

Client library development guidelines

This is a proposal for workflows that client libraries should support to make the experience with each library similar and users can easily adopt examples and workflows.

For best experience libraries should still embrace best practices common in their environments. This means clients can...

• choose which kind of casing they use (see below).
• feel free to implement aliases for methods.

Conventions

Casing

Clients can use snake_case, camelCase or any method used commonly in their environment. For example, the API request to get a list of collections can either be names get_collections or getCollections. This applies for all names, including scopes, method names and parameters.

Scopes

Each method belongs to a scope. To achieve this in object-oriented (OO) programming languages, methods would be part of a class. If programming languages don't support scopes, you may need to simulate it somehow to prevent name collisions, e.g. by adding a prefix to the method names (like in the "procedural style" example below). Best practices for this will likely evolve over time.

Example for the version method in openEO:

• Procedural style: openeo_version()
• Object-oriented style:

  OpenEO obj = new OpenEO();
  obj.version();
  

If you can't store scope data in an object, you may need to pass these information as argument(s) to the method.

Example:

• Procedural style:

  $connection = openeo_connect("https://openeo.org");
  openeo_get_capabilities($connection);
  

• Object-oriented style:

  OpenEO obj = new OpenEO();
  Connection con = obj.connect("https://openeo.org");
  con.getCapabilities();
  

Scope categories

Each scope is assigned to a scope category, of which there are three:

• Root category: Contains only the scope openEO.
• API category: Mostly methods hiding API calls to the back-ends. Methods may be implemented asynchronously. Contains the scopes Connection, File, Job, ProcessGraph, Service.
• Content: Mostly methods hiding the complexity of response content. Methods are usually implemented synchronously. Currently contains only the scope Capabilities. Method names should be prefixed if name collisions are likely.

Method names across ALL the scopes that belong to the root or API categories MUST be unique. This is the case because the parameter in hasFeature(method_name) must be unambiguous.

Method names of scopes in the Content category may collide with method names of scopes in the root/API categories, as is the case with version() (relates to (1) the client library version in openEO scope and (2) the API version in Connection scope).

Parameters

The parameters usually follow the request schemas in the openAPI specification. The parameters should follow their characteristics, for example regarding the default values.

Some methods have a long list of (optional) parameters. This is easy to implement for languages that support named parameters such as R. Other languages may have problems implementing this natively as they need to fill many parameters with default values. For example creating a job in R with a budget would lead to such a method call: createJob(process_graph = ..., null, budget = 123), but in PHP it would be: createJob(..., null, null, null, null, null, 123). This is not an ideal behaviour, therefore client developers might want to consider passing parameters in coupled in a dictionary or class to emulate named parameters. The example in PHP could be improved to createJob([process_graph => ..., null, budget => 123]).

ToDo: Allow sorting and other useful operations for lists?

Method mappings

Note: Subscriptions and some scopes for response JSON objects are still missing. We are open for proposals.

Parameters with a leading ? are optional.

Scope: openEO (root category)

Description	Client method
Connect to a back-end, including authentication. Returns Connection.	connect(url, ?auth_type, ?auth_options)
Get client library version.	version()

Parameters

• auth_type in connect: null, basic or oidc (non-exclusive). Defaults to null (no authentication).
• auth_options in connect: May hold additional data for authentication, for example a username and password for basic authentication.

Scope: Connection (API category)

Description	API Request	Client method
Get the capabilities of the back-end. Returns Capabilities.	GET /	capabilities()
List the supported output file formats.	GET /output_formats	listFileTypes()
List the supported secondary service types.	GET /service_types	listServiceTypes()
List all collections available on the back-end.	GET /collections	listCollections()
Get information about a single collection.	GET /collections/{name}	describeCollection(name)
List all processes available on the back-end.	GET /processes	listProcesses()
Authenticate with OpenID Connect (if not specified in connect).	GET /credentials/oidc	authenticateOIDC(?options)
Authenticate with HTTP Basic (if not specified in connect).	GET /credentials/basic	authenticateBasic(username, password)
Get information about the authenticated user.	GET /me	describeAccount()
Lists all files from a user. Returns a list of File.	GET /files/{user_id}	listFiles(?user_id)
Creates a (virtual) file. Returns a File.	None	createFile(path, ?user_id)
Validates a process graph.	POST /validate	validateProcessGraph(process_graph)
Lists all process graphs of the authenticated user. Returns a list of ProcessGraph.	GET /process_graphs	listProcessGraphs()
Creates a new stored process graph. Returns a ProcessGraph.	POST /process_graphs	createProcessGraph(process_graph, ?title, ?description)
Executes a process graph synchronously.	POST /preview	execute(process_graph, ?output_format, ?output_parameters, ?budget)
Lists all jobs of the authenticated user. Returns a list of Job.	GET /jobs	listJobs()
Creates a new job. Returns a Job.	POST /jobs	createJob(process_graph, ?output_format, ?output_parameters, ?title, ?description, ?plan, ?budget, ?additional)
Lists all secondary services of the authenticated user. Returns a list of Service.	GET /services	listServices()
Creates a new secondary service. Returns a Service.	POST /services	createService(process_graph, type, ?title, ?description, ?enabled, ?parameters, ?plan, ?budget)

Parameters

• user_id in listFiles and createFile: Defaults to the user id of the authenticated user.
• options in authenticateOIDC: May hold additional data required for OpenID connect authentication.

Scope Capabilities (Content category)

Should be prefixed with Capabilities if required. In non-object-oriented paradigms it is likely required as version() in this scope and the scope OpenEO could collide. For example, version() in this scope could be named openeo_capabilities_version() in procedural style.

Description	Field	Client method
Get openEO version.	version	version()
List all supported features / endpoints.	endpoints	listFeatures()
Check whether a feature / endpoint is supported.	endpoints > ...	hasFeature(method_name)
Get default billing currency.	billing > currency	currency()
List all billing plans.	billing > plans	listPlans()

Parameters

• method_name in hasFeature: The name of a client method in any of the scopes that are part of the API category. E.g. hasFeature("describeAccount") checks whether the GET /me endpoint is contained in the capabilities response's endpoints object.

Scope: File (API category)

The File scope internally knows the user_id and the path.

Description	API Request	Client method
Download a user file.	GET /files/{user_id}/{path}	downloadFile(target)
Upload a user file.	PUT /files/{user_id}/{path}	uploadFile(source)
Delete a user file.	DELETE /files/{user_id}/{path}	deleteFile()

Parameters

• target in downloadFile: Path to a local file or folder.

Scope: Job (API category)

The Job scope internally knows the job_id.

Description	API Request	Client method
Get all job information.	GET /jobs/{job_id}	describeJob()
Update a job.	PATCH /jobs/{job_id}	updateJob(?process_graph, ?output_format, ?output_parameters, ?title, ?description, ?plan, ?budget, ?additional)
Delete a job	DELETE /jobs/{job_id}	deleteJob()
Calculate an time/cost estimate for a job.	GET /jobs/{job_id}/estimate	estimateJob()
Start / queue a job for processing.	POST /jobs/{job_id}/results	startJob()
Stop / cancel job processing.	DELETE /jobs/{job_id}/results	stopJob()
Get document with download links.	GET /jobs/{job_id}/results	listResults(?type)
Download job results.	GET /jobs/{job_id}/results > ...	downloadResults(target)

Parameters

• type in listResult: Either json or metalink (non-exclusive). Defaults to json.
• target in downloadResults: Path to a local folder.

Scope: ProcessGraph (API category)

The ProcessGraph scope internally knows the pg_id (process_graph_id).

Description	API Request	Client method
Get all information about a stored process graph.	GET /process_graphs/{pg_id}	describeProcessGraph()
Update a stored process graph.	PATCH /process_graphs/{pg_id}	updateProcessGraph(?process_graph, ?title, ?description)
Delete a stored process graph.	DELETE /process_graphs/{pg_id}	deleteProcessGraph()

Scope: Service (API category)

The Service scope internally knows the service_id.

Description	API Request	Client method
Get all information about a secondary web service.	GET /services/{service_id}	describeService()
Update a secondary web service.	PATCH /services/{service_id}	updateService(?process_graph, ?title, ?description, ?enabled, ?parameters, ?plan, ?budget)
Delete a secondary web service.	DELETE /services/{service_id}	deleteService()

Processes

The processes a back-end supports may be offered by the clients as methods in its own scope. The method names should follow the process names, but the conventions listed above can be applied here as well, e.g. converting filter_bands to filterBands. As parameters have no natural or technical ordering in the JSON objects, clients must come up with a reasonable ordering of parameters if required. This could be inspired by existing clients. The way of building a process graph from processes heavily depends on the technical capabilities of the programming language. Therefore it may differ between the client libraries. Follow the best practices of the programming language, e.g. support method chaining if possible.

Workflow example

Some simplified example workflows using different programming styles are listed below. The following steps are executed:

1. Loading the client library.
2. Connecting to a back-end and authenticating with username and password via OpenID Connect.
3. Requesting the capabilities and showing the implemented openEO version of the back-end.
4. Showing information about the "Sentinel-2A" collection.
5. Showing information about all processes supported by the back-end.
6. Building a simple process graph.
7. Creating a job.
8. Pushing the job to the processing queue.
9. After a while, showing the job details, e.g. checking the job status.
10. Once processing is finished, downloading the job results to the local directory /tmp/job_results/.

R (functional style)

library(openeo)

con = connect("https://openeo.org", "username", "password")
cap = capabilities()
cap %>% version()
con %>% describeCollection("Sentinel-2A")
con %>% listProcesses()

processgraph = process("get_collection", name = "Sentinel-2A") %>% 
  process("filterBbox", west = 672000, south = 5181000, east = 652000, north = 5161000, crs = "EPSG:32632") %>%
  process("filterDaterange", extent = c("2017-01-01T00:00:00Z", "2017-01-31T23:59:59Z")) %>%
  process("NDVI", nir = "B4", red = "B8A") %>%
  process("minTime")

job = con %>% createJob(processgraph)
job %>% startJob()
job %>% describeJob()
job %>% downloadResults("/tmp/job_results/")

Python (mixed style)

import openeo

con = openeo.connect("https://openeo.org", "username", "password")
cap = con.capabilities()
print cap.version()
print con.describe_collection("Sentinel-2A")
print con.list_processes()

processes = con.get_processes();
pg = processes.get_collection(name = "Sentinel-2A");
pg = processes.filter_bbox(pg, west = 672000, south = 5181000, east = 652000, north = 5161000, crs = "EPSG:32632")
pg = processes.filter_daterange(pg, extent = ["2017-01-01T00:00:00Z", "2017-01-31T23:59:59Z"])
pg = processes.NDVI(pg, nir = "B4", red = "B8A")
pg = processes.min_time(pg)

job = con.create_job(pg.graph)
job.start_job()
print job.describe_job()
job.download_results("/tmp/job_results/")

Java (object oriented style)

import org.openeo.*;

OpenEO obj = new OpenEO();
Connection con = obj.connect("https://openeo.org", "username", "password");
Capabilities cap = con.capabilities();
System.out.println(cap.version());
System.out.println(con.describeCollection("Sentinel-2A"));
System.out.println(con.listProcesses());

ProcessGraphBuilder pgb = con.getProcessGraphBuilder()
// Chain processes...
ProcessGraph processGraph = pgb.buildProcessGraph();

Job job = con.createJob(processGraph);
job.startJob();
System.out.println(job.describeJob());
job.downloadResults("/tmp/job_results/");

PHP (procedural style)

require_once("/path/to/openeo.php");

$connection = openeo_connect("http://openeo.org", "username", "password");
$capabilities = openeo_capabilities($connection);
echo openeo_capabilities_version($capabilites);
echo openeo_describe_collection($connection, "Sentinel-2A");
echo openeo_list_processes($connection);

$pg = openeo_process($pg, "get_collection", ["name" => "Sentinel-2A"]);
$pg = openeo_process($pg, "filter_bbox", ["west" => 672000, "south" => 5181000, "east" => 652000, "north" => 5161000, "crs" => "EPSG:32632"]);
$pg = openeo_process($pg, "filter_daterange", ["extent" => ["2017-01-01T00:00:00Z", "2017-01-31T23:59:59Z"]]);
$pg = openeo_process($pg, "NDVI", ["red" => "B4", "nir" => "B8A"]);
$pg = openeo_process($pg, "min_time");

$job = openeo_create_job($connection, $pg);
openeo_start_job($job);
echo openeo_describe_job($job);
openeo_download_results($job, "/tmp/job_results/");