RGG: a general GUI Framework for R scripts.

BACKGROUND: R is the leading open source statistics software with a vast number of biostatistical and bioinformatical analysis packages. To exploit the advantages of R, extensive scripting/programming skills are required.
RESULTS: We have developed a software tool called R GUI Generator (RGG) which enables the easy generation of Graphical User Interfaces (GUIs) for the programming language R by adding a few Extensible Markup Language (XML) - tags. RGG consists of an XML-based GUI definition language and a Java-based GUI engine. GUIs are generated in runtime from defined GUI tags that are embedded into the R script. User-GUI input is returned to the R code and replaces the XML-tags. RGG files can be developed using any text editor. The current version of RGG is available as a stand-alone software (RGGRunner) and as a plug-in for JGR.
CONCLUSION: RGG is a general GUI framework for R that has the potential to introduce R statistics (R packages, built-in functions and scripts) to users with limited programming skills and helps to bridge the gap between R developers and GUI-dependent users. RGG aims to abstract the GUI development from individual GUI toolkits by using an XML-based GUI definition language. Thus RGG can be easily integrated in any software. The RGG project further includes the development of a web-based repository for RGG-GUIs. RGG is an open source project licensed under the Lesser General Public License (LGPL) and can be downloaded freely at (http://rgg.r-forge.r-project.org).

Background

R is an environment in which statistical techniques can be implemented [1]. R is an open source project, licensed under the General Public License, and is used by a growing number of researchers in the field of bioinformatics/biostatistics. Many new statistical procedures are published with corresponding R scripts or packages. Over 1,600 packages can be obtained from the Comprehensive R Archive Network [2] and further 300 bioinformatics packages can be obtained from the website of the Bioconductor project [3]. Notably, microarray analysis has been greatly improved by several packages like limma [4], affy [5], siggenes/samr [6], pamr [7], limmaGUI [8] and by software applications like BRB-ArrayTools [9] or by web applications like CARMAweb [10], MIDAW [11], and RACE [12] (all of these are using R and Bioconductor). R is based on an effective programming language that includes conditionals, loops, user-defined recursive functions as well as input and output facilities. The user frequently needs to change R code like the one for setting the working directory, reading an input file and changing parameters to new values, which usually requires profound knowledge of the R scripting language. For many applications, large parts of the code remain unchanged.

There are different approaches for the GUI-based usage of R: an option is the use of GUI toolkits such as Tcl/Tk [13], Gtk2 [14], Java Swing [15] and wxWidgets [16] employing the R packages tcltk [17], RGtk2 [18], rJava [19] and RwxWidgets [20]. Every toolkit has special characteristics that require the programming of its own custom GUI. gWidgets [21] serve as an abstraction of the named R packages and offers a standard gateway. A GUI that is written using gWidgets is easily compatible with these toolkits and can thus be executed. All these packages including gWidgets use R to describe GUIs and require R to generate the GUI. Such GUIs depend on R and are less flexibly integrated in other softwares.

Stand-alone frontends present another approach for the GUI-based usage of R. Rcmdr [22] and Rkward [23] runs on Linux and has a similar usability to Rcmdr yet with a richer interface. The user interface of Rkward was developed using the Qt Toolkit [24] in C++ [25]. Like in RGG, GUIs are described with XML making it easier for the end-user to expand Rkward with further GUI functions. Rpad [26] is a web-based analysis program which integrates R to an Apache web server on the backend. The R code embedded in the HTML is sent back to the server for statistical evaluation. Extending these tools to new functions requires programming skills in HTML, JavaScript [27] and/or PHP [28].

In this context, this project's aim is to develop a general GUI Framework for R. This should be independent from the GUI toolkits and integrating software and make R based statistical computing available to a wider audience less familiar with R programming.

Implementation

The R GUI Generator (RGG) framework allows the R user to generate Graphical User Interfaces (GUIs) for R scripts. RGG uses a component-based approach and consists of two parts: an XML [29] (Extensible Markup Language)-based GUI definition language to use the existing components (GUI elements) and a GUI engine that is implemented in Java.

RGG – GUI Engine

Within the RGG framework, the GUI engine is the interpreter of an RGG file which generates the GUI in runtime. GUI engines are language-specific and contain the implementation of GUI elements. The current engine was implemented in Java using the Java Swing Library.

However, new engines using other programming languages including R based on the same RGG GUI definition language specification may further be implemented. Thus the RGG GUI definition language is independent from the GUI toolkit.

Java based RGG GUI engine

Figure 1 shows the software architecture of the RGG GUI engine. It consists of four parts: the GUI elements, the RGGModel, the RGGPanelModel, and the RGGPanelBuilder. The GUI engine creates the GUI elements for each XML tag in the rgg file. These GUI elements are added to the RGGModel from which the new R script is generated. The RGGPanelModel is created from the RGGModel and contains the GUI model. It has a two-dimensional structure for laying out the GUI widgets in rows and columns. The RGGPanelBuilder generates the GUI using the RGGPanelModel. The GUI element consists of three subcomponents, an RElement, a VisualComponent, and an RElementFactory. The RElementFactory receives XML tags including their attributes from the RGG file and initializes the RElements and their visual components. The VisualComponent draws the GUI elements on the screen and the RElement converts the user inputs into the corresponding R code.

Figure 1

Software Architecture of Java based RGG GUI Engine. The GUI has four main parts: GUI elements, RGGModel, RGGPanelModel, and RGGPanelBuilder. The engine creates GUI elements for each XML tag. These GUI elements are added to the RGGModel from which the new R script is then generated. The RGGPanelModel is created from the RGGModel. RGGPanelBuilder generates the GUI using RGGPanelModel. The GUI element has three subparts: RElement, VisualComponent, and RElementFactory. RElementFactory initializes the RElement and its visual parts, the VisualComponent. VisualComponent draws the GUI elements on the screen and RElement converts the user inputs into the corresponding R code.

RGG – GUI definition Language

The RGG GUI definition language is based on XML and contains all basic GUI elements. RGG files are thus XML files. Rules applying to an XML file also apply to RGG files. To distinguish RGG files from other XML files, the root tag must be set to . In between the two tags and the user can combine GUI elements and R code. There are common GUI elements such as , , or more complex and specific elements like a dataframe/matrix editor or an importer of microarray data, the . Each GUI element consists of a given number of GUI widgets. Frequently, the basic elements like (a label and a text field), (a label, a text field, and a button), consist of two or three GUI widgets. However, there are no limitations to the complexity of GUI elements. A file containing both R code and GUI elements is saved with the suffix .rgg. Such RGG files serve as templates for the GUI engine. The GUI engine loads the RGG file and dynamically creates and arranges GUI elements from the XML tags. Each GUI element converts the user settings (e.g. text in a ) into the corresponding R code; i.e., the XML tags in the RGG file are replaced by the corresponding R codes. In this way a new R script is generated which is then executed in R. Several XML tags leave the R script unchanged, but improve the GUI layout. These tags include the container elements and . The GUI components and elements can be arranged both horizontally and vertically, i.e. spread across several columns and rows of the GUI. By default, GUI components and elements are arranged vertically. To ensure horizontal arrangement of child elements, the container element can be used, whereas the vertical arrangement can be restored using . The tag introduces a horizontal line to separate groups of independent GUI elements, and the tag modifies the spatial separation of single GUI elements. Each GUI element has its attributes, which aids its fine-tuning. Some attributes are common in many GUI elements such as "var" and "label". The var attribute's value corresponds to the R variable name that the returned user input is assigned to. The label attribute defines the label of the element in the GUI. Each of the more than fifteen GUI elements and their definition by XML tags are described in detail on the project webpage and an HTML page (provided as Additional file 1) gives an overview of the GUI elements and their attributes. New GUI elements are continuously uploaded and documented on the website. The development of RGG was inspired by XUL, several components such as the and concept were directly adapted from the XUL. The reason why we could not use XUL directly is related to the choice of the programming language JAVA which should make it easier to integrate RGG in other JAVA based softwares (e.g. Taverna) [30].

JGR Plug-in & RGGRunner

The current version of RGG is available as a stand-alone software (RGGRunner) and as a plug-in for JGR (Figure 2), which is a Java frontend for R [31]. RGG generates an R script that is executed locally by RGGRunner or JGR-Plug-in. However, new execution environments can be developed, which provide a server-client architecture e.g. via Rserve. RGG does not have a server-client architecture yet, and runs R scripts locally (by JGR-Plug-in and RGGRunner). RGG files can be developed in any text editor; however, using the RGG plug-in for JGR, the developer can control both the R prompt and the nascent GUI. The RGG framework can be integrated into any software supporting R and Java. Users with limited R programming skills can operate RGG using the stand-alone application RGGRunner (provided as Additional File 2 and 3), where any given GUI can be imported and executed. Further, the source code of an RGG file and the generated R code can be viewed directly within RGGRunner. The RGGRunner is available for all major operating systems, such as Linux, Windows and Mac OS X.

Figure 2

Concept of R GUI Generator (RGG). For the development of R Graphical User Interfaces (GUIs) we recommend the RGG plug-in for JGR, however, any text editor can be used. Combining an R script with predefined GUI definition tags (RGG elements) results in an rgg file. The stand-alone software RGGRunner interprets the rgg file and generates the GUI for the end-user with no R skills.

Outlook

The RGG project is going to provide a well-organized and documented web-repository for RGG files. In addition, the RGG framework will be integrated into the Taverna environment which enables the composition of bioinformatic workflows [32]. RGG is an open source project under the Lesser General Public License. The source code of these elements and of the GUI engine is freely available via a public subversion repository (see project web-site). Downloads and installation instructions are available at the project homepage .

Simple example

It may be desirable that a successful R script evolves into a GUI based function. Figure 3A gives an example of a self-explaining R-script for a Fisher's exact test on a 2 × 2 contingency table, a test frequently used in biomedical research. To run the Fisher test in R with different count data, the numeric values have to be changed in this R script. The user has to change the R script accordingly, which requires knowledge of R scripting. An RGG-GUI was created for the Fisher's exact test using a text editor or, alternatively, the RGG plug-in for JGR. This GUI includes the RGG GUI elements matrix editor , dropdown menu and a header text