Adding Server-side Analysis Tasks to JS9

A JS9 web site can define analysis tools that will be loaded into the user's JS9 Analysis menu. These analysis tasks are then executed by the back-end (CGI or node.js) server using the original FITS data, with results sent back to the browser for display.

To learn how to install the back-end helper, see: Installing a Server-side Helper.

Analysis tools are defined within files contained in the analysis-plugins directory. If a file in this directory ends with a ".json" extension, it is assumed to be an analysis definition file. The following sample analysis definition files will be found in the analysis-plugins directory by default:

If you plan to allow dynamic conversion of FITS files to PNG representation files, the fits2png.json file should be left in the plugin directory. The other two are optional and should be removed if funtools is not available.

Analysis definition files are in JSON format (http://www.json.org/) and consist of an array of JavaScript objects, one for each analysis tool. The properties which define an analysis tool are:

See the sample json files in the analysis-plugins directory for the JSON syntax of this file.

The files property is a rule that determines whether an analysis task is available for a particular file. The following rules are defined:

Examples: Note that all checks are case-insensitive.

The action string contains the command to execute. For security reasons, this action is never executed directly. Instead, the first argument of the command must be a wrapper script contained in the analysis-wrappers directory. The js9Xeq wrapper script contained therein is the default script for running JS9 analysis. The second argument in the action string is a command id, which is executed within the command-processing case statement of the js9Xeq. This wrapper script checks for valid commands and arguments, and then executes the specified command in a way that protects against malicious attack. You can modify this script to meet your needs by adding case statements for each of the new tasks you want to run.

Alternatively, you can run your analysis tasks through your own wrapper script, by placing it in the analysis-wrappers directory. Using your own wrapper script allows you to perform your own special setup (e.g., setting environment variables, cd to work directory) before executing a task. Obviously you can use js9Xeq as an example and modify it to suit your needs.

In either case, note that execution of a wrapper script is accomplished by pre-pending the path of the analysis wrapper directory to the wrapper script name and executing that full pathname directly. This prevents a would-be attacker from executing anything except the wrapper itself. Of course, you still have to be careful to avoid insecure shell coding in your wrapper script. See The World Wide Web Security FAQ for advice on writing CGI scripts. Note that JS9 scripts generally use bash or sh. Please make sure your bash and sh programs are properly patched to protect against the bash bug (CVE-2014-6271).

The action command line can contain macros (strings prefixed with the $ character) that will be expanded by the browser before the action is sent to the server for processing. Typically, these macros are used to specify filenames, regions, and (possibly) user parameters. The following intrinsic macros are defined:

The three regions macros return region info in the current wcs system. To specify an alternate wcs system, append "physical", "image", or "wcs" in parentheses, e.g. $regions(physical).

For example, to run the funcnts program, specifying the original FITS file as well as source and background regions, use:

    funcnts $filename $sregions $bregions

Many analysis tasks do not require the user to select additional parameter options and can can be executed immediately. Other command require user-selected options. These latter options should be put into a web form, whose location is specified by the purl option. See the params/histplot.html file for an example.

Parameter options in a parameter form should have unique ids. These ids can then be added as macros to the action command line by prefixing them with a $. For example, the histplot command is:

    js9Xeq histplot '$filename[$regions]' '$norm $bwidth' $column $bins
Here, the norm, bwidth, column, and bins parameters are taken from the web form for histplot. The $filename and $regions macros are intrinsic to JS9.

The input element of the analysis form should run the JS9.SubmitAnalysis() command, which will serialize the form, pass it to the macro expander, and then send the resulting command string to the server for execution. See histplot.html for sample syntax.

The rtype property determine how JS9 handles return values:

Sample scripts such as funcnts2flot and funhist2plot generate a JSON string that contains a valid flot object. For example, the radial profile plot contains x, y, and error values:
{
  "label" : "surf_bri(...) vs. avg_radius(...)", 
  "points" : {"errorbars" : "y", "yerr" : {"show" : "true", "color" : "red"}},
  "data": [[x1, y1, e1], [x2, y2, e2], ... [xn, xn, en]]
}
while the histogram plot contains only x and y values:
{
  "label" : "counts vs. ...bin", 
  "data": [[x1, y1], [x2, y2], ... [xn, xn]]
}
If your analysis task generates one of these two plot formats, JS9 will display the plot data. By default, the plot is generated using the jQuery flot library (see http://www.flotcharts.org/), which is supplied with JS9.

You also can use the plotly library for plotting. To do this, set the JS9.globalOpts.plotLibrary property to "plotly" and load the library in the web page header. (Due to its relatively large size, plotly is not packaged in the JS9 support file.)

Plotting is still somewhat experimental and rudimentary. For example, flot (but not plotly) simply uses the x and y keys to toggle between linear and log scaling. We will be adding features and improvements with time. Please work with us to improve plotting support.

Finally, the hidden value determines whether or not a task is displayed in the Analysis menu. If hidden value is true, the analysis task is not shown. However, it still is available to the JS9.RunAnalysis() routine. The fits2png command in the fits2png.json definition file is an example of a hidden command: it is run indirectly by JS9 and need not be listed as a user task.

Adding Analysis Tasks Directly to Your Web Page

By default, server-side JS9 analysis tasks are executed using the Analysis menu, but they also can be executed directly from the web page by means of HTML elements (buttons, forms, etc.) To do this, a web page author simply creates the desired interface and calls either JS9.RunAnalysis() (for buttons, menus, etc.) or JS9.SubmitAnalysis() (for forms). These are described below. See the js9analysis.html page for an example.

The JS9.RunAnalysis() routine is used to execute an analysis task and return the results for further processing within the web page. The JS9.SubmitAnalysis() routine works similarly, but it is used with a form: it automatically serializes the form values and passes them to the analysis macro expander so that these values can be put into the analysis command line. The calling sequences for these routines are similar:

    JS9.RunAnalysis(aname, [options], returnFunc);
    JS9.SubmitAnalysis(this, aname, returnFunc);
where:

The returnFunc() routine is a callback function to process the returned data from the analysis task. The calling sequence is:

    returnFunc(stdout, stderr, errcode, aobj)
where: Typically, you would check the errcode and/or stderr string first and issue an error message if there is an error. Otherwise, the stdout string can be processed based on the return type (rtype) of output (e.g., "text" or "plot"). For plotting, you can use the flot functionality that is loaded into JS9, or you can use your own chosen plotting package. See js9analysis.html for a simple example.

Temporary Work Directories for Users

By default, server-side analysis is initiated in the directory in which Node.js or the CGI script is run. For tasks not creating temp files, this usually is sufficient. For other cases, it can be desirable to isolate users in their own temporary work directories. To do this automatically (i.e. without the task itself having to change directories), set the following property in the globalOpts object of the js9Prefs.json file:

You also can set the following globalOpts configuration parameters in js9Prefs.json: When workDir is configured, user sub-directories will be created in the workDir directory to contain temporary files. Analysis tasks will be started in the work sub-directory.

By default, the work directory is removed when the web page is unloaded, but this can be changed by setting rmWorkDir to false.

Note that the workDirQuota is not a hard quota: JS9 tasks such as load proxy enforce their own checks using this value. For example, load proxy only checks whether the quota has been exceeded before loading the next file. This means that users can load one file that will bring the total size above the quota. Thus, if the quota is 10Mb and the work directory is empty, the load proxy will load a 50Mb file. But it will not more load files until already-loaded files have been closed in JS9 (and thereby deleted in the temp directory) so as to bring the total under the quota.

The Proxy Load Service

NB: use this service with care. It allows users to consume disk resources!

For security reasons, JavaScript contained in one web page can access data in another web page only if both web pages have the same origin (i.e., basically coming from the same host). This policy is called the Same Origin Policy. This means that JS9 cannot load a FITS file from an arbitrary URL without using special techniques.

One such technique is to use a proxy server: the URL is not loaded directly into JS9, but instead is copied back to the server from which JS9 itself was loaded. The file is then retrieved by JS9 so that the "same origin" rules are not violated.

A CGI or Node.js back-end server can be set up as a proxy server by setting the following globalOpts properties in the js9Prefs.json file (before starting the server):

When the back-end server supports the proxy service, you will see a menu option in the File menu called open link via proxy ...

Enter a FITS URL into the proxy dialog box and press the Load button. The data file will be retrieved by the server and stored in a directory specifically tied to the web page. (The directory and its contents will be deleted when the page is unloaded.) JS9 then will load the file from this directory. Note that since the file resides on the back-end server, all back-end analysis defined on that server is available.

This technique can also be used FITS links from other sources. For example, if you have a FITS file on Google Drive, Dropbox or even a static web page, you can generate/copy the link and paste it into the proxy dialog box for display. (Dropbox files also can utilize the open link via CORS ... option, which downloads the data directly to JS9, bypassing the proxy server.)

Note that individual proxy FITS files are deleted from the working directory when they are closed in JS9 using the File:close image menu option.

Last updated: September 6, 2016