--- author: "Chris Reudenbach" title: "Basic usage of link2GI" date: "`r Sys.Date()`" editor_options: chunk_output_type: console output: html_document: theme: united toc: true rmarkdown: default pdf_document: latex_engine: xelatex toc: true urlcolor: blue vignette: > %\VignetteIndexEntry{Basic usage of link2GI} %\VignetteEncoding{UTF-8}{inputenc}\ %\VignetteEngine{knitr::knitr} --- # Brute force search Automatic searching and finding of installed GIS software binaries is done by the `find` functions. Depending on your operating system and the number of installed versions you will get a data frame with the binary and module folders. ```{r, eval=FALSE} # find all SAGA GIS installations at the default search location require(link2GI) saga <- link2GI::findSAGA() saga ``` Same with `GRASS` and `OTB` ```{r, eval=FALSE} require(link2GI) if (Sys.info()["sysname"] == "Windows") { grass <- link2GI::findGRASS(searchLocation = "C:/") otb <- link2GI::findOTB(searchLocation = "C:/") } else { grass <- link2GI::findGRASS(searchLocation = "/usr/bin",quiet = FALSE) otb <- link2GI::findOTB(searchLocation = "~/apps/otb911/",quiet = FALSE) } grass otb ``` The `find' functions provide an overview of the installed software. These functions do not create links or change settings. # Setting up project structures If you are just calling link2GI on the fly, i.e. for a single temporary operation, there is no need to set up folders and project structures. If you are working on a more complex project, it might be helpful to have a fixed structure. The same goes for existing `GRASS` projects that need to be in specific mapsets and locations. A simple (you can call it dirty) approach is the `initProj` function, which creates folder structures (if not existing) and sets global variables (if desired) containing the paths as strings. ```{r, eval=FALSE} require(link2GI) dirs <- link2GI::initProj( root_folder = tempdir(), standard_setup = "baseSpatial" ) dirs <- link2GI::setupProj( root_folder = getwd(), folders = c("data/level0", "data/level1", "output", "run"), code_subfolder = c("src", "src/functions") ) dirs ``` # linkSAGA - Find and set up 'SAGA' API bindings In the past it was quite tedious to link the correct `SAGA GIS` version. Since version 1.x.x of `RSAGA` things are much better. The new `RSAGA::rsaga.env()` function is to get the first `RSAGA` version in the search path. It is also possible to pass the version number as shown below. Storing the result in appropriate variables will even allow you to easily switch between different `SAGA GIS` installations. ```{R eval=FALSE} require(link2GI) # find all SAGA GIS installations and take the first one saga1 <- link2GI::linkSAGA() saga1 ``` # Find and set up GRASS 7/8 API bindings ## Important note: GRASS runtime environment GRASS GIS relies on a correctly initialized **runtime environment** (PATH, GISBASE, PROJ, GDAL, Python bindings). R (or RStudio) must be started from an environment where these variables are already set. Otherwise, `rgrass` and GRASS command-line calls may fail. ### Windows (OSGeo4W) If GRASS is installed via **OSGeo4W**, R or RStudio **must be started from the OSGeo4W Shell**. OSGeo4W initializes required variables such as `OSGEO4W_ROOT`, `PATH`, `PROJ_LIB`, and `GDAL_DATA`. ### Linux On Linux, GRASS environment variables are usually set by system startup scripts or shell profiles. If GRASS was installed manually, via custom builds, containers, or non-standard locations, R must be started from the same shell session where GRASS is available (e.g. after `grass --text` or sourcing GRASS startup scripts). ### macOS When using Homebrew or standalone GRASS installations, R must be started from a shell that has GRASS on the PATH. GUI launches may miss required environment variables. ## `linkGRASS` `linkGRASS` initializes the session environment and system paths for easy access to `GRASS GIS 7.x./8.x`. The correct setting of spatial and projection parameters is done automatically either by using an existing and valid `raster` or `terra`, `sp` or `sf` object or manually by providing a list of minimum required parameters. These properties are used to initialize either a temporary or a permanent `rgrass` environment, including the correct `GRASS 7/8` database structure. If you do not specify any of the above, `linkGRASS` will create an EPSG:4326 worldwide site. The most time consuming part on Windows systems is the search process. This can easily take 10 minutes or more. To speed up this process, you can also provide a correct parameter set. The best way to do this is to call `findGRASS` manually. Then call `linkGRASS` with the returned version arguments of your choice. The `linkGRASS` function tries to find all valid `GRASS GIS` binaries by analyzing the `GRASS GIS` startup script files. After identifying the `GRASS GIS` binaries, all necessary system variables and settings are generated and passed to a temporary `R` environment. If you have more than one valid installation and run `linkGRASS` with the arguments `select_ver = TRUE`, you will be asked to select one. ### Standard full search usage The most common use of `GRASS` is for a single call or algorithm. The user is not interested in setting all the parameters. `linkGRASS`/`findGRASS` does an automatic search and finds all the `GRASS` binaries using the georeferenced-dataset object for spatial referencing and other necessary settings. **NOTE:** This is the highly recommended linking procedure for all on-the-fly invocations of `GRASS`. Please also note that if more than one `GRASS` installation is found, the one with the highest version number is automatically selected. Take a look at the following examples, which show a typical call for the well-known `sp` and `sf` vector data objects. Starting with `sp`. ```{r, eval=FALSE} # get meuse data as sp object and link it temporary to GRASS require(link2GI) require(sf) require(sp) crs = 28992 # get data data(meuse) meuse_sf = st_as_sf(meuse, coords = c("x", "y"), crs = crs, agr = "constant") # Automatic search and find of GRASS binaries # using the meuse sp data object for spatial referencing # This is the highly recommended linking procedure for on the fly jobs # NOTE: if more than one GRASS installation is found the highest version will be selected link2GI::linkGRASS(meuse_sf,epsg = crs,quiet = FALSE) ``` Now do the same with `sf` based data. ```{r, eval=FALSE} require(link2GI) require(sf) # get data nc <- st_read(system.file("shape/nc.shp", package="sf")) terra::crs(nc) # Automatic search and find of GRASS binaries # using the nc sf data object for spatial referencing # This is the highly recommended linking procedure for on the fly jobs # NOTE: if more than one GRASS installation is found the highest version will be selected grass<-linkGRASS(nc,returnPaths = TRUE) ``` The second most common situation is to use an existing `GRASS` site and project, either with existing data sets or manually provided parameters. ```{r, eval=FALSE} require(link2GI) require(sf) # proj folders root_folder<-tempdir() paths<-link2GI::createFolders(root_folder = root_folder, folders = c("project1/")) # get data nc <- st_read(system.file("shape/nc.shp", package="sf")) # CREATE and link to a permanent GRASS folder at "root_folder", location named "project1" linkGRASS(nc, gisdbase = root_folder, location = "project1", quiet = FALSE) # ONLY LINK to a permanent GRASS folder at "root_folder", location named "project1" linkGRASS(gisdbase = root_folder, location = "project1", gisdbase_exist = TRUE, quiet = FALSE ) # setting up GRASS manually with spatial parameters of the nc data epsg = 28992 proj4_string <- sp::CRS(paste0("+init=epsg:",epsg)) linkGRASS(spatial_params = c(178605,329714,181390,333611,proj4_string@projargs),epsg=epsg,quiet = FALSE) # creating a GRASS gisdbase manually with spatial parameters of the nc data # additionally using a peramanent directory "root_folder" and the location "nc_spatial_params " epsg = 4267 proj4_string <- sp::CRS(paste0("+init=epsg:",epsg))@projargs linkGRASS(gisdbase = root_folder, location = "nc_spatial_params", spatial_params = c(-84.32385, 33.88199,-75.45698,36.58965,proj4_string),epsg = epsg) ``` ### Typical for specified search paths and OS The full disk search can be tedious, especially on Windows it can easily take 10 minutes or more. So it is helpful to specify a search path to narrow down the search. To search for `GRASS` installations in the home directory, you can use the following command. **Manual Linking Linux** ```{r, eval=FALSE} # Link the GRASS installation and define the search location linkGRASS(nc, search_path = "~/apps/otb911") ``` If you already did a full search and kow your installation fo example using the command `findGRASS` you can use the result directly for linking. ```{r, eval=FALSE} findGRASS() ``` ``` instDir version installation_type 1 /usr/lib/grass83 8.3.2 grass ``` ```{r, eval=FALSE} linkGRASS(nc,c("/usr/lib/grass83","8.3.2","grass"),epsg = 4267, quiet = FALSE) ``` **Manual Linking Windows** ```{r, eval=FALSE} # Link the GRASS installation and define the search location linkGRASS(nc, search_path = "C:", quiet = FALSE) ``` If you already did a full search and kow your installation fo example using the command `findGRASS` you can use the result directly for linking. ```{r, eval=FALSE} findGRASS() ``` ``` instDir version installation_type 1 C:/OSGeo4W 8.4.1 osgeo4w ``` ```{r, eval=FALSE} linkGRASS(nc,c("C:/OSGeo4W","8.4.1","osgeo4w"),epsg = 4267, quiet = FALSE) ``` ### Specific examples Finally, some more specific examples related to interactive selection or OS-specific settings. #### Manual version selection Manually select the `GRASS` installation and use the meuse `sf` object for spatial referencing. If you only have one installation it is directly selected. ```{r, eval=FALSE} linkGRASS(nc, ver_select = TRUE) ``` #### Creating a permanent gisdbase folder Create and link a permanent `GRASS` gisdbase (folder structure) in "~/temp3" with the default mapset "PERMANENT"" and the location "project1". Use the `sf` object for all spatial attributes. ```{r, eval=FALSE} linkGRASS(x = nc, gisdbase = "~/temp3", location = "project1") ``` #### Using a permanent gisdbase folder Link to the permanent `GRASS` gisdbase (folder structure) in "~/temp3" with the default mapset "PERMANENT" and the location named "project1". Use the formerly referencend nc `sf` object parameter for all spatial attributes. ```{r, eval=FALSE} linkGRASS(gisdbase = "~/temp3", location = "project1", gisdbase_exist = TRUE) ``` #### Manual Setup of the spatial attributes Setting up `GRASS` manually with spatial parameters of the meuse data ```{r, eval=FALSE} linkGRASS(spatial_params = c(178605,329714,181390,333611, "+proj=sterea +lat_0=52.15616055555555 +lon_0=5.38763888888889 +k=0.9999079 +x_0=155000 +y_0=463000 +no_defs +a=6377397.155 +rf=299.1528128 +towgs84=565.4171,50.3319,465.5524, -0.398957,0.343988,-1.8774,4.0725 +to_meter=1"),epsg = 28992) ``` # A typical use case for the Orfeo Toolbox wrapper link2GI supports the use of the Orfeo Toolbox with a simple list-based wrapper function. Actually, two functions parse the module and function syntax dumps and generate a command list that can be easily modified with the necessary arguments. If you have installed it in a user home directory you need to adrees this: Usually you have to get the module list first: ```{r, eval=FALSE} # link to the installed OTB Linux HOME directory otblink<-link2GI::linkOTB(searchLocation = "~/apps/") # get the list of modules from the linked version algo<-parseOTBAlgorithms(gili = otblink) algo <- link2GI::otb_capabilities(gili = otblink) ``` Based on the modules of the current version of `OTB', you can then select the module(s) you want to use. ```{r, eval=FALSE} ## for the example we use the edge detection, ## ------------------------------------------------------------ ## 1) Select an OTB algorithm by name pattern ## ------------------------------------------------------------ ## grep() returns *all* matching algorithm names as a character vector. ## This is intentional for exploration, but NOT valid for execution. algoKeyword <- grep("edge", algo, value = TRUE, ignore.case = TRUE) ## Inspect matches (important!) algoKeyword length(algoKeyword) ## ------------------------------------------------------------ ## 2) Select exactly ONE algorithm (mandatory) ## ------------------------------------------------------------ ## Explicitly pick one algorithm from the matches. ## This avoids ambiguity and guarantees a scalar character value. algo <- algoKeyword[[1]] ## ------------------------------------------------------------ ## 3) Read OTB help text (capabilities) ## ------------------------------------------------------------ ## This parses the OTB -help output and returns the raw help text. ## This is the *authoritative source* for parameters. caps <- link2GI::otb_capabilities(algo = algo, gili = otblink) ## Print the help text to understand available parameters cat(paste(caps$text, collapse = "\n")) ## ------------------------------------------------------------ ## 4) Parse parameters into a structured table ## ------------------------------------------------------------ ## otb_args_spec() converts the help text into a data.frame ## with keys, types, defaults, and mandatory flags. spec <- link2GI::otb_args_spec(algo = algo, gili = otblink) ## Inspect the relevant parameter metadata spec[, c("key", "mandatory", "default", "class")] ## ------------------------------------------------------------ ## 5) Build a command template with valid parameters only ## ------------------------------------------------------------ ## otb_build_cmd() creates a named list with: ## - required parameters ## - optional parameters with defaults (if requested) ## - no execution yet cmd <- link2GI::otb_build_cmd( algo, gili = otblink, include_optional = "defaults", require_output = TRUE ) ## Inspect the command structure str(cmd) ## ------------------------------------------------------------ ## 6) Show the exact CLI command that WOULD be executed ## ------------------------------------------------------------ ## retCommand = TRUE prints the full otbApplicationLauncherCommandLine call ## without running it. This is crucial for transparency and debugging. cat(link2GI::runOTB(cmd, otblink, retCommand = TRUE), "\n") ``` This is a minimal discovery workflow that queries the linked OTB installation and returns the full list of available applications. The result is a plain character vector, so you can filter it (e.g., with grep()) and then inspect parameters for a single application using the introspection helpers. ```{r eval=FALSE} require(link2GI) require(terra) require(listviewer) # 0) Link OTB otblink <- link2GI::linkOTB(searchLocation = "~/apps/") root_folder <- tempdir() fn <- system.file("ex/elev.tif", package = "terra") # 1) Choose the application (must be a single character scalar) algoKeyword <- "EdgeExtraction" # 2) Create a command template with valid keys (defaults included) cmd <- link2GI::otb_build_cmd( algo = algoKeyword, gili = otblink, include_optional = "defaults", require_output = TRUE ) # 3) Set mandatory arguments (same values as legacy example) cmd[["in"]] <- fn cmd[["filter"]] <- "touzi" cmd[["channel"]] <- "1" # 4) Set explicit on-disk output (recommended API: otb_set_out) out_file <- file.path(root_folder, paste0("out", cmd[["filter"]], ".tif")) cmd <- link2GI::otb_set_out(cmd, gili = otblink, key = "out", path = out_file) # 5) Run the algorithm and read output as a raster retStack <- link2GI::runOTB(cmd, gili = otblink, retRaster = TRUE) # 8) Plot result plot(retStack) ``` # Usecases presented on the GEOSTAT August 2018 > **Important note** > > The use cases presented at **GEOSTAT (August 2018)** are **outdated**. > They reflect the software ecosystem, interfaces, and platform assumptions > available at that time and should **not** be interpreted as representing > the current state of `link2GI`, GRASS GIS, or related toolchains. During the GEOSTAT 2018 (see https://opengeohub.org) in Prague some more complex use cases have been presented. ## Find slides and materials - [Presentation slides](https://gisma.github.io/link2gi2018/link2gi2018.html#1) - [Github Repository](https://github.com/gisma/link2gi2018) ## The examples - Basic usage of `SAGA` and `OTB` calls - [SAGA & OTB basic usecase](https://github.com/gisma/link2gi2018/blob/master/R/usecases/saga-otb/useCaseSAGA-OTB.R) - Wrapping a [GRASS GIS example](https://neteler.gitlab.io/grass-gis-analysis/02_grass-gis_ecad_analysis/) of Markus Neteler as presented on GEOSTAT 2018 - [Analysing the ECA&D climatic data - reloaded](https://github.com/gisma/link2gi2018/blob/master/R/usecases/grass/useCaseGRASS-Neteler2018.R) - Performing a `GRASS` based cost analysis on a huge cost raster - [Beetle spread over high asia](https://github.com/gisma/link2gi2018/blob/master/R/usecases/cost-analysis/useCaseBeetle.R) - Deriving a canopy height model using a mixed API approach - [Canopy Height Model from UAV derived point clouds](https://github.com/gisma/link2gi2018/blob/master/R/usecases/uav-pc/useCaseCHM.R)