다음을 통해 공유


DeployR API Overview

Applies to: DeployR 8.x (See comparison between 8.x and 9.x)

Looking to deploy with Machine Learning Server? Start here.

The DeployR API exposes the R platform as a service allowing the integration of R statistics, analytics, and visualizations inside Web, desktop and mobile applications. This API is exposed by the DeployR server, a standards-based server technology capable of scaling to meet the needs of enterprise-grade deployments.

With the advent of DeployR, the full statistics, analytics and visualization capabilities of R can now be directly used inside Web, desktop and mobile applications.

Looking for a specific API call? Look in this online API reference.

R for Application Developers

While data scientists can work with R directly in a console window or IDE, application developers need a different set of tools to use R inside applications. The DeployR API exposes R analytics Web services, making the full capabilities of R available to application developers on a simple yet powerful Web services API.

As an application developer integrating with DeployR-managed analytics Web services, typically your interest is in executing R code, not writing it. Data scientists with R programming skills write R code. With one-click in the DeployR Repository Manager, this R code can be turned into a DeployR-managed analytics Web service. Once R code is exposed by DeployR as a service, an application can make API calls to pass inputs to the service, execute the service, and retrieve outputs from the service. Those outputs can include R object data, R graphics output such as plots and charts, and any file data written to the working directory associated the current R session.

Each time a service is executed on the API, the service makes use of an R session that is managed by DeployR as a project on behalf of the application. Depending on the nature and requirements of your application, you can choose to execute services on anonymous or authenticated projects.

To simplify integration of R analytics Web services using the DeployR API, we provide several client libraries, which are currently available for Java, JavaScript and .NET developers. A major benefit of using these client libraries is that they simplify making calls, encoding data, and handling response markup on the API.

Users on the API

The User APIs exist principally to facilitate user authentication with the DeployR server. Additionally, the /r/user/autosave call can be used to enable or disable autosave semantics on persistent projects for the duration of the users HTTP session.

Authenticated Users

One of the first steps for most typical applications using this API is to provide a mechanism for users to authenticate with the DeployR server by signing-in and signing-out of the application.

To sign in, a user must provide username and password credentials. These credentials are verified by the DeployR server using the /r/user/login call. Credentials are matched against user account data stored in the DeployR database or against user account data stored in LDAP, Active Directory, or PAM directory services.

If these credentials are verified by the DeployR server, then we say that the user is an authenticated user. An authenticated user is granted access to the full API, allowing the user to work on projects, submit or schedule jobsand work with repository-managed files and scripts.

Pre-Authenticated Users

There are situations where a user may have been reliably authenticated by some external system such as SiteMinder prior to accessing a DeployR-enabled Web application. These types of pre-authentication scenarios are now supported by the server. Consequently, DeployR-enabled Web applications should omit sign-in and sign-out logic when pre-authenticated users are expected.

Anonymous Users

While many require the security and control associated with authenticated users, there are cases when an application may want to offer specific services to users without ever establishing a formal verification of the user's identity.

In such situations, we say that the user is an anonymous user. Typically an anonymous user is an unauthenticated visitor to a DeployR-enabled Web application.

Anonymous users are only granted access to a single API call, /r/repository/script/execute.


Projects on the API

A project on the DeployR API represents an R session.

Most R users are accustomed to working with R interactively in an R console window. In this environment, users can input commands to manipulate, analyze, visualize, and interpret object and file data in the R session. The set of objects in the R session is collectively known as the workspace. The set of files in the R session is collectively known as the working directory. The R session environment also supports libraries, which allows the functionality found in R packages to be loaded by the user on-demand.

The DeployR environment supports this same set of functionalities by introducing the concept of projects on the API. As with working in an R console window, all operations on project APIs are synchronous, where requests are processed serially and blocked until completion. The project management APIs facilitate basic lifecycle management for projects.

DeployR supports a number of different types of projects, each of which is designed to support distinct workflows within client applications.

The following sections discuss the different types of projects available:


### Anonymous Projects

An anonymous project is a project created by an anonymous user. There are two types of anonymous project: stateless projects and HTTP blackbox projects. The types of projects can be created using the following API calls:

API Call Description
/r/repository/script/execute Executes a repository-managed script on an anonymous project
/r/repository/script/render Executes a repository-managed script on an anonymous project and renders outputs to HTML

Anonymous users are not permitted to work directly with the Project APIs. Those APIs are only available to authenticated users.

Stateless Projects

Stateless projects are useful when an application needs to make a one-off request on an R session. It may help to think of a stateless project as a disposable R session, that is used once and then discarded. Once the service request has responded with the results of the execution the R session is shutdown and permanently deleted by the server. There are many types of application workflows that can benefit from working with stateless projects.

HTTP Blackbox Projects

With the introduction of HTTP blackbox projects, stateful R sessions are now available to anonymous users. HTTP blackbox projects are useful when an application needs to maintain the same R session for the duration of an anonymous user's HTTP session. Once the HTTP session expires or the server detects that the anonymous user has been idle (default: 15-minutes idle timeout) the HTTP blackbox project and all associated state are permanently deleted by the server. There can only be one HTTP blackbox project live on an HTTP session at any given time.

To execute an R script on an HTTP blackbox project, enable the blackbox parameter on the following calls: /r/repository/script/execute and /r/repository/script/render

There are two types of data that can be returned when executing an R script on an HTTP blackbox project:

  • Unnamed plots generated by the R graphics device on the execution are returned as results
  • Named workspace objects identified on the robjects parameter can be returned as DeployR-encoded R objects

All files in an HTTP blackbox project's working directory are completely hidden from the user.

These calls also support a recycle parameter that can be used when working with HTTP blackbox projects. When this parameter is enabled, the underlying R session is recycled before the execution occurs. Recycling an R session deletes all R objects from the workspace and all files from the working directory. The ability to recycle an HTTP blackbox project gives an anonymous user control over the R session lifecycle.

To interrupt an execution on the HTTP blackbox project on the current HTTP session, use the /r/repository/script/interrupt.


### Authenticated Projects

An authenticated project is a project created by an authenticated user. There are three types of authenticated project:

  • Temporary Project
  • User Blackbox Project
  • Persistent Project

Authenticated projects can be created using the following API calls:

API Call Description
/r/project/create Creates a new temporary, user blackbox or persistent project
/r/project/pool Creates a pool of temporary projects

Authenticated users own projects. Each authenticated user can create zero, one or more projects. Each project is allocated its own workspace and working directory on the server and maintains its own set of R package dependencies along with a full R command history. A user can execute R code on a project using the /r/project/execute/code and /r/project/execute/script calls and retrieve the R command history for the project using the /r/project/execute/history call.

Given an authenticated user can do everything an anonymous user can do on the API, it is possible for an authenticated user to create and work with anonymous projects as described in the section Anonymous Projects.

Temporary Projects

Temporary projects are useful when an application needs to maintain the same R session for the duration of an authenticated user's session within the application. Once the user makes an explicit call on /r/project/close or signs-out of the application the temporary project and all associated state are permanently deleted by the server.

A user can create a temporary project by omitting the name parameter on the /r/project/create call or they can create a pool of temporary projects using the /r/project/pool call.

All unnamed projects are temporary projects.

User Blackbox Projects

A user blackbox project is a special type of temporary project that limits API access on the underlying R session. User blackbox projects are most useful when an application developer wants to associate a temporary project with a user without granting that user unfettered access to the underlying R session.

There are two types of data that can be returned when executing an R script on a user blackbox project:

  • Unnamed plots generated by the R graphics device on the execution are returned as results
  • Named workspace objects identified on the robjects parameter can be returned as DeployR-encoded R objects

All files in a user blackbox project's working directory are completely hidden from the user.

User blackbox projects can be created using the blackbox parameter on the following API calls:

API Call Description
/r/project/create Creates a new project
/r/project/pool Creates a pool of temporary projects

User blackbox projects permit only the following subset of project-related API calls:
API Call Description
/r/project/execute/script Executes a script on project
/r/project/execute/interrupt Interrupts a code execution on project
/r/project/ping Pings a project
/r/project/recycle Recycles a project R session
/r/project/close Closes a project

This set of API calls on user blackbox projects is collectively known as the User Blackbox API Controls.

A user blackbox project also ensure that none of the R code from the R scripts that are executed on that project is ever returned in the response markup.

Persistent Projects

Persistent projects are useful when an application needs to maintain the same R session across multiple user sessions within the application. A persistent project is stored indefinitely in the server unless it is explicitly deleted by the user.

Each time a user executes R code on their persistent project, the accumulated R objects in the workspace and files in the working directory from previous service executions remain available.

Whenever users return to the server, their own persistent projects are available so that users can pick up where they last left off. In this way, persistent projects can be developed over days, weeks, months even years.

The server stores the following state for each persistent project:

  • Project workspace
  • Project working directory
  • Project package dependencies
  • Project R command history and associated results

An authenticated user can create a persistent project by specifying a value for the name parameter on the /r/project/create call. Alternatively, if a user is working on a temporary project then that project can become persistent once the user makes a call on /r/project/save which has the effect of naming the project.

All named projects are persistent projects.


### Live Projects on the Grid

A live project is any project, temporary or persistent, that is actively in use on the DeployR grid. By definition, all temporary projects are live projects. A user can have zero, one or more projects live on the grid at any one time.

To ensure a fair distribution of server runtime resources across all users, the DeployR administrator can limit the number of live projects permitted for a given user at any one time. When a user reaches his or her live project limit, all further API calls on new projects are rejected until one or more of that user's current live projects are closed using the /r/project/close call. Calls that are rejected due to live project limits indicate an appropriate Grid Resource Error on the response code.


### Working with Project Management APIs

Creating Projects

There are a number of ways to create projects on the API, including the following:

API Call Description
/r/project/create Create a new project
/r/project/saveas Clone an existing project
/r/project/import Import a project archive

Saving Projects

There are a number of ways to save projects on the API, including the following:

API Call Description
/r/project/save Save an existing project
/r/project/saveas Clone an existing project
/r/project/export Export a project archive

Closing Projects

To close a live project, use the /r/project/close API call. Closing a live project (temporary or persistent) deactivates the project, which in turn removes it from the grid and releases all associated server runtime resources. Remember, that closing a temporary project will permanently delete that project.

Deleting Projects

To delete a persistent project, use the /r/project/delete API call.

Project Archives

A project archive is generated by exporting the entire state of a persistent project into a compressed archive file using the /r/project/export API call. Once exported, this file resides outside of the DeployR server. Project archives are useful both as a general backup mechanism and as a means of sharing work between colleagues, particularly colleagues working on different DeployR installations.

A user can import a project archive back into any compatible DeployR server in order to produce a new, persistent project using the /r/project/import API call.


### Project Ownership & Collaboration

As noted, each authenticated user can create zero, one or more projects. Each newly created project is by default privately owned, and visible only to its creator. Such user-based privacy is a central aspect of the DeployR security model. However, there are scenarios in which a more flexible access model for projects would benefit users by facilitating collaboration among them.

Persistent Project Read-Only Collaboration

A project owner may wish to share a project by granting read-only access to fellow authenticated users on the same DeployR server. Once read-only access has been granted, it facilitates discussions, demonstrations and so on.

To facilitate these kinds of workflows the /r/project/about/update call allows users to toggle the read-only "shared" access controls on any of their private projects. Using this call to enable shared/read-only access on a project makes the project visible to all authenticated users. Using this call to disable shared/read-only access on a project makes the project invisible to all but the project owner(s).

Persistent Project Read-Write Collaboration

At times a project owner may also wish to grant read-write access on a project to fellow authenticated users sharing the same DeployR server so they can actively collaborate on the development of a project. This type of project collaboration would be beneficial, for example, as part of a planned multistage development or simply to permit work on the project to continue when the original project owner goes on vacation.

To facilitate these kinds of workflows a project owner can now grant authorship rights on a project to one or more authenticated users using the /r/project/grant call. Once a user has been granted authorship rights on a project they effectively co-own the project and can view, modify and even delete the project. Where multiple users have been granted authorship rights on a project the server ensures that only one user can be actively working on the project at any one time.

Deleting a project with multiple authors simply revokes authorship rights to the project for the caller. After the call, all other authors on the project still have full access to the project. If the last remaining author deletes a project, then the project is permanently deleted.


### Project Execution APIs

Callers can use the project execution APIs to facilitate the execution of R code and the retrieval of the project's execution history associated results.

R Code Execution

There are two ways to execute R code on a project:

Execute a block of R code directly using the /r/project/execute/code call. Execute a repository-managed script using the /r/project/execute/script call. Important! Any project execution can be interrupted using the /r/project/execute/interrupt call.

R Code Execution History

To retrieve the R code execution history on a project, use the /r/project/execute/history call.

R Code Execution Artifacts

Artifacts are files, that are explicitly named in R code and written to the project's working directory.

To retrieve an artifact on a project, use the /r/project/directory/download call.

R Code Execution Results

Results are files, that are implicitly named in R code, generated by the R graphics device and stored as part of a project's execution history.

To retrieve results on a project, use the /r/project/execute/result/download call.


### Project Workspace APIs

The project workspace APIs facilitate working with objects found in the project workspace.

Workspace Object Retrieval

Retrieve a summary list of objects found in the workspace using the /r/project/workspace/list call. Retrieve a DeployR-encoded representation of one or more objects found in the workspace using the /r/project/workspace/get call.

Workspace Object Creation

Typically objects are created in the workspace when commands are executed. However, there are a number of APIs that allow the direct creation of objects in the workspace:

Upload an object into the workspace from a file on the users computer using the /r/project/workspace/upload call. Transfer an object into the workspace from a URL using the /r/project/workspace/transfer call. Push an object into the workspace from a DeployR-encoded object using the /r/project/workspace/push call.

Workspace Objects and the Repository

Workspace objects can be stored directly to the repository using the /r/project/workspace/store call. Repository-managed files containing R binary objects can be loaded directly to the workspace using the /r/project/workspace/load call.


### Project Directory APIs

The project directory APIs facilitate working with files found in the project working directory.

Directory File Retrieval

API Call Description
/r/project/directory/list Retrieve a summary list of files found in the working directory
/r/project/directory/download Retrieve individual files found in the working directory
/r/project/directory/download Retrieve a compressed archive containing all files found in the working directory

Directory File Creation

Typically files are created in the working directory when commands are executed. However, there are a number of APIs that allow the direct creation of files in the working directory:

API Call Description
/r/project/directory/upload Upload a file into the working directory from a file on the users computer
/r/project/directory/transfer Transfer a file into the working directory from a URL
/r/project/directory/write Write a text file into the working directory

Directory Files and the Repository
API Call Description
/r/project/directory/store Store directory files directly to the repository
/r/project/directory/load Load repository-managed files directly into the working directory

### Project Package APIs

The project package APIs facilitate R package dependency management for projects.

Directory File Retrieval

API Call Description
/r/project/package/list Retrieve a list of all packages loaded for a project
/r/project/package/attach Load individual packages for a project
/r/project/package/detach Unload individual packages for a project

Jobs on the API

Working with R interactively in an R console window or working with projects on the API are both examples of synchronous working environments where a user can make a request that is blocked until processing completes and an appropriate response is generated and returned. When working with R interactively, the response is displayed as output in the R console window. When working with projects on the API, the response is well-formed markup on the response stream.

However, there are times when it can be advantageous to permit users to make requests without requiring that they wait for responses. For example, consider long-running operations that could take hours or even days to complete.

The DeployR environment supports these types of long-running operations by introducing the concept of jobs on the API. DeployR managed jobs support the execution of commands in the background on behalf of users.

The Job APIs facilitate working with DeployR managed jobs. Jobs support the execution of commands in the background on behalf of users.

Submitting Versus Scheduling Jobs

Jobs can be submitted for immediate background execution using the /r/job/submit call. Jobs can be scheduled for execution at some later date or on a recurring basis using the /r/job/schedule call.

Job Scheduling Priority

When a user submits or schedules a job, they can identify a high, medium, or low (default) priority for their job. Each priority level has its own dedicated jobs queue managed by the jobs scheduler.

Jobs in higher priority job queues are guaranteed to be executed on the grid ahead of jobs in lower priority job queues. Put another way, jobs in higher priority job queues are assigned free slots on the grid ahead of jobs on lower priority queues. Jobs on the same jobs queue are executed on the grid in the order in which they were submitted.

High, medium, and low priority job queues allow users to expedite higher priority jobs by 'jumping the queue', and to expedite jobs to best fit specific scheduling needs.

Retrieving Job Results

Users can query the status of their jobs using the /r/job/query call that returns the current status for a given job. Users can cancel a scheduled job or an actively running job using the /r/job/cancel call.

Regardless of whether a job runs to completion or whether a job is interrupted or aborted by the user the DeployR server creates a new project for the user by default. This project will be named after the job that created it and contains a complete workspace, working directory, package dependencies, command history, and associated results generated by that job. The user can then work with the full set of Project APIs to manipulate, analyze, visualize, and interpret all of the objects, files, and generated plots resulting from the job.

However, if the storenoproject parameter is set to true when submitting the job then a project is not created when the job completes. Using the storenoproject parameter can be very resource efficient and improve throughput when the only data you care about has already been stored in the repository using the storeobjects and/or storefiles parameters.

If the storenoproject parameter is not set to true, then a project is always created to hold the results of a job unless the job experiences a runtime error during execution, for example, the R code being executed on the job has syntax errors or attempts to reference undefined objects. In such cases, see the statusMsg property on a failed job for an indication of the underlying cause of the problem.

All completed, interrupted, and aborted jobs result in new projects for users.

Understanding the Jobs Lifecycle

The following diagram provides an overview of the full set of Job APIs and indicates the lifecycle of jobs managed by those APIs:

Repository on the API

The DeployR environment offers each authenticated user their own versioned file storage by introducing the concept of a repository on the API. The repository provides a persistent store for user files of any type, including binary object files, plot image files, data files, simple text files, and files containing blocks of R code that are referred to as repository-managed scripts.

Because the repository supports a versioned file system, a full version history for each file is maintained and any version of a file can be retrieved upon request.

By default, each private repository store exposes a single repository directory, known as the root directory.

Users can store files in and retrieve files from their root directory. In addition, users can manage their own set of custom and archived directories. Working with the root, custom and archived directories is detailed in the following section.


### Repository Directories The Repository Directory APIs facilitate working with repository-managed directories. Repository-managed directories fall into two distinct categories: + User Directories + System Directories
#### User Directories Each authenticated user maintains their own set of private user directories. The set of user directories can be further divided into the following user directory types:
  • Root directory
  • Custom directories
  • Archived directories

By default, each authenticated user has access to their own root directory in the repository. When working with repository Files, all API calls default to operating on the root directory if a specific custom or archived directory is not specified on the directory parameter.

Each authenticated user can also create zero, one or more custom directories. Each custom directory becomes a subdirectory of the root directory. Custom directories can be useful to help manage logical groups of files across multiple projects, customers, applications etc.

Users can also archive the contents of their root directory or the contents of any of their custom directories within archived directories. This can be a useful way to preserve old but valuable files over time while reducing day-to-day directory clutter.

By default, archived directories do not appear in the response markup when a user makes a call to /r/repository/directory/list. However, when the archived parameter is enabled on that call, any archived directories belonging to the user appears in the response markup.

System Directories

System directories are virtual directories that exist only in the context of the response markup on the /r/repository/directory/list call. These directories are provided as a convenient way for a user working with the Repository Directory APIs to get access to repository-managed files that are shared or made public by other users.

There are just three system directories:

  • Restricted directory. The Restricted system directory maintains a list of files that have been shared-with-restrictions by other authenticated users. These files have had their access controls set to restricted access.

  • Shared directory. The Shared system directory maintains a list of files that have been shared by other authenticated users. These files have had their access controls set to shared access.

  • Public directory. The Public system directory maintains a list of files that have been published by other authenticated users. These files have had their access controls set to public access.

Important: As all files appearing in the system directories belong to other users, these files cannot be copied, moved, modified, archived, or deleted using the Repository Directory APIs.


### Repository-Managed Files

Each user has access to a private repository store. Each file placed in that store is maintained indefinitely by the server unless it is explicitly deleted by the user.

Repository-managed files can be easily loaded by users directly into anonymous projects, authenticated projects as well as into jobs. For example, a binary object file in the repository can be loaded directly into a project workspace using the /r/project/workspace/load call. A data file can be loaded directly into a project working directory using the /r/project/directory/load call.

Conversely, objects in a project workspace and files in a project working directory can be stored directly to the repository using the /r/project/workspace/store and /r/project/directory/store calls respectively.

Repository-managed files can also be loaded directly into the workspace or working directory ahead of executing code on anonymous projects or an asynchronous job.

Repository File Ownership & Collaboration

By default, files in a user's private repository are visible only to that user. Such user-based privacy is a central aspect of the DeployR security model. However, there are situations that can benefit from a more flexible access model for repository-managed files.

Repository File Read-Only Collaboration

A user may wish to share a file in their private repository by granting read-only access to fellow authenticated and/or anonymous users. Once read-only access has been granted to users, they can download and view the file. In the case of a repository-managed script users can also execute the file.

To facilitate these kinds of workflows the /r/repository/file/update call allows users to toggle the read-only access controls on any of their repository-managed files.

Using this call to enable shared access on a file makes the file visible to authenticated users. Using this call to enable published access on a file makes the file visible to authenticated and anonymous users.

Repository File Read-Write Collaboration

At times the owner of a repository-managed file may also wish to grant read-write access on a file so that fellow authenticated users who are sharing the same DeployR server can actively collaborate on the development of a file. For example, consider a situation where the owner of a repository-managed script wishes to collaborate more closely on its design and implementation with a select group of authenticated users.

To facilitate this kind of collaborative workflow a user can now grant authorship rights on a repository file to one or more authenticated users using the /r/repository/file/grant call. Once a user has been granted authorship rights on a repository-managed file the user now co-owns the file and has the ability to download, modify and even delete the file.

Important

When modifying a repository-managed file with multiple authors it is the responsibility of the users in the collaborative workflow to ensure the overall consistency of the file. Deleting a repository-managed file with multiple authors simply revokes authorship rights to the file for the caller. After the call all other authors on the file will still have full access to the file. If the last remaining author deletes a repository-managed file, then the file is permanently deleted.

Repository File APIs

The Repository File APIs facilitate working with repository-managed files. Because the repository supports versioned file storage, multiple versions of any file can exist at any one time and any version can be retrieved upon request.

Each repository-managed file is uniquely identified on the API by the following parameters:

  • filename - repository file name
  • author - repository file author
  • directory - (optional) repository directory name, if omitted, defaults to root directory.
  • version - (optional) repository file version

Access to each repository-managed file is controlled by a set of properties on the file: restricted, shared, and published respectively. These properties can be used to control the visibility of the file to other users, both authenticated and anonymous users. There are four distinct access control levels:

  • Private - the default access level, the file is visible to it's author(s) only.
  • Restricted - the file is visible to authenticated users that have been granted one or more of the roles indicated on the restricted property of the file.
  • Shared - the file is visible to all authenticated users when the shared property is true.
  • Public - the file is visible to all authenticated and all anonymous users when the published property is true.

Repository File Retrieval

API Call Description
/r/repository/file/list Retrieve a list of the files in the repository
/r/repository/file/download Retrieve individual files in the repository
/r/repository/directory/list Retrieve a list of files per repository-managed directory
/r/repository/directory/download Retrieve one or more files from a repository-managed directory

Repository File Creation
API Call Description
/r/repository/file/upload Upload a file into the repository from the users computer
/r/repository/file/transfer Transfer a file into the repository from a URL
/r/repository/file/write Write a text file into the repository

Repository File Version Control
API Call Description
/r/repository/file/list Retrieve a list of the files in the repository
/r/repository/file/diff Retrieve a generated diff between any version of a text-based file and the latest version of that file
/r/repository/file/revert Revert to an older version of specific file in the repository

### Repository-Managed Scripts

Repository-managed scripts are a special type of repository-managed file. Any file with a .r or a .R extension is identified by the server as a repository-managed script.

These scripts are essentially blocks of R code with well-defined inputs and outputs. While scripts are technically also repository-managed files, they are designed to be exposed as an executable on the API.

Scripts can be created, managed, and deployed using the standard Repository APIs or directly within the* DeployR Repository Manager*. Refer to the Repository Manager Help for further details.

Authenticated users can execute scripts within the context of any project using the /r/project/execute/script call. Both authenticated and anonymous users can execute scripts within the context of anonymous projects using the /r/repository/script/execute and /r/repository/script/render calls.

Repository-managed scripts are files in the repository so all API calls described in the section Working with Repository Files are available to create and manage repository-managed scripts.

Working with the Repository Script APIs

The Repository Script APIs provide some script-specific functionality for repository-managed scripts. Scripts are blocks of R code with well-defined inputs and outputs. While scripts are technically also repository-managed files, scripts differ from other repository-managed files as they perform a specific function that can be exposed as an executable on the API.

Important! Repository-managed scripts are files in the repository so all API calls described in the section Working with Repository Files are available to create and manage repository-managed scripts.


## Event Stream

The event stream API is unique within the DeployR API as it supports push notifications from the DeployR server to client applications. Notifications correspond to discrete events that occur within the DeployR server. Rather than periodically polling the server for updates, a client application can subscribe once to the event stream and then receive event notifications pushed by the server.

There are four distinct event categories:

  • Stream Lifecycle events
  • Execution events
  • Job Lifecycle events
  • Management events

### Stream Lifecycle Events Stream lifecycle events occur when a client application successfully connects to or disconnects from an event stream on the server. There are two stream lifecycle events:
  • streamConnectEvent. The data passed on the streamConnectEvent allows the client application to determine the type of event stream connection established with the server: authenticated, anonymous, or management event stream. The nature of an event stream connection determines the nature of the events that can be delivered on that stream. More details can be found on event stream types below.

  • streamDisconnectEvent. The streamDisconnectEvent informs a client application that the event stream has been disconnected by the server. Following a streamDisconnectEvent no further events arrive at the client on that event stream.


### Execution Events Execution events occur when an R session is executing R code (on behalf of an anonymous project, authenticated project or on behalf of a job) or when an execute API call fails. There are three distinct types of Execution events:
  • executionConsoleEvent. An executionConsoleEvent pushes R console output generated by an R session to the event stream. Console events are automatically pushed to the event stream each time you execute R code.

  • executionRevoEvent. An executionRevoEvent pushes DeployR-encoded R object data generated by an R session to the event stream. These events require an explicit call to a DeployR-specific R function called revoEvent(). To retrieve the DeployR-encoded value of any R object in your workspace as a executionRevoEvent, add a call to the revoEvent() function in your R code or script and pass the object name as a parameter to this function, for example:

    x <- rnorm(10)
    revoEvent(x)
    
  • executionErrorEvent. An executionErrorEvent pushes an error message when an execute API call fails for any reason. These events occur when the caller attempts to execute R code that is syntactically invalid or when parameters such as inputs or preloads passed on the call are invalid.


### Job Lifecycle Events Job lifecycle events occur each time a Job transitions to a new state in its lifecycle, for example when moving from QUEUED to RUNNING state or when moving from RUNNING to COMPLETED or FAILED state.

There is just one type of lifecycle event, jobLifecycleEvent. Job lifecycle events make it simple for client applications to track the progress of background jobs without having to continuously poll the server for status updates.


### Management Events Management events occur when important runtime conditions are detected by the server. These are the distinct types of Management events:
  • gridActivityEvent. A gridActivityEvent is pushed to the event stream when a slot on the grid is either activated or deactivated on behalf of an anonymous project, authenticated project or on behalf of a job. This type of event provides real-time visibility onto grid activity at runtime.

  • gridWarningEvent. A gridWarningEvent is pushed to the event stream when activity on the grid runs up against limits defined by the Concurrent Operation Policies under Server Policies in the DeployR Administration Console or whenever demands on the grid exceed available grid resources. This type of event signals resource contention and possibly even resource exhaustion and should therefore be of particular interest to the administrator for a DeployR instance.

  • gridHeartbeatEvent. A gridHeartbeatEvent is pushed periodically by the server. This type of event provides a detailed summary of slot activity across all nodes on the grid. This event when used in conjunction with gridActivityEvent and gridWarningEvent provides 100% visibility across all live grid activity at runtime.

  • securityLoginEvent. The securityLoginEvent is pushed to the event stream when users logs in to the server.

  • securityLogoutEvent. The securityLogoutEvent is pushed to the event stream when users log out of the server.

Authenticated, Anonymous, and Management Event Streams

The nature of an event stream connection determines the nature of the events that can be delivered on that stream. Both the current authenticated status of the caller on the /r/event/stream API call and the parameters passed on that call determine the ultimate nature of the event stream connection. The following event stream connection types exist:

  • Authenticated event stream
  • Anonymous event stream
  • Management event stream

All streams push a streamHandshakeEvent event when a connection if first established. The Authenticated event stream pushes execution and job lifecycle events related to a specific authenticated user. The Anonymous event stream pushes only execution events within the anonymous HTTP Session. The Management event stream pushes server-wide management events.


## DeployR Web Services API Overview

While it is not necessary to understand the internal architecture of the DeployR server in order to use this API the following overview is provided in order to lend context to server administrators intending to support the API and to client application developers intending to use the API.

R User Web, Desktop, Mobile Apps

The R user web, desktop, and mobile apps represent client applications built using the DeployR API. The API is a fully standardized Web services interface using JSON over HTTP. This means that any piece of software that is both capable of connecting to the server and parsing either JSON can become a client.

To make life easier for the client developers using this API, DeployR also provides several client libraries. These client libraries, currently available for Java, JavaScript and .NET developers, simplify making calls, encoding data and handling response markup on the API.

DeployR Public API

This document describes the DeployR Public API and the complete set of API services provided for users, projects, jobs, repository-managed files, and scripts and the event stream. Using this API directly or by taking advantage of the DeployR client libraries developers can integrate R-based analytics into their client applications.

DeployR Administration Console

The DeployR Administration Console is a browser-based administrators tool used to customize the deployment configuration for the server, tune server runtime behaviors, and facilitate the integration of client applications on the API. For more information, refer to the Administration Console documentation.

DeployR Grid Management Framework

The grid management framework provides load balancing capabilities for intensive R-compute environments. This framework manages a scalable network of collaborating nodes where each node on the grid contributes its own resources (processor, memory, and disk). Each node can be used by the server to execute R analyses on behalf of client applications. For more information, see section Managing the Grid.

Spring 3 Framework & J2EE Container

The Spring Framework is a lightweight, modular framework for building enterprise Java applications. The DeployR server builds on top of the Spring Framework in order to deliver advanced R-based analytics services on-demand in a secure, robust manner.

Database

The DeployR server uses a database to manage the reliable persistence of all user, project, and repository data.


### Architecture Overview

The DeployR API exposes a set of services via a Web service interface using HTTP(S). You can send HTTP GET and POST requests to invoke services on the API. In a nutshell:

  • HTTP GET method is expected when retrieving data on the API

  • HTTP POST method is required when submitting, updating, or deleting data on the API

Most API calls respond with well-formed markup. On such calls, you can request the response markup in JSON format. Some API calls return binary data such as image plots, structured data like CSV files, or gzip compressed archives. In all cases, the type of data being returned will always be indicated by the Content-Type header on the response.

Additionally, each request returns a meaningful HTTP status code on the response. For a list of response codes that can be indicated on this API and that should be handled by all applications, refer to section API Response Codes.


### API Parameter Overview

For each service call, there is a documented list of required and optional parameters. Parameter values are expected to be UTF-8 compliant and URL-encoded.

The DeployR API currently supports the JSON format for data exchange. Consequently, each method requires the use of the format parameter in order to specify how the request and response data on the call are encoded.

In the next release, the XML format will no longer be supported.

While all parameters are specified as a name/value pair, some parameter values require complex data. Whenever complex data is required, a JSON schema defines how these values are to be encoded. For more information on the schema definitions and specific examples of relevant service calls, see section Web Service API Data Encodings.


### API Response Code Overview

Each DeployR API call responds with a meaningful HTTP status code. Applications using this API should be implemented to handle each of these status codes as appropriate.

HTTP Success Codes

  • 200 OK

HTTP Error Codes
  • 400 Bad Request: invalid data on call

  • 401 Unauthorized Access: caller has insufficient privileges

  • 403 Forbidden Access: caller is unauthorized

  • 405 HTTP Method Disallowed: disallowed HTTP method on call

  • 409 Conflict: project is currently busy on call, concurrent call rejected

  • 503 Service Temporarily Unavailable: HTTP session temporarily invalidated


API Call Status & Error Codes

When an API call returns an HTTP 200 status code the application should still test the success property in the response markup on each call to determine whether the DeployR server was able to successfully execute the service on behalf of the caller.

If the success property in the response markup indicates failure (with a value of false), then the application can inspect the error and errorCode properties in the response markup to determine the underlying cause of that failure.

The error property provides a plain text message describing the underlying failure. The errorCode property indicates the nature of the underlying error. Possible values for the errorCode property are shown here:

  • 900 General Server Error: runtime error.

  • 910 Grid Resource Error: maximum number of concurrent live projects exceeded for authenticated user

  • 911 Grid Resource Error: maximum number of concurrent live projects exceeded for anonymous user

  • 912 Grid Resource Error: maximum number of concurrent live jobs exceeded for authenticated user

  • 913 Grid Resource Error: grid resources temporarily at exhaustion.

  • 914 Grid Resource Error: grid runtime boundary limit exceeded.

  • 915 Grid Resource Error: grid node unresponsive.

  • 916 Grid Resource Error: grid node R session unresponsive.

  • 917 Grid Resource Error: grid node version incompatible with server.

  • 940 Authentication Error: username/password credentials provided are invalid.

  • 941 Authentication Error: user has insufficient privileges to log in on the API.

  • 942 Authentication Error: another user has already authenticated on the current HTTP session.

  • 943 Authentication Error: user account has been disabled by the system administrator

  • 944 Authentication Error: user account has been temporarily locked by the system administrator

  • 945 Authentication Error: user account password has expired, requires reset

To understand how grid resource errors occur, refer to the sections Managing the Grid and Managing Server Policies.

Sample API (JSON) response markup indicating error on call:

{
    "deployr": {
        "response": {
            "call": "/r/user/login",
            "success": false,
            "error": "Bad credentials",
            "errorCode": 900,
            "httpcookie": "2606B516198D7B2F62B43ADBA15F29C2"
        }
    }
}

Explore the API

To help developers familiarize themselves with the full set of APIs DeployR ships with a Web-based API Explorer tool. This tool allows developers to explore the DeployR API in an interactive manner.

For more information, see the documentation on the API Explorer Tool.

API Call Overview

Each of the following "Working with" sections in the API Reference detail the individual API calls by category. For each call described in this reference guide, you can review the following information:

  • Call REST Endpoint

  • Call Description

  • Call HTTP Method

  • Call Request Format

  • Call Parameter List

  • Example Call Parameters

  • Example Call Response Markup (JSON)

Encode R Object Data for use on the API

DeployR-specific encodings are used to encode R object data for use on the API.

Encoded object data can be returned from the server in the response markup on the following calls:

Encoded object data can also be sent to the server as parameter values on the following calls:

Each encoded object is defined by the following properties:

  • name - the name of the encoded object

  • type - the DeployR-specific encoding type for the encoded object

  • rclass - the underlying R class for the encoded object

  • value - the data encapsulated by the encoded object

When encoded objects are returned from the server in the response markup the purpose of the type property is to indicate to developers how to decode the encapsulated data for use within a client application. For example, an encoded vector object returned on the /r/project/workspace/get call appears in the response markup as follows:

{
	"name": "sample_vector",
	"type": "vector",
	"rclass": "character",
	"value": [
		"a",
		"b",
		"c"
	]
}

In the preceding example, the encoded vector, named sample_vector , indicates a type property with a value of vector. This informs the client application that the value property on this encoded object contains an array of values. Those values match the values found in the vector object in the workspace.

When encoded objects are sent as parameters to the server, the client must name each object, and then provide values for the type and value properties for each of those objects. The rclass property is determined automatically by the server and need not be specified by the client application. For example, passing both an encoded logical and an encoded matrix as inputs on the /r/project/workspace/push call is achieved as follows:

{
	"sample_logical" : {
		"type": "primitive",
		"value" : true
	},
	"sample_matrix": {
		"type": "matrix",
		"value": [
			[
				1,
				3,
				12
			],
			[
				2,
				11,
				13
			]
		]
	}
}

In the case of primitive encodings, the type property is optional and can therefore be omitted from the markup being sent to the server.


DeployR Encoding Types

The complete list of supported DeployR-encoding types is shown here. As indicated, these encodings can encode a wide range of different classes of R object:

  • primitive type encodes R objects with class: character, integer, numeric and logical.

  • date type encodes R objects with class: Date, POSIXct, and POSIXlt.

  • vector type encodes R objects with class: vector.

  • matrix type encodes R objects with class: matrix.

  • factor type encodes R objects with class: ordered and factor.

  • list type encodes R objects with class: list.

  • dataframe type encodes R objects with class: data.frame.


Sample DeployR Encodings

The following R code is provided for demonstration purposes. The code creates a wide range of R objects. After the following code, you will find sample JSON response markup that demonstrates how each of the objects generated here is encoded on the API.

# primitive character
x_character <- "c"
# primitive integer
x_integer <- as.integer(10)

# primitive numeric (double)
x_numeric <- 5

# logical
x_logical <- T

# Date
x_date <- Sys.Date()

# "POSIXct" "POSIXt"
x_posixct <- Sys.time()

# vector (character)
x_character_vector <- c("a","b","c")

# vector (integer)
x_integer_vector <- as.integer(c(10:25))

# vector (numeric)
x_numeric_vector <- as.double(c(10:25))

# matrix
x_matrix <-  matrix(c(1,2,3, 11,12,13), nrow = 2, ncol=3)

# ordered factor
x_ordered_factor <- ordered(substring("statistics", 1:10, 1:10), levels=c("s","t","a","c","i"))

# unordered factor
x_unordered_factor <- factor(substring("statistics", 1:10, 1:10), levels=c("s","t","a","c","i"))

#list
x_list <- list(c(10:15),c(40:45))

# data frame
x_dataframe <- data.frame(c(10:15),c(40:45))

JSON Encodings on API Response Markup

The following response markup contains annotations that describe each of the supported DeployR R object encodings:

{
	"deployr": {
		"response": {
			"success": true,
			"workspace": {
		  "objects" : [
		//
		//  DeployR Primitive Encodings 
		//

				{
					"name": "x_character",
					"type": "primitive",
					"rclass": "character",
					"value": "c"
				},

				{
					"name": "x_integer",
					"type": "primitive",
					"rclass": "integer",
					"value": 10
				},

				{
					"name": "x_numeric",
					"type": "primitive",
					"rclass": "numeric",
					"value": 5
				},

				{
					"name": "x_logical",
					"type": "primitive",
					"rclass": "logical",
					"value": true
				},


		//
		//  DeployR Date Encodings 
		//

				{
					"name": "x_date",
					"type": "date",
					"rclass": "date",
					"value": "2011-10-04",
					"format": "yyyy-MM-dd"
				},

				{
					"name": "x_posixct",
					"type": "date",
					"rclass": "POSIXct",
					"value": "2011-10-05",
					"format": "yyyy-MM-dd HH:mm:ss Z"
				},


		//
		//  DeployR Vector Encodings 
		//

				{
					"name": "x_character_vector",
					"type": "vector",
					"rclass": "character",
					"value": [
						"a",
						"b",
						"c"
					]
				},

				{
					"name": "x_integer_vector",
					"type": "vector",
					"rclass": "integer",
					"value": [
						10,
						11,
						12,
						13,
						14,
						15
					]
				},

				{
					"name": "x_numeric_vector",
					"type": "vector",
					"rclass": "numeric",
					"value": [
						10,
						11,
						12,
						13,
						14,
						15
					]
				},


		//
		//  DeployR Matrix Encodings 
		//

				{
					"name": "x_matrix",
					"type": "matrix",
					"rclass": "matrix",
					"value": [
						[
							1,
							3,
							12
						],
						[
							2,
							11,
							13
						]
					]
				},


		//
		//  DeployR Factor Encodings 
		//

				{
					"name": "x_ordered_factor",
					"type": "factor",
					"rclass": "ordered",
					"value": [
						"s",
						"t",
						"a",
						"t",
						"i",
						"s",
						"t",
						"i",
						"c",
						"s"
					],
					"labels": [
						"s",
						"t",
						"a",
						"c",
						"i"
					],
					"ordered": true
				},

				{
					"name": "x_unordered_factor",
					"type": "factor",
					"rclass": "factor",
					"value": [
						"s",
						"t",
						"a",
						"t",
						"i",
						"s",
						"t",
						"i",
						"c",
						"s"
					],
					"labels": [
						"s",
						"t",
						"a",
						"c",
						"i"
					],
					"ordered": false
				},


		//
		//  DeployR List Encodings 
		//

				{
					"name": "x_list",
					"type": "list",
					"rclass": "list",
					"value": [
						{   "name" : "1",
							"value": [
								10,
								11,
								12,
								13,
								14,
								15
							],
							"type": "vector",
							"rclass": "numeric"
						},
						{   "name" : "2",
							"value": [
								40,
								41,
								42,
								43,
								44,
								45
							],
							"type": "vector",
							"rclass": "numeric"
						}
					]
				},


		//
		//  DeployR Dataframe Encodings 
		//

				{
					"name": "x_dataframe",
					"type": "dataframe",
					"rclass": "data.frame",
					"value": [
						{
							"type": "vector",
							"name": "c.10.15.",
							"rclass": "numeric",
							"value": [
								10,
								11,
								12,
								13,
								14,
								15
							]
						},
						{
							"type": "vector",
							"name": "c.40.45.",
							"rclass": "numeric",
							"value": [
								40,
								41,
								42,
								43,
								44,
								45
							]
						}
					]
				}

			  ]
		},
			"call": "/r/project/workspace/get",
			"project": {
				"project": "PROJECT-00c512b6-84a4-45ff-b94a-134cca8e0584",
				"descr": null,
				"owner": "testuser",
				"live": true,
				"origin": "Project original.",
				"name": null,
				"projectcookie": null,
				"longdescr": null,
				"ispublic": false,
				"lastmodified": "Wed, 5 Oct 2011 06:31:37 +0000"
			}
		}
	}
}

Sample JSON Encodings for Input Parameters on API

{
	"x_character": {
		"type": "primitive",
		"value": "c"
	},
	"x_integer": {
		"type": "primitive",
		"value": 10
	},
	"x_double": {
		"type": "primitive",
		"value": 5.5
	},
	"x_logical": {
		"type": "primitive",
		"value": true
	},
	"x_date": {
		"type": "date",
		"value": "2011-10-04",
		"format": "yyyy-MM-dd"
	},
	"x_posixct": {
		"type": "date",
		"value": "2011-10-05 12:13:14 -0800",
		"format": "yyyy-MM-dd HH:mm:ss Z"
	},
	"x_character_vector": {
		"type": "vector",
		"value": [
			"a",
			"b",
			"c"
		]
	},
	"x_integer_vector": {
		"type": "vector",
		"value": [
			10,
			11,
			12
		]
	},
	"x_double_vector": {
		"type": "vector",
		"value": [
			10,
			11.1,
			12.2
		]
	},
	"x_matrix": {
		"type": "matrix",
		"value": [
			[
				1,
				3,
				12
			],
			[
				2,
				11,
				13
			]
		]
	},
	"x_ordered_factor": {
		"type": "factor",
		"ordered": true,
		"value": [
			"s",
			"t",
			"a",
			"t",
			"i",
			"s",
			"t",
			"i",
			"c",
			"s"
		],
		"labels": [
			"s",
			"t",
			"a",
			"c",
			"i"
		]
	},
	"x_unordered_factor": {
		"type": "factor",
		"ordered": false,
		"value": [
			"s",
			"t",
			"a",
			"t",
			"i",
			"s",
			"t",
			"i",
			"c",
			"s"
		],
		"labels": [
			"s",
			"t",
			"a",
			"c",
			"i"
		]
	},
	"x_list": {
		"type": "list",
		"value": [
			{
				"name": "first",
				"value": [
					10,
					11,
					12
				],
				"type": "vector"
			},
			{
				"name": "second",
				"value": [
					40,
					41,
					42
				],
				"type": "vector"
			}
		]
	},
	"x_dataframe": {
		"type": "dataframe",
		"value": [
			{
				"name": "first",
				"value": [
					10,
					11,
					12
				],
				"type": "vector"
			},
			{
				"name": "second",
				"value": [
					40,
					41,
					42
				],
				"type": "vector"
			}
		]
	}
}

DeployR Encodings Made Easy

To simplify life for those client developers using The DeployR API, we provide several client libraries for Java, JavaScript, and .NET developers. A major benefit of using these client libraries is that they greatly simplify the creation of DeployR-encoded object data inputs to the server and the parsing of DeployR-encoded object data as outputs from the server.


## API Change History
### DeployR for Microsoft R Server 8.0.5
  1. All authenticated APIS now require a Cross-Site Request Forgery (CSRF) token in the HTTP request header named X-XSRF-TOKEN for POST requests.

    The X-XSRF-TOKEN can be obtained from a successful authentication /r/user/login API call. The X-XSRF-TOKEN value can be retrieved from either the HTTP response header or the X-XSRF-TOKEN property in the response markup. The X-XSRF-TOKEN name|value must then be included in the HTTP request header for any future authenticated API calls, otherwise an HTTP 403 error is given.

    Example:

    {
    "deployr": {
     "response": {
       "call": "/r/user/login",
       "success": true,
       "user": {
         "username": "george",
         "displayname": "George Best",
         "permissions": {
             "scriptManager": true,
             "powerUser": true,
             "packageManager": true,
             "administrator": false,
             "basicUser": true
         },
         "cookie": null
       },
       "limits": {
         "maxIdleLiveProjectTimeout": 1800,
         "maxConcurrentLiveProjectCount": 170,
         "maxFileUploadSize": 268435456
       },
       "uid": "UID-5CB911405DC6EB0F8E283990F7969E63",
       "X-XSRF-TOKEN": "53708abe-c5c2-4091-9ced-6314d49de0a3"
     }
    }
    }
    
  2. Affecting all APIS, the httpcookie property has been removed from the response markup in place of the unique request identifier uid (which is not a cookie).

    {
       "deployr": {
           "response": {
               "call": "/r/user/release",
               "success": true,
               "whoami": "george",
               “uid”: “UID-07D91F6C510E10ED84E6AC52CD0B6D1B"
           }
       }
    }
    
  3. Removal of the Repository Shell Script APIs. /r/repository/shell/execute has been removed

Note: Going forward application/xml will become depreciated, supporting only application/json in the response content-type


### DeployR 8.0.0
  • Rebranding and Volume Licensing

  • This release of DeployR primarily focused on:

    • rebranding of the product from Revolution Analytics to Microsoft.

    • defects


### DeployR 7.4.1

Standard Execution Model Changes

The DeployR API standard execution model has been updated. A new on-execution parameter has been added:

  • artifactsoff - (optional) when enabled, artifacts generated in the working directory are neither cached to the database or reported in the response markup

This change applies across all execution APIs:


### DeployR 7.4

Grid Cluster Targeted Executions

Grid node "cluster" names, used to denote the runtime characteristics of a node, can be assigned by the DeployR admin, using the Administration Console, to individual nodes or groups of nodes on the DeployR grid, for example, "hi-mem" or "hi-cpu". This feature is DeployR Enterprise only.

By identifying a value on a new cluster parameter client applications can request tasks be executed on nodes within a specific cluster on the grid on the following calls:

Cluster names are case-inssensitive so "Hi-Mem" and "hi-mem" indicate the same cluster. If the cluster indicated on any of these calls fails to match an actual cluster name assigned by the admin on the DeployR grid the call is rejected.

Project Phantom Executions

Phantom executions are a special type of execution on a project that avoids per-execution meta-data being created by the server in the database. This helps avoid database resource exhaustion on the server when high-throughput, long-lived project pools are in use, such as pools used by the RBroker Framework. Client applications can request tasks be executed as phantom executions by enabling the value of the new phantom parameter on the following calls:

A phantom execution will not appear in the project history returned by the /r/project/execute/history call.

External Repository Support

DeployR has always provided API support for working with repository-managed files and scripts that have been uploaded into the DeployR Default Repository. DeployR 7.4 server introduces support for a new repository store known as the DeployR External Repository.

Unlike the default repository which is backed by the MongoDB database, the new external repository is simply a directory on disk from where the DeployR server can list and load files and execute scripts.

Managing files in the external repository is as simple as creating, editing, copying, moving, and deleting files on disk. All interactions with the external repository on the API are read-only, so files and scripts can be listed, loaded and even executed but not modified in any way on the API.

The DeployR External Repository can be used to provide a seamless bridge between existing file control systems, such as git and svn, and the DeployR server.

Listing external repository-managed files and scripts is supported using a new external parameter on the following set of APIs:

Loading and executing external repository-managed files and scripts use the existing set of filename, directory, and author parameters on the following calls:

The base directory for the external repository can be found here on DeployR server:

$DEPLOYR_HOME/deployr/external/repository

Each user can maintain private, shared, and public files in the external repository at the following locations respectively:

// Private external repository files.
$DEPLOYR_HOME/deployr/external/repository/{username}
// Shared external repository files.
$DEPLOYR_HOME/deployr/external/repository/shared/{username}

// Public external repository files.
$DEPLOYR_HOME/deployr/external/repository/public/{username}

// Note, sub-directories are supported by the external repository,
// for example, the following files are all privately owned by testuser:
$DEPLOYR_HOME/deployr/external/repository/testuser/plot.R
$DEPLOYR_HOME/deployr/external/repository/testuser/demo/score.R
$DEPLOYR_HOME/deployr/external/repository/testuser/demo/model.rData

A small number of sample files are deployed to the external repository following each new DeployR 7.4 installation and you may try out the external repository using new support found in the latest API Explorer.

For more information about working with the new DeployR External Repository, post your questions to the forum.

Default and External Repository Filters

Each of the following APIs has been updated to support more sophisticated filters on results:

Using the directory and categoryFilter independently or together the results returned on these calls can be filtered to specific subsets of available files, for example:

(1) Show me all of the data files across all of my directories in the default repository

categoryFilter=data

(2) Show me all of the files in the tutorial directory in the default repository

directory=tutorial

(3) Show me binary R files within the example-fraud-score directory in the default repository

directory=example-fraud-score
categoryFilter=R

(4) Show me just R scripts within my private demo directory in the external repository

directory=external:root:demo
categoryFilter=script

(5) Show me just R scripts within testuser's public plot directory in the external repository

directory=external:public:testuser:plot
categoryFilter=script

Repository Shell Script Execution

The following new Repository Shell API has been added:

  1. /r/repository/shell/execute | Executes a repository-managed shell script on the DeployR server.

For more information about working with this new API, please post your questions to the forum.


### DeployR 7.3

User Grid Resource Release API

The following new User API has been added:

  1. /r/user/release | Releases server-wide grid resources held by the currently authenticated user

Standard Execution Model Changes

The DeployR API standard execution model has been updated. A new pre-execution parameter has been added:

  • preloadbydirectory - (optional) comma-separated list of repository directory names from which to load the complete set of files in each directory into the working directory prior to execution

Also, a new on-execution parameter has been added:

  • enableConsoleEvents - (optional) when enabled R console events are delivered on the event stream for the current execution

The introduction of the enableConsoleEvents is a breaking change for those using the event stream. Prior versions of DeployR enabled R console events by default and there was no way on the API to disable these events. As of 7.3, R console events on the event stream are disabled by default and if necessary, must be explicitly requested by enabling this new parameter.

These two changes apply across all execution APIs:

The new preloadbydirectory parameter is also supported on the following APIs: