM data, especially if you are working in a for-profit capacity.

Main package functions

The package is composed of the following main functions:

  1. oe_providers(): Show which OSM providers are available;
  2. oe_match(): Match an input place with one of the files stored by the OSM providers;
  3. oe_download(): Download the chosen file;
  4. oe_vectortranslate(): Convert between .pbf and .gpkg formats;
  5. oe_read(): Read .pbf and .gpkg files;
  6. oe_get(): Match, download, (vector)translate, and import data, all in one step.

For many users who just want to get OSM data quickly, oe_get() may be sufficient, as covered in the README. We will demonstrate each function in turn, following the same order in which they are typically used. As you can see, the name of the most important functions in this package start with oe_* prefix, which means that you can easily use auto-completion features (with Rstudio or similar IDE(s)).

oe_providers(): List providers

oe_providers() lists the providers that are currently available with the version of osmextract you have installed.

Each element in the column database_name is a data object that is packaged with osmextract. You can read a detailed description of each provider data running, for example, ?geofabrik_zones or ?bbbike_zones.

Perhaps, the best known bulk OSM data provider is Geofabrik, and its extracts are summarised as a data.frame in the packaged object geofabrik_zones.

Note that in addition to being a data frame with rows and columns, geofabrik_zones is also an sf object, as defined in the package of the same name. When working with sf objects, it makes sense to have the package loaded:

That gives you access to many functions for working with geographic vector data of the type provided by osmextract. Each row of data in an sf object contains a geometry, representing the area covered by a provider zone, meaning you can plot the data as follows:

The plot above shows how the provider divides geographic space into discrete chunks. Different providers have other zoning systems. For example:

As shown in the visualisation above of zones in Europe, that provider offers rectangular extracts of the major cities. We are working on adding support for manually selected regions from the BBBike website (see

Check the “Comparing the supported OSM providers” vignette for some simple guidelines on how to choose the best provider.

oe_match(): Match an input place with an OSM extract

The function oe_match() takes in input a string through the parameter place, and it returns a named list of length two with the URL and the size (in bytes) of a .osm.pbf1 file representing a geographical zone stored by one of the supported providers. For example:

The geographical zone is chosen by calculating the Approximate String Distance (?adist()) between the input place and one of the fields in the provider’s dataset. Then, the function selects the closest match. By default, oe_match() uses the name field and Geofabrik provider, but you can select a different field via the argument match_by. We refer to the providers’ help pages for a detailed description of all available fields. If you are using Geofabrik provider, a useful and interesting alternative field is represented by the (unique and unambiguous) iso3166-1 alpha2 codes:

There are a few scenarios where the iso3166-1 alpha2 codes in geofabrik_data cannot be used since there are no per-country extracts (e.g. Israel and Palestine):

For this reason, we coded a function named oe_match_pattern() to explore the matching operations for all available providers according to a pre-defined pattern. It returns a named list where the names are the id(s) of the supported OSM providers and the values are the matched names. For example:

The default field is name, but we can change that as follows:

If we set full_row = TRUE, then oe_match_pattern() will return the complete row(s) from each provider’s data:

We can combine the two functions as follows:

The parameter max_string_dist (default value is 1) represents the maximum tolerable distance between the input place and the closest match in match_by column. This value can always be increased to help the matching operations, but that can lead to false matches:

The parameter max_string_dist is always set to 0 if match_by argument is equal to iso3166_1_alpha2 or iso3166_2.

If the approximate string distance between the closest match and the input place is greater than max_string_dist, then oe_match() will also check the other supported providers. For example:

Finally, if there is no exact match with any of the supported providers and match_by argument is equal to "name", then oe_match() will use the Nominatim API to geolocate the input place and perform a spatial matching operation (explained below):

Finding zones based on geographic inputs

The input place can also be specified using an sf, sfc, or bbox object with arbitrary CRS2, as documented in the following example. oe_match() will return a named list of length two with the URL and the size of a .pbf file representing a zone that geographically contains the sf or sfc object (or an error if the input is not contained into any geographical area).

If the input place intersects multiple geographically nested areas and the argument level is equal to NULL (the default value), then oe_match() automatically returns the extract with the highest level. In particular, we could roughly say that smaller geographical areas are associated with higher level(s). For example, level = 1 may correspond to continent-size extracts, 2 is for countries, 3 represents the regions and 4 the subregions:

If there are multiple OSM extract intersecting the input place at the same level, then oe_match() will return the area whose centroid is closest to the input place.

If you specify more than one geometry into the sf or sfc object, then oe_match() will select an area that contains all of them.

The same operations work with LINESTRING or POLYGON objects:

The input place can also be specified using a numeric vector of coordinates. In that case, the CRS is assumed to be EPSG:4326:

Finally, to reduce unnecessary computational resources and save bandwidth/electricity, we will use a small OSM extract in subsequent sections that can be matched as follows:

oe_download(): Download OSM extracts

The oe_download() function is used to download .pbf files representing OSM extracts. It takes in input a URL, through the parameter file_url, and it downloads the requested data in a directory (specified by the parameter download_directory):

The argument provider can be omitted if the input file_url is associated with one of the supported providers. The default value for download_directory is tempdir() (see ?tempdir), but, if you want to point to a directory that will persist, you can add OSMEXT_DOWNLOAD_DIRECTORY=/path/for/osm/data in your .Renviron file, e.g. with:

You can always check the default download_directory used by oe_download() with:

We strongly advise you setting a persistent directory since downloading and converting (see the next sub-section) .pbf files are expensive operations, that can be skipped if the functions detect that the requested extract was already downloaded and/or converted.

More precisely, oe_download() runs several checks before actually downloading a new file, to avoid overloading the OSM providers. The first step is the definition of the path associated with the input file_url. The path is created by pasting together the download_directory, the name of the chosen provider (specified by provider argument or inferred from the input URL), and the basename() of the URL. For example, if file_url is equal to "", and download_directory = "/tmp/, then the path is built as /tmp/geofabrik_italy-latest.osm.pbf. In the second step, the function checks if the new path/file already exists (using file.exists()) and, in that case, it returns the path, without downloading anything3. Otherwise, it downloads a new file (using download.file() with mode = "wb") and then it returns the path.

oe_vectortranslate(): Convert to gpkg format

The oe_vectortranslate() function translates a .pbf file into .gpkg format4. It takes in input a string representing the path to an existing .pbf file, and it returns the path to the newly generated .gpkg file. The .gpkg file is created in the same directory as the input .pbf file and with the same name. The conversion is performed using ogr2ogr through vectortranslate utility in sf::gdal_utils().

We decided to adopt this approach following the suggestions of the maintainers of GDAL. Moreover, GeoPackage files have database capabilities like random access and querying that are extremely important for OSM data (see below).

Let’s start with an example. First, we download the .pbf file associated with ITS example:

its_pbf = oe_download(its_details$url, provider = "test")
#> The chosen file was already detected in the download directory. Skip downloading.
list.files(oe_download_directory(), pattern = "pbf|gpkg")
#>  [1] "bbbike_Muenchen.gpkg"                      "bbbike_Muenchen.osm.pbf"                  
#>  [3] "geofabrik_Baku.gpkg"                       "geofabrik_Baku.osm.pbf"                   
#>  [5] "geofabrik_Chicago.gpkg"                    "geofabrik_Chicago.osm.pbf"                
#>  [7] "geofabrik_Helsinki.gpkg"                   "geofabrik_Helsinki.osm.pbf"               
#>  [9] "geofabrik_Leeds.gpkg"                      "geofabrik_Leeds.osm.pbf"                  
#> [11] "geofabrik_andorra-latest.gpkg"             "geofabrik_andorra-latest.osm.pbf"         
#> [13] "geofabrik_belgium-latest.gpkg"             "geofabrik_belgium-latest.osm.pbf"         
#> [15] "geofabrik_centro-latest.gpkg"              "geofabrik_centro-latest.osm.pbf"          
#> [17] "geofabrik_england-latest.gpkg"             "geofabrik_england-latest.osm.pbf"         
#> [19] "geofabrik_goa-latest.gpkg"                 "geofabrik_goa-latest.osm.pbf"             
#> [21] "geofabrik_greater-london-latest.gpkg"      "geofabrik_greater-london-latest.osm.pbf"  
#> [23] "geofabrik_isle-of-wight-latest.gpkg"       "geofabrik_isle-of-wight-latest.osm.pbf"   
#> [25] "geofabrik_liechtenstein-latest.gpkg"       "geofabrik_liechtenstein-latest.osm.pbf"   
#> [27] "geofabrik_lombardia-latest.osm.pbf"        "geofabrik_luxembourg-latest.gpkg"         
#> [29] "geofabrik_luxembourg-latest.osm.pbf"       "geofabrik_maine-latest.osm.pbf"           
#> [31] "geofabrik_malta-latest.gpkg"               "geofabrik_malta-latest.osm.pbf"           
#> [33] "geofabrik_nord-ovest-latest.gpkg"          "geofabrik_nord-ovest-latest.osm.pbf"      
#> [35] "geofabrik_portugal-latest.gpkg"            "geofabrik_portugal-latest.osm.pbf"        
#> [37] "geofabrik_vatican_city-latest.gpkg"        "geofabrik_vatican_city-latest.osm.pbf"    
#> [39] "geofabrik_wales-latest.gpkg"               "geofabrik_wales-latest.osm.pbf"           
#> [41] "geofabrik_west-yorkshire-latest.gpkg"      "geofabrik_west-yorkshire-latest.osm.pbf"  
#> [43] "openstreetmap_fr_lombardia-latest.gpkg"    "openstreetmap_fr_lombardia-latest.osm.pbf"
#> [45] "test_its-example.gpkg"                     "test_its-example.osm.pbf"

and then we convert it to .gpkg format:

its_gpkg = oe_vectortranslate(its_pbf)
#> The corresponding gpkg file was already detected. Skip vectortranslate operations.
list.files(oe_download_directory(), pattern = "pbf|gpkg")
#>  [1] "bbbike_Muenchen.gpkg"                      "bbbike_Muenchen.osm.pbf"                  
#>  [3] "geofabrik_Baku.gpkg"                       "geofabrik_Baku.osm.pbf"                   
#>  [5] "geofabrik_Chicago.gpkg"                    "geofabrik_Chicago.osm.pbf"                
#>  [7] "geofabrik_Helsinki.gpkg"                   "geofabrik_Helsinki.osm.pbf"               
#>  [9] "geofabrik_Leeds.gpkg"                      "geofabrik_Leeds.osm.pbf"                  
#> [11] "geofabrik_andorra-latest.gpkg"             "geofabrik_andorra-latest.osm.pbf"         
#> [13] "geofabrik_belgium-latest.gpkg"             "geofabrik_belgium-latest.osm.pbf"         
#> [15] "geofabrik_centro-latest.gpkg"              "geofabrik_centro-latest.osm.pbf"          
#> [17] "geofabrik_england-latest.gpkg"             "geofabrik_england-latest.osm.pbf"         
#> [19] "geofabrik_goa-latest.gpkg"                 "geofabrik_goa-latest.osm.pbf"             
#> [21] "geofabrik_greater-london-latest.gpkg"      "geofabrik_greater-london-latest.osm.pbf"  
#> [23] "geofabrik_isle-of-wight-latest.gpkg"       "geofabrik_isle-of-wight-latest.osm.pbf"   
#> [25] "geofabrik_liechtenstein-latest.gpkg"       "geofabrik_liechtenstein-latest.osm.pbf"   
#> [27] "geofabrik_lombardia-latest.osm.pbf"        "geofabrik_luxembourg-latest.gpkg"         
#> [29] "geofabrik_luxembourg-latest.osm.pbf"       "geofabrik_maine-latest.osm.pbf"           
#> [31] "geofabrik_malta-latest.gpkg"               "geofabrik_malta-latest.osm.pbf"           
#> [33] "geofabrik_nord-ovest-latest.gpkg"          "geofabrik_nord-ovest-latest.osm.pbf"      
#> [35] "geofabrik_portugal-latest.gpkg"            "geofabrik_portugal-latest.osm.pbf"        
#> [37] "geofabrik_vatican_city-latest.gpkg"        "geofabrik_vatican_city-latest.osm.pbf"    
#> [39] "geofabrik_wales-latest.gpkg"               "geofabrik_wales-latest.osm.pbf"           
#> [41] "geofabrik_west-yorkshire-latest.gpkg"      "geofabrik_west-yorkshire-latest.osm.pbf"  
#> [43] "openstreetmap_fr_lombardia-latest.gpkg"    "openstreetmap_fr_lombardia-latest.osm.pbf"
#> [45] "test_its-example.gpkg"                     "test_its-example.osm.pbf"

The vectortranslate operation can be customised in several ways modifying the parameters layer, extra_tags, osmconf_ini, vectortranslate_options, boundary and boundary_type.

layer argument

The .pbf files processed by GDAL are usually categorized into 5 layers, named points, lines, multilinestrings, multipolygons and other_relations5. The oe_vectortranslate() function can covert only one layer at a time. Nevertheless, several layers with different names can be stored in the same .gpkg file. By default, the function will convert the lines layer (which is the most common one according to our experience), but you can change that using the parameter layer.

The .pbf files always contain all five layers:

while, by default, oe_vectortranslate convert only the lines layer:

We can add another layer as follows:

osmconf_ini and extra_tags

The arguments osmconf_ini and extra_tags are used to modify how GDAL reads and processes a .pbf file. More precisely, several operations that GDAL performs on a .pbf file are governed by a CONFIG file, that you can check at the following link. The package stores a local copy which is used as the standard CONFIG file.

The basic components of OSM data are called elements and they are divided into nodes, ways or relations. Hence, for example, the code at line 7 of that CONFIG file is used to determine which ways are assumed to be polygons if they are closed.

The parameter osmconf_ini can be used to specify the path to a different CONFIG file in case you need more control over GDAL operations. See the next sub-sections for an example. If osmconf_ini is equal to NULL (the default), then oe_vectortranslate() function uses the standard CONFIG file.

Another example can be presented as follows. OSM data is usually described using several tags, i.e. pairs of two items: a key and a value. The code at lines 33, 53, 85, 103, and 121 of the default CONFIG file determines, for each layer, which tags are explicitly reported as fields, while all the other tags are stored in the other_tags column (see here for more details). The parameter extra_tags (default value: NULL) governs which tags are explicitly reported in the .gpkg file and are omitted from the other_tags field. The default tags are always included (unless you modify the CONFIG file or the vectortranslate_options). Please note that the argument extra_tags is ignored if osmconf_ini is not NULL (since we do not know how you generated the new .ini file).

Lastly, the oe_get_keys() function can be used to check all keys that are stored in the other_tags field for a given .gpkg or .pbf file. For example,

Starting from version 0.3.0, if you set values = TRUE, then oe_get_keys returns the values associated to each key (we also defined an ad-hoc printing method):

Check ?oe_get_keys for more details.

We can always re-create the .gpkg file adding one or more new tags:

Check the next sections for more complex, useful, and realistic use-cases.

vectortranslate_options argument

The parameter vectortranslate_options is used to control the arguments that are passed to ogr2ogr via sf::gdal_utils() when converting between .pbf and .gpkg formats. The utility ogr2ogr can perform various operations during the translation process, such as spatial filters or queries. These operations can be tuned using the vectortranslate_options argument. If NULL (default value), then vectortranslate_options is set equal to c("-f", "GPKG", "-overwrite", "-oo", paste0("CONFIG_FILE=", osmconf_ini), "-lco", "GEOMETRY_NAME=geometry", layer). Explanation:

  • "-f", "GPKG" says that the output format is GPKG. This is mandatory for GDAL < 2.3;
  • "-overwrite is used to delete an existing layer and recreate it empty;
  • "-oo", paste0("CONFIG_FILE=", osmconf_ini) is used to modify the open options for the .osm.pbf file and set the path of the CONFIG file;
  • "-lco", "GEOMETRY_NAME=geometry" adjust the layer creation options for the .gpkg file, modifying the name of the geometry column;
  • layer indicates which layer should be converted.

Starting from version 0.3.0, the options c("-f", "GPKG", "-overwrite", "-oo", "CONFIG_FILE=", paste0("CONFIG_FILE=", osmconf_ini), "-lco", "GEOMETRY_NAME=geometry", layer) are always appended at the end of vectortranslate_options unless you explicitly set different default parameters for the arguments -f, -oo and -lco.

boundary and boundary_type arguments

According to our experience, spatial filters are the most common operations added to the (default) vectortranslate process (usually to select a smaller area lying in a larger OSM extract). Hence, starting from version 0.3.0, we defined two new arguments named boundary and boundary_type that can be used to easily apply a spatial filter directly when converting the compressed OSM extract. These new arguments are exemplified in the next sections and can help all users creating less verbose vectortranslate_options.

Other notes

By default, the vectortranslate operations are skipped if oe_vectortranslate() function detects a file having the same path as the input file, .gpkg extension and a layer with the same name as the parameter layer with all extra_tags. In that case, the function will return the path of the .gpkg file. This behaviour can be overwritten by setting force_vectortranslate = TRUE. If the arguments osmconf_ini, vectortranslate_options or boundary parameters are not NULL, the vectortranslate operations are never skipped.

Starting from sf version 0.9.6, if quiet argument is equal to FALSE, then oe_vectortranslate() will display a progress bar during he vectortranslate process.

oe_read(): Read-in OSM data

The oe_read() function is a wrapper around oe_download(), oe_vectortranslate(), and sf::st_read(). It is used for reading-in a .pbf or .gpkg file that is specified using its path or URL.

So, for example, the following code can be used for reading-in the its-gpkg file:

If the input file_path points to a .osm.pbf file, the vectortranslate operations can be skipped using the parameter skip_vectortranslate. In that case, oe_read() will ignore the conversion step.

We can see that the output data includes nine fields (i.e. the default tags), while the previous example had 11 fields (i.e. the default tags + bicycle and foot tags, that were added to the .gpkg file a few chunks above).

We can also read an object starting from a URL:

Please note that if you are reading from a URL which is not linked with any of the supported providers, you need to specify the provider parameter. The test_its-example.osm.pbf file already exists in the download_directory, but we forced the download and vectortranslate operations.

oe_get(): Do it all in one step

To simplify the steps outlined above, while enabling modularity if needs be, we packaged them all into a single function that works as follows:

The function oe_get() is a wrapper around oe_match() and oe_read(), and it summarizes the algorithm that we use for importing OSM extracts:

  1. Match the input place with the URL of a .pbf file through oe_match();
  2. If necessary, download the corresponding .pbf file using oe_download();
  3. Convert it into .gpkg format using oe_vectortranslate();
  4. Read-in one layer of the .gpkg file using sf::st_read().

The following commands (not evaluated here) show how oe_get() can be used to import the OSM extracts associated with the desired input place, after downloading the .pbf file and performing the vectortranslate operations. We suggest you run the commands and check the output.

The arguments osmconf_ini, vectortranslate_options, boundary, boundary_type, query and wkt_filter (the last two arguments are defined in sf::st_read()) can be used to further optimize the process of getting OSM extracts into R.


The following example shows how to create an ad-hoc CONFIG file, which is used by GDAL to read a .pbf file in a customised way. First, we load a local copy of the default osmconf.ini file, taken from the following link.

Then, we modify the code at lines 18 and 21 asking GDAL to report all nodes and ways (even without any significant tag).

We change also the code at lines 45 and 53, removing the osm_id field and changing the default attributes:

Another relevant parameter that could be customised during the creating of an ad-hoc osmconf.ini file is closed_ways_area_polygons (see lines 5-7 of the default CONFIG file). We can now write a local copy of the custom_osmconf_ini file:

and read the ITS Leeds file with the new osmconf.ini file:

If we compare it with the default output:

we can see that there are 2 extra features in the sf object that was read-in using the customized CONFIG file (i.e. 191 features instead of 189 since we set "report_all_nodes=yes" and "report_all_ways=yes") and just 4 field: highway, lanes (see the code a few chunks above), z_order (check the code here), and other_tags.

Please note that the argument extra_tags is always ignored (with a warning message), if you are using an ad-hoc osmconf.ini file:

vectortranslate_options + boundary and boundary_type

The parameter vectortranslate_options is used to modify the options that are passed to ogr2ogr. This is extremely important because if we tune the vectortranslate_options parameter, then we can analyse small parts of an enormous .pbf files without fully reading it in memory.

The first example, reported in the following chunk, shows how to use the argument -t_srs to modify the CRS of the output .gpkg object (i.e. transform from EPSG:4326 to EPSG:27700) while performing vectortranslate operations:

The default CRS of all OSM extracts obtained by Geofabrik and several other providers is EPSG:4326, i.e. latitude and longitude coordinates expressed via WGS84 ellipsoid, while the code EPSG:27700 indicates the British National Grid. Hence, the parameter -t_srs can be used to transform geographical data into projected coordinates, which may be essential for some statistical software like spatstat. The same operation can also be performed in R with the sf package (e.g. ?st_transform()), but the conversion can be slow for large spatial objects. Please note that the default options (i.e. c("-f", "GPKG", "-overwrite", "-oo", "CONFIG_FILE=", paste0("CONFIG_FILE=", osmconf_ini), "-lco", "GEOMETRY_NAME=geometry", layer)) are internally appended to the vectortranslate_options argument.

The next example illustrates how to apply an SQL-like query during the vectortranslate process. More precisely, we can use the arguments -select and -where to create an SQL-like query that is run during the vectortranslate process. Check here for more details on the OGR SQL dialect.

First of all, we need to build a character vector with the options that will be passed to ogr2ogr:

and then we can process the file:

The same procedure can be repeated using an ad-hoc osmconf.ini file.

These arguments are fundamental if you need to work with a small portion of a bigger .pbf file. For example, the following code (not run in the vignette) is used to extract all primary, secondary and tertiary roads from the .pbf file of Portugal stored by Geofabrik servers. After downloading the data, it takes approximately 35 seconds to run the code using an HP ENVY Notebook with Intel i7-7500U processor and 8GB of RAM using Windows 10:

while the classical approach (also not run in the vignette) is slower and provides identical results:

Starting from version 0.3.0, the arguments boundary and boundary_type can be used to perform spatial filter operations during the vectortranslate process. In particular, a spatial boundary can be created using an sf or sfc object (with POLYGON or MULTIPOLYGON geometry) via the argument boundary:

This is the output, where the bounding box was highlighted in black, the intersecting streets in red and all the other roads in grey.

Finally, the argument boundary_type can be used to select among different types of spatial filters. For the moment we support only two types of filters: "spat" (default value) and "clipsrc". The former option implies that the spatial filter selects all features that intersect a given area (as shown above), while the latter option implies that the features are also cropped. In both cases, the polygonal boundary must be specified as an sf or sfc object.

The following example shows how to download from Geofabrik servers the .pbf extract associated with Malta and apply a spatial filter while performing vectortranslate operations. We select and clip only the road segments that intersect a 5 kilometres circular buffer centred in La Valletta, the capital.

The options -t_srs, -select and -where have the same interpretation as before. The spatial filter may return invalid LINESTRING geometries (due to the cropping operation). For this reason, the -nlt and PROMOTE_TO_MULTI options are used to override the default geometry type and promote the LINESTRING(s) into MULTILINESTRING(s). You can use st_cast() to convert the MULTILINESTRING into LINESTRING (which may be important for some packages or functions).

The following map represent the result, where we highlight the bounding circle and the road segments within using a dark-red colour, while all the other road segments are coloured in grey.

The process takes approximately 1 or 2 seconds, while the equivalent R code, reported below, is slower:

The time difference gets more and more relevant for larger OSM extracts. Moreover, the R code crops the road segments using a rectangular boundary instead of the proper circular polygon:

query and wkt_filter arguments

The last two options that we introduce are query and wkt_filter. They are defined in the R package sf and represent a useful compromise between the GDAL and the R approaches explained above, especially when a user needs to apply different queries to the same (typically small or medium-size) OSM extract. In fact, the two parameters create regular queries and spatial filters, respectively, that are applied immediately before reading-in the .gpkg file. The following code, for example, mimics the operations illustrated above, reading-in the road segments that intersect the circular buffer defined around La Valletta:

This is the output and we can see that it applies a circular spatial filter but it doesn’t crop the features:

This approach has its pros and cons. First of all, it’s slightly slower than the GDAL routines, mainly because several unnecessary features are being converted to the .gpkg format. Hence, it may become unfeasible for larger .pbf files, probably starting from 300/500MB. We will test more cases and add more benchmarks in the near future. On the other side, it does not require a new time-consuming ogr2ogr conversion every time a user defines a new query. For these reasons, this is the suggested approach for querying a small OSM extract.

Last but not least, we can use the function hstore_get_value to extract one of the tags saved in the other_tags column without using ogr2ogr and rerunning the oe_vectortranslate() function::

# No extra tag
colnames(oe_get("ITS Leeds", quiet = TRUE))
#>  [1] "osm_id"     "name"       "highway"    "waterway"   "aerialway"  "barrier"    "man_made"  
#>  [8] "z_order"    "other_tags" "geometry"

# Check extra tags
oe_get_keys("ITS Leeds")
#>  [1] "surface"             "lanes"               "bicycle"             "lit"                
#>  [5] "access"              "oneway"              "maxspeed"            "ref"                
#>  [9] "foot"                "natural"             "lanes:backward"      "lanes:forward"      
#> [13] "source:name"         "step_count"          "lanes:psv:backward"  "alt_name"           
#> [17] "layer"               "motor_vehicle"       "tunnel"              "bridge"             
#> [21] "covered"             "incline"             "lanes:psv"           "service"            
#> [25] "turn:lanes"          "turn:lanes:forward"  "frequency"           "indoor"             
#> [29] "lcn"                 "level"               "maxheight"           "operator"           
#> [33] "power"               "source:geometry"     "substation"          "turn:lanes:backward"
#> [37] "voltage"             "website"

# Add extra tag
  "ITS Leeds", 
  provider = "test", 
  query = "SELECT *, hstore_get_value(other_tags, 'bicycle') AS bicycle FROM lines"
#> The input place was matched with: ITS Leeds
#> The chosen file was already detected in the download directory. Skip downloading.
#> The corresponding gpkg file was already detected. Skip vectortranslate operations.
#> Reading query `SELECT *, hstore_get_value(other_tags, 'bicycle') AS bicycle FROM lines' from data source `C:\Users\Utente\Documents\osm-data\test_its-example.gpkg' using driver `GPKG'
#> Simple feature collection with 189 features and 10 fields
#> Geometry type: LINESTRING
#> Dimension:     XY
#> Bounding box:  xmin: -1.562458 ymin: 53.80471 xmax: -1.548076 ymax: 53.81105
#> Geodetic CRS:  WGS 84
#>  [1] "osm_id"     "name"       "highway"    "waterway"   "aerialway"  "barrier"    "man_made"  
#>  [8] "z_order"    "other_tags" "bicycle"    "geometry"

Other providers

The package supports downloading, reading and extracting OpenStreetMap data from various providers. A list of providers can be found at The first provider supported was Geofabrik. The second was bbbike. The package can be extended to support additional providers, as seen in the following commit that adds a working provider.

For information on adding new providers to the package, see the providers vignette.

More on OpenStreetMap

There is a world of knowledge, convention and wisdom contained in OSM data that we hope this package helps you discover and use this knowledge for public benefit. To learn more about the structure of OSM data and the various tagging systems and conventions, the Elements page on the OSM wiki is an ideal place to start. You will find much more excellent content on the OSM wiki pages.

Contributing to OSM

The final thing to say in this introductory vignette is that as a citizen-led project like Wikipedia, OSM relies on a participatory culture, where people not only consume but contribute data, to survive. On that note, we urge anyone reading this to at least sign-up to get an OSM account at

We highly recommend contributing to the world’s geographic commons. The step from being a user to being a contributor to OSM data is a small one and can be highly rewarding. If you find any issues with OSM data, people in the OpenStreetMap will be very happy for you to correct the data. Once logged-in, you can contribute by using editors such as the excellent ID editor, which you can get to by zooming into anywhere you want at and clicking “Edit”.

To learn more about contributing to the amazing OSM community, we recommend checking out the OSM Beginners Guide.

  1. The .pbf format is a highly optimised binary format used by OSM providers to store and share OSM extracts.

  2. If the input spatial object has no CRS, then oe_match() raises a warning message and sets CRS = 4326.

  3. The parameter force_download can be used to override this behaviour in case you need to update an old OSM extract.

  4. The GeoPackage (.gpkg) is an open, standards-based, platform-independent, portable, self-descripting, compact format for transferring geospatial information. See here.

  5. Check the first paragraphs here for more details.