# Generating KNIME Nodes¶

With the help of GenericWorkflowNodes one can make his/her command line applications available in KNIME as nodes that can be included as parts of a bigger workflow. In this tutorial you will learn how to make your command line application available as a KNIME node.

## Prerequisites¶

** Eclipse KNIME SDK**
git
Apache Ant
The Generic KNIME Plugins project uses Apache Ant as the build system. On Linux and Mac, you should be able to install it through your package manager. For Windows, see the Apache Ant Downloads (note that samtools does not work on Windows so you will not be able to follow through with this tutorial on Windows).

## Overview¶

KNIME nodes are shipped as Eclipse plugins. The GenericKnimeNodes (GWN) package provides the infrastructure to automatically generate such nodes from the description of their command line. The description of the command line is kept in XML files called Common Tool Descriptor (CTD) files.

Tip

For SeqAn Apps, thanks to the seqan::ArgumentParser class generating a plugin directory is completely automatic if you are using the SeqAn build infrastructure. All you have to do is run make with target prepare_workflow_plugin. Read the tutorial Generating SeqAn KNIME Nodes for more details.

The input of the GWN package is a directory tree with the following structure:

plugin_dir
│
├── plugin.properties
│
├── descriptors (place your ctd files and mime.types here)
├    ├── app1_name.ctd
├    ├── app2_name.ctd
├    ├── ...
├    ├── mime.types
│
│
├── icons (the icons to be used must be here)
│
├── DESCRIPTION (A short description of the project)
│
├── LICENSE (Licensing information of the project)
│

### The file plugin.properties¶

This file contains the plugin configuration. Look at the following plugin.properties file as an example:

# the package of the plugin
pluginPackage=net.sf.samtools

# the name of the plugin
pluginName=SamTools

# the version of the plugin
pluginVersion=0.1.17

# the path (starting from KNIMEs Community Nodes node)
nodeRepositoyRoot=community

executor=com.genericworkflownodes.knime.execution.impl.LocalToolExecutor
commandGenerator=com.genericworkflownodes.knime.execution.impl.CLICommandGenerator

When creating your own plugin directory, you only have to update the first three properties:

pluginPackage
A Java package path to use for the Eclipse package.
pluginName
A CamelCase name of the plugin.
pluginVersion
Version of the Eclipse plugin.

### The Directory descriptors¶

The descriptors directory contains two types of files i.e. the CTD files for each application in our plugin and a mime.types file.

### MIME Types¶

mime.types file is a text file that contains a mapping between MIME types and file extensions. Every file extension, to be either used or produced by the applications in the plugin, has to be registered here. Each line contains the definition of a MIME type. The name of the mime type is followed (separated by a space) by the file extensions associated with the file type. The following example shows how the content of a mime.types file looks like.

application/x-fasta fa fasta
application/x-fastq fq fastq
application/x-sam sam
application/x-bam bam

Important

There may be no ambiguous mappings, like giving a single extension for both application/x-fasta and application/x-fastq in the example shown above. An extension should be mapped to a single application.

For every application with the name app_name, there is one CTD file called ${app_name}.ctd. These files contain the command line description of an application in XML format. They also decide which command-line arguments will be input/output ports or configuration entries in the node to be generated. Tip For applications developed in SeqAn or applications using the seqan::ArgumentParser for parsing their command-line arguments a CTD file can be generated using a hidden parameter -write-ctd. for example: ./seqan_app_name -write-ctd seqan_app_name.ctd Look at the tutorials Make Your SeqAn App KNIME Ready and Generating SeqAn KNIME Nodes . Below is an example of a CTD file for a SortBam tool for sorting BAM files. <?xml version="1.0" encoding="UTF-8"?> <tool name="SortBam" version="0.1.17" category="SAM and BAM" docurl="http://samtools.sourceforge.net/samtools.shtml"> <executableName>samtools</executableName> <description><![CDATA[SAMtools BAM Sorting.]]></description> <manual><![CDATA[samtools sort]]></manual> <docurl>Direct links in docs</docurl> <cli> <clielement optionIdentifier="sort" isList="false" /> <clielement optionIdentifier="-f" isList="false" /> <!-- Following clielements are arguments. You should consider providing a help text to ease understanding. --> <clielement optionIdentifier="" isList="false"> <mapping referenceName="bam_to_sam.argument-0" /> </clielement> <clielement optionIdentifier="" isList="false"> <mapping referenceName="bam_to_sam.argument-1" /> </clielement> <clielement optionIdentifier="" isList="false"> <mapping referenceName="bam_to_sam.argument-2" /> </clielement> </cli> <PARAMETERS version="1.4" xsi:noNamespaceSchemaLocation="http://open-ms.sourceforge.net/schemas/Param_1_4.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"> <NODE name="bam_to_sam" description="SAMtools BAM to SAM conversion"> <ITEM name="argument-0" value="" type="input-file" required="true" description="Input BAM file." supported_formats="*.bam" /> <ITEM name="argument-1" value="" type="output-file" required="true" description="Output BAM file." supported_formats="*.bam" /> <ITEM name="argument-2" value="" type="string" required="true" description="Sort by query name (-n) instead of position (default)" restrictions=",-n" /> </NODE> </PARAMETERS> </tool> Hint If a <clielement> does provides an empty optionIdentifier then it is a positional argument without a flag (examples for parameters with flags are -n 1, --number 1). If a <clielement> does not provide a <mapping> then it is passed regardless of whether has been configured or not. The samtools_sort_bam tool from above does not provide any configurable options but only two arguments. These are by convention called argument-0 and argument-1 but could have any name. Also, we always call the program with sort -f as the first two command line arguments since we do not provide a mapping for these arguments. Click more… to see the description of the tags and the attributes in the CTD XML file: CTD Tags attributes: /tool The root tag. /tool@name The CamelCase name of the tool as shown in KNIME and part of the class name. /tool@version The version of the tool. /tool@category The path to the tool’s category. /tool/executableName The name of the executable in the payload ZIP’s bin dir. /tool/description Description of the tool. /tool/manual Long description for the tool. /tool/docurl URL to the tool’s documentation. /tool/cli Container for the <clielement> tags. These tags describe the command line options and arguments of the tool. The command line options and arguments can be mapped to parameters which are configurable through the UI. The parameters are stored in tool/PARAMETERS /tool/cli/clielement There is one entry for each command line argument and option. /tool/cli/clielement@optionIdentifier The identifier of the option on the command line. For example, for the -l option of ls, this is -l. /tool/cli/clielement@isList Whether or not the parameter is a list and multiple values are possible. One of true and false. /tool/cli/clielement/mapping Provides the mapping between a CLI element and a PARAMETER. /tool/cli/clielement/mapping@referenceName The path of the parameter. The parameters <ITEM> s in tool/PARAMETERS are stored in nested <NODE> tags and this gives the path to the specific parameter. /tool/PARAMETERS Container for the <NODE> and <ITEM> tags. The <PARAMETERS> tag is in a diferent namespace and provides its own XSI. /tool/PARAMETERS@version Format version of the <PARAMETERS> section. /tool/PARAMETERS/…/NODE A node in the parameter tree. You can use such nodes to organize the parameters in a hierarchical fashion. /tool/PARAMETERS/…/NODE@advanced Boolean that marks an option as advanced. /tool/PARAMETERS/…/NODE@name Name of the parameter section. /tool/PARAMETERS/…/NODE@description Documentation of the parameter section. /tool/PARAMETERS/…/ITEM Description of one command line option or argument. /tool/PARAMETERS/…/ITEM@name Name of the option. /tool/PARAMETERS/…/ITEM@value Default value of the option. When a default value is given, it is passed to the program, regardless of whether the user touched the default value or not. /tool/PARAMETERS/…/ITEM@type Type of the parameter. Can be one of string, int, double, input-file, output-path, input-prefix, or output-prefix. Booleans are encoded as string with the restrictions attribute set to "true,false". /tool/PARAMETERS/…/ITEM@required Boolean that states whether the parameter is required or not. /tool/PARAMETERS/…/ITEM@description Documentation for the user. /tool/PARAMETERS/…/ITEM@supported_formats A list of supported file formats. Example: "*.bam,*.sam". /tool/PARAMETERS/…/ITEM@restrictions In case of int or double types, the restrictions have the form min:, :max, min:max and give the smallest and/or largest number a value can have. In the case of string types, restrictions gives the list of allowed values, e.g. one,two,three. If the type is string and the restriction field equals "true,false", then the parameter is a boolean and set in case true is selected in the GUI. A good example for this would be the -l flag of the ls program. ### The Directory payload¶ The directory payload contains ZIP files with the executable tool binaries. Usually there is one ZIP file for each platform (Linux, Windows, and Mac Os X) architecture (32Bit or 64Bit)combination. The names of the files are binaries_${plat}_${arch}.zip where${plat} is one of lnx, win, or mac, and ${arch} is one of 32 and 64. In this way the appropriate binaries will be used based on the system that KNIME is running on. Some, even all, of the zip files representing different architectures can be missing. That means the node will not be functional on a KNIME instance installed on the architecture(s) corresponding to missing zip file(s). Nevertheless the payload directory has to be present even if it is empty. Each ZIP file contains a directory /bin which is used as the search path for the binary given by <executableName> and an INI file /binaries.ini which can be used to set environment variables before executing any of tools. ### The Directory icons¶ Here we put icons for different purposes. • An image file with a name category.png (15x15 px): an icon for categories in the KNIME node explorer tree. • An image file with a name splash.png (50x50 px): an icon to be displayed in the KNIME splash screen. • An image file with a name app_name.png (15x15 px): an icon to be displayed with the corresponding node of the application app_name. This is done once for each app but it’s optional. ## Running Example¶ We will adapt some functions from the samtools package to KNIME: BamToSam This tool will execute samtools view -o${OUT} ${IN}. SamToBam This tool will execute samtools view -Sb -o${OUT} ${IN}. SortBam This tool will execute samtools sort -o${OUT} ${IN}. ## Preparation: Building samtools and Downloading GenericKnimeNodes¶ We will work in a new directory knime_samtools (we will assume that the directory is directly in your$HOME for the rest of the tutorial. First we need to download samtools-1.3 from http://www.htslib.org/download/. Alternatively you can also use this direct link or use either of wget or curl utilities from your command line as follows.

wget

curl

Now let us extract and build samtools

knime_samtools # tar -jxvf samtools-1.3.tar.bz2
...
knime_samtools # cd samtools-1.3
samtools-1.3 # ./configure
samtools-1.3 # make
...
samtools # ls -l samtools
-rwxr-xr-x 1 user group 1952339 May  7 16:36 samtools
samtools # cd ..
knime_samtools #

knime_samtools # git clone git://github.com/genericworkflownodes/GenericKnimeNodes.git

## Preparation: Installing KNIME File Handling¶

We need to install support for file handling nodes in KNIME. In order to do that - Launch your Eclipse-KNIME-SDK

Tip

Launching Eclipse-KNIME-SDK

• If you are on a linux system you should browse to your knime_eclipse_3.1 installation double click on eclipse executable.
• If you are using MacOS X then you can launch the KNIME SDK 3.1 as you would launch any other application.
• Open the window for installing Eclipse plugins; in the program’s main menu: Help > Install New Software....
• On the install window enter http://www.knime.org/update/3.1 into the Work with: field, enter file into the search box, and finally select KNIME File Handling Nodes in the list.
• Then, click Next and follow through with the installation of the plugin. When done, Eclipse must be restarted.; in the program’s main menu: Help > Install New Software....

## Obtaining the Demo Workflow Plugin Directory¶

Please download the file workflow_plugin_dir.zip to your knime_samtools directory and look around in the archive. Also have a look into binaries_*_*.zip files in payload. The structure of this ZIP file is similar to the one explained in the Overview section at the beginning of this tutorial.

## Creating an Exclipse Plugin from the Plugin Directory¶

The next step is to use GKN to create an Eclipse plugin from the workflow plugin directory. For this, change to the directory GenericKnimeNodes that we cloned using git earlier. We then execute ant and pass the variables knime.sdk with the path to the KNIME SDK that you downloaded earlier and plugin.dir with the path of our extracted workflow_plugin_dir directory.

Tip

Path to Eclipse-KNIME-SDK - If you are on a linux system the path to knime.sdk is the path to your knime_eclipse_3.1.0 - If you are using MacOS X then the default path to knime.sdk is /Applications/KNIME\ SDK\ 3.1.0.app/

knime_samtools # cd GenericKnimeNodes
GenericKnimeNodes # ant -Dknime.sdk=${HOME}/eclipse_knime_3.1 \ -Dplugin.dir=$HOME/knime_samtools/workflow_plugin_dir

This generates an Eclipse plugin with wrapper classes for our nodes. The generated files are within the generated_plugin directory of the directory GenericKnimeNodes.

## Importing the Generated Projects into Eclipse¶

On the KNIME Eclipse SDK window go to the menu File > Import.... In the Import window, select General > Existing Project Into Workspace

On the pop-up dialog, click Browse... next to Select root directory.

Then, select the directory of your “GenericWorkflowNodes” checkout. The final dialog should then look as follows.

Clicking finish will import (1) the GKN classes themselves and (2) your generated plugin’s classes.

Now, the packages of the GKN classes and your plugin show up in the left Package Explorer pane of Eclipse.

Hint

Information: Synchronizing ant build result with Eclipse.

Since the code generation happens outside of Eclipse, there are often problems caused by Eclipse not recognizing updates in generated java files. After each call to ant, you should clean all built files in all projects by selecting the menu entries Project > Clean..., selecting Clean all projects, and then clicking OK.

Then, select all projects in the Package Explorer, right-click and select Refresh.

## Launching Eclipse with our Nodes¶

Finally, we have to launch KNIME with our plugin. We have to create a run configuration for this. Select Run > Run Configurations....

In the Run Configurations window, select Eclipse Application on the left, then create the small New launch configuration icon on the top left (both marked in the following screenshot). Now, set the Name field to “KNIME”, select Run an application and select org.knime.product.KNIME_APPLICATION in the drop down menu. Finally, click Run.

Your tool will show up in the tool selector in community/SAM and BAM.

Here is an example KNIME workflow with the nodes that we just created.