3270 Application Modernization Using OpenLegacy, Part 3: Generating the Application Model

by Frank Griffin, Senior Software Developer for Treehouse Software 

This is the third in a series of blog posts concerning the Treehouse partnership with OpenLegacy. As discussed previously, OpenLegacy is a way to wrap legacy mainframe applications for presentation to Java, Web, and Mobile clients.

In the last post, we covered the process of using OpenLegacy run-emulation to monitor 3270 traffic as you interacted with CICS to use the CEMT transaction to obtain information about one of the standard CICS transaction programs. This resulted in a trail file – an XML document which described the interaction in detail. This time, we’ll go through the process of generating the application model, which is a set of Java classes that represent each 3270 screen and provide methods for reading and writing the fields on the screen. These classes will be used later by the service parts of the application which interact with the 3270 screens on behalf of the web client.

If you still have the OpenLegacy IDE open from when you did run-emulation, you’ll need to refresh the CICSWeb project to make it aware of the new trail file, which has been created in src/test/resources/trails. Right-click on the project and select “Refresh”:


OpenLegacy will detect that a new trail has been created, and will ask if you want it analyzed.


We’re going to select “No” at this point, in order to go through the model generation process in detail. This will return you to the usual IDE view of the project. Expand the project and drill down through src, test, resources, and trails to expose the newly-created trail file:


and double-click on the trail file to open it using the Trail Editor:

You’ll notice that the editor opens in Snapshot mode by default, showing the first screen in the trail file. If you want to see the actual XML source, you can click the Source tab at the bottom left of the editor window and toggle back and forth.

In order to see the index of screen thumbnails, maximize the editor window. Then right-click on Screen 1, and select Generate Screen Model:



Selecting Generate Screen Model will get you a prompt:


Uncheck Generate JUnit Test, and click OK. This will open the model generation dialog for Screen 1:


At the top, the Fields table should contain one entry for each field on the screen which can contain dynamic data, that is data that can change from client request to client request. These fields may be unprotected (designating input fields to be populated with input data from the client) or protected (designating output fields to be populated with output data to be sent to the client). In our case, there is only one: the command field on line 24. OpenLegacy will have chosen a default name for the field based on assumed usage or text appearing to the left which might be a label, and if the name has multiple segments, e. g. New Password, the name will appear in camel-case (newPassword) since the name will be used as a field name and as part of method names in the generated Java code.

The Identifiers table to the right will contain OpenLegacy’s best guess as to the fields on the screen which uniquely identify the screen. OpenLegacy can’t really know which protected fields on the screen represent static data that can be used to identify the screen and dynamic fields which may change, so this is one of the areas where you may have to dive in and correct the default guesses. Your screen model must include at least one Identifier; otherwise, OpenLegacy can’t identify a screen. For screen matching, identifiers are additive, so you can specify as many as needed to identify the screen uniquely within the trail file. Conversely, if OpenLegacy has tagged a field as an Identifier but the field can change from use to use, you’ll need to delete it as an Identifier by selecting it and clicking on the X at the top right of the table.

Note that OpenLegacy has chosen a name for the screen (ApplicationDeveloperSystem) based on text it found on the screen as an Identifier. This will be used as part of the Java class name for this screen, so it will need to be unique among the screens in the trail file.

In our case, the guesses for Screen 1 are adequate, so we’ll leave everything as is. Scroll the window down, if necessary, to expose the Cancel/OK buttons:


Once you’ve finished, click OK at the bottom of the window, which will open a properties dialog:


Much of the data on display here is the same stuff you saw on the previous screen, e. g., Name, Display Name, but other data has been taken from the trail file, such as the screen size. If you select the Fields and Identifiers tabs at the bottom, you’ll see the same Fields and Identifiers as well, but with additional trail data. Select the Fields tab, and note that the position of the field on the screen is recorded, as is the fact that the field is unprotected (“editable”):


Going back to the General tab, select Screen Navigation rather than Screen Entity:


Most of these fields are beyond the scope of this discussion, but note particularly the Accessed From field. I mentioned before that OpenLegacy creates a Java class for each screen, with methods that can be called to interact with fields on the screen. For application simplicity, users of these methods do not have to navigate to the desired screen manually. You can simply call the method associated with the field you want, and OpenLegacy will figure out what it has to do in the way of 3270 screen navigation to get there. It does this by using Accessed From and other navigation data here, and correlates this information across all of the screens in the trail. In this case, Accessed From is blank because this screen is obtained simply by connecting to the mainframe.

Lastly, notice the tab in the Screen Editor for this display: it says “ApplicationDeveloperSystem.java”, because you are looking at a graphical rendition not of the trail file, but of actual Java code which OpenLegacy has generated. Remember that Eclipse editors are capable of showing the same file in multiple ways. To illustrate this, select the bottom notebook tab marked “Source (read only)”:


The tokens starting with “@” are called “annotations” in Java, and Java provides language support for reading annotation data in compiled class code without having to decode the class file yourself. Note that all of the data displayed in the other notebook tabs is represented here as attribute fields in one annotation or another. OpenLegacy generated this code from the trail file data for this screen and compiled it when we clicked OK. Then, the Screen Editor filled in the fields on the notebook tabs by retrieving the annotation data from the compiled code. Once we clicked OK, we were done with this screen’s section of the trail file; the reference location for all of the screen’s data from now on is the generated Java code. If we make any changes to this data using the Screen Editor, saving the changes will result in regenerating the Java code with new values in the annotations and recompiling it.

The tab says “(read-only)” because any manual edits you make here won’t be saved. During Generate Model, OpenLegacy is dealing only with automatic code generation. Manual modifications and additions, if needed, are done later using the standard editor.

You’ll note that in the Trail Editor, screens are marked either “(IN)” or “(OUT)” depending on whether they came in from the mainframe or went out to the mainframe from the emulator. Only (IN) screens have a Generate Screen Model option in their right-click menus.

Moving on, let’s close ApplicationDeveloperSystem.java, right-click on Screen 3 back in the Trail Editor, and select Generate Screen Model:


We see considerably more activity here than on Screen 1. OpenLegacy has assigned this screen the name SignonToCics based on the top banner line, and because of the labels and the presence of non-display fields, it has selected a Screen Type of “Login Entity” rather than “General”. As before, various fields and identifiers have been pre-selected for you, and based on labels and field attributes “user” has been given the status of a userid field and “password” has been designated a password field. This means that these fields will be filled in using authentication data supplied by the client. Note that the community version of OpenLegacy supports only userid/password authentication, so no special status has been given to the “groupid” or “newPassword” fields since client protocols, e. g. HTTP, don’t make any provision for them.

Once again, the default choices for Fields and Identifiers look suitable, so we’ll leave them as is. Scroll down and click OK to generate the code and open the Screen Editor, then select the “Source” tab:


This looks pretty much like before, but as we have more fields, we have more @ScreenField annotations.

However, we have some customization to do here as regards navigation. Select the “General” tab again and “Screen Navigation”, where we will need to fill in Accessed From to indicate that we are accessed from ApplicationDeveloperSystem and we will need to indicate that we are accessed from that screen by entering “l cicsts51” in the “command” field of the previous screen.

It’s worth noting that if we had chosen to allow OpenLegacy to do Generate Model on the entire trail file at once, automatically moving from one screen to the next, it would have filled these fields in for us. We need to do it manually here only because we are processing each screen standalone.

To fill in Accessed From, we don’t just type the previous screen name; OpenLegacy needs to know the name of the Java class that represents the previous screen, not just the screen name. Click on the “Browse” button next to Accessed From, and you’ll get a standard Eclipse class browser with nothing displayed. Type “Application” in the top input field, and Eclipse will display all of the known classes beginning with that string:


Double-click on ApplicationDeveloperSystem in the list, and that will be filled in as Accessed From.

Now, it may seem counter-intuitive to be dealing with “l cicsts51” in Screen 3, since in the trail file it was entered on Screen 2. However, it shows up here because the Screen Navigation tab for Screen 3 describes what needs to be done to get to Screen 3, not what happens once you’re there, so this tab is where we stipulate that the fixed string “l cicsts51” has to be entered in “command”. We do this by scrolling down to Assigned Fields and clicking on Add, which will add a blank row to the Assigned Fields table. Then just type a Name of “command” and a Value of “l cicsts51”:


Note that as a result of these activities, the Screen Editor tab for SignonToCics.java now has an asterisk indicating that it has been modified. Change to the Source tab, and you’ll see that it is unchanged from before; then save the changes by pressing CTRL-S, and watch how the source changes, adding a @ScreenNavigation annotation with your changes:


Close the SignonToCics.java Screen Editor tab and return to the Trail Editor. Now proceed to Screen 5, right-click and select Generate Screen Model. Surprise – nothing happens ! This is because Screen 5 has no Fields, and therefore OpenLegacy has nothing with which to work. In this case, we’ll need to define the screen ourselves, using right-click and New Screen Model, which will give us a display with no Fields or Identifiers filled in:


We’ll assign a Screen Name of SignonIsComplete. We need to have at least one Field (whether it will see any use or not), and the unprotected field at the right end of line 24 will do nicely. To add a Field, you must select/trace it on the screen display by left-click and dragging the mouse, and then click on one of the Field Type icons at the top of the Fields table. We’ll select the area and click on the S icon (String):


We’ll also need an Identifier, and the first part of the message will suffice. To add an Identifier, trace it out on the screen and click on the + icon at the top of the Identifier table:


Scroll down and click OK; no customization is needed, so you can just close SignonIsComplete.java. Note that we did not fill in Accessed From or any assignable fields in Screen Navigation, because all data entered on the SignonToCics screen is dynamic (userid and password), and the AID key which gets us here is the default (ENTER).

Screen 7 is another challenge. It results from hitting CLEAR to Screens 5/6, and has neither Fields nor Identifiers that OpenLegacy can recognize, so Generate Screen Model once again has nothing to work with, and we must use New Screen Model, which we’ll name “Empty”. Adding a Field is just like before; we’ll use the first line of the blank screen, and rename the default name of “newField1” to “command”. However, to provide the Identifier, we must use a trick since there is no identifying text present. Since this screen can be differentiated from the others by the fact that it begins with an unprotected field in row 1 column 1, we can define a zero-length identifier there (start (1,1) end (1,1)) and stipulate that the field must be editable. Adding the Field:


Adding the identifier is not so straightforward, as there is no way to trace a zero-length field. We’ll do it after the fact using the Screen Editor.

Scroll down and click OK to open the Screen Editor, and select the Identifiers tab at the bottom to display the empty Identifiers table. Click “Add” to add a blank Identifier row to the table and change the row/column from 0/0 to 1/1. Type “true” in the Value column (this is the value of the attribute, not the Identifier text). Now click on the Attributes column, which will give you a dropdown list of attribute types, and select “Editable”:


There’s one more bit of configuration needed here. The navigation from Screens 5/6 to Screen 7 doesn’t involve any assigned field, but it does involve hitting the CLEAR key (ESCAPE in the browser emulator), which is not the default AID (ENTER) for screen navigation. Thus, we need to go to Screen Navigation, set Accessed From to SignonIsComplete, and Terminal Action to ESCAPE using the dropdown list:


One screen left – Screen 9. Close Empty.java, right-click on Screen 9, and select Generate Screen Model:


While we could return all of this information to the client, for simplicity we’re just going to return the program length and use count. The Fields and Identifiers on this screen are not standard OpenLegacy Fields and Identifiers, because the “command” field must be created dynamically using the program name supplied by the client, and the program length and use count are both parts of a larger 3270 field, e. g. “Leng(0000000624)” and not just “0000000624”. But no matter, this will give us the chance to expand our experience with OpenLegacy configuration.

We don’t have to have a Field, but we’ll leave the “command” field selected by OpenLegacy, even though we have no use for it. Note that it’s an example of a Field which is protected, but whose contents will change dynamically depending on the program name supplied by the client. In fact, none of the other chosen Fields are needed either, so we’ll delete them by highlighting them one by one and clicking on the X at the top right of the Fields table:


Now, we come to the Identifiers. None of them are suitable, because “Pro” hardly uniquely identifies the screen, and the other two will change dynamically depending on the program name supplied. So, we’ll delete them all just as we did the Fields – select and click X – and add a unique Identifier of the fixed portion of the CEMT command by tracing it and clicking + at the top left of the Identifiers table:


Now we need to add two additional Fields for the data we wish to extract. The values for the length and use count are integers, so in each case trace the integer portion of the field (within the parentheses) and click on the I (Integer) icon on top of the Fields table to create the Fields. Select the fieldnames and overtype the defaults with “length” and “useCount” respectively:


Scroll down and click OK to generate the model and open the Screen Editor. No additional configuration is needed here, so it can just be closed.

We’re done generating the Screen Models. But wait, you say, don’t we have to create the CEMT command from the program name ? And doesn’t Screen 9 need an Accessed From of Empty ?

The answer is no in both cases. The job of the Screen Model is just to provide program getter/setter access to screen fields of interest. The classes in the Model will be used later on by the other part of the generated application, called the Service, to provide dynamic data to the 3270 emulator and retrieve output field values.

And Accessed From is only needed if static data (like “l cicsts51”) needs to be entered or if the transition AID to the screen differs from the default of ENTER. In our case, the CEMT command to be entered on Empty will be built dynamically by the Service and the Empty class setter for the “command” field (the “command” on Empty, not the one on ResultsOvertypeToModify) will be used to inject the command into the Empty screen. The transition AID for the Empty to ResultsOvertypeToModify transition is the default of ENTER, so neither condition applies.

That wraps up this post. OpenLegacy has taken our trail file, and with a little GUI-only tweaking from us has generated a complete set of Java classes with simple-to-use methods like setCommand( “CEMT INQUIRE PROG(“ + programname + “)”) and getLength() which will drive the legacy application through the 3270 emulator to supply our client’s input data and retrieve his output data. If you’re so inclined, you can see this Java code by expanding the src/main/java directory tree under the project.

In the next post, I’ll cover generating the Service portion of the application and making the few needed code modifications.

3270 Application Modernization Using OpenLegacy, Part 2: Creating the Trail File

by Frank Griffin, Senior Software Developer for Treehouse Software 

This is the second in a series of blog posts concerning the Treehouse partnership with OpenLegacy (See Part 1 here). As discussed last time, OpenLegacy is a way to wrap legacy mainframe applications for presentation to Java, Web, and Mobile clients.

We’ve already covered the conceptual workings of wrapping 3270 applications in general, so this time we’ll examine the first steps of wrapping an actual 3270 application in detail.

If you haven’t already downloaded OpenLegacy, you’ll need to do so in order to follow along in this Proof of Concept. Go to http://www.openlegacy.com and select Resources → Download. You’ll need to provide your name and an email address, which will get you a userid and password for the download site at ftp://ftp.openlegacy.org .

CICS is one of the bulwarks of legacy 3270 applications. These days, it also supports a variety of other communications protocols, but for 3270 it acts very much like a Windows/DOS/Unix command line. You type a transaction name followed by optional arguments, and hit ENTER. CICS looks up the transaction name in a table to find the application program associated with it, and runs that program. The program writes to the terminal, either in line mode or using formatted 3270 screens, and may request further input from the user and provide additional terminal output.

CICS provides a number of utility transaction names with the product. We’ll be looking at two of them today: CESN (sign on to CICS as a known user), and CEMT (general multi-purpose transaction for querying and controlling aspects of CICS). We’ll construct a demo application by connecting to CICS, signing on with the CESN transaction and issuing the transaction CEMT INQUIRE PROG(xxxxxxxx) to obtain information about a specific CICS transaction program. The input to the application will be a userid, password, and application program name. The output will be some of the data associated with that program.

Open the OpenLegacy IDE, and choose New → OpenLegacy Project. Select “Screens” as the backend solution type, “Integration Web” for the frontend solution type, “Mainframe” for the host type, and supply your host’s name or IP address:



Now, we need to run an emulation session to “teach” OpenLegacy how to navigate. Right-click on the project and select OpenLegacy → run-emulation. This will set several things in motion. It will start an internal web server within the IDE on port 1512 and launch an instance of your default browser with a URL of http://localhost:1512. When the browser connects to the web server, it will execute a servlet that will use the open-source s3270 scripting emulator to connect to your mainframe and report the 3270 datastream to the servlet, which will convert the screen to HTML and send it to the browser.

The result is that your browser instance will appear to be a 3270 screen, and you’ll be able to interact with it as if it were a real terminal or terminal emulator. As you do, the servlet is sitting in the middle, logging all terminal activity between you and the mainframe. On our mainframe, the first screen you’ll see is the standard VTAM Logon screen:


We logon to CICS (the VTAM logon screen and the command you use may differ at your site), and depending on how CICS is configured, you may need to type “CESN” explicitly, or CESN may be started for you automatically:


Fill in a valid Userid and Password, and hit ENTER:


Clear the screen (ESC key in the browser emulator), as CICS transactions have to be invoked from a clear screen, type “CEMT INQUIRE PROG(DFHZNEP)” (DFHZNEP is a dummy network error program supplied with CICS), and hit ENTER:



The CEMT transaction has displayed all sorts of information about the DFHZNEP program, which is what we wanted.

At this point, we could continue with 3270 navigation to exit and signoff from CICS, but for simplicity in this example, we’ll just cut off the emulation here. This will cause OpenLegacy to disconnect the TN3270 session, which will cut us loose from CICS; if this were TSO, we’d have to follow through or risk leaving disconnected TSO sessions active.

Note that depending on the size of your browser window you may have to scroll down to make the “Logoff | Flip” links visible. Now click on “Logoff”, which will disconnect the terminal, shut down the emulator, and create the trail file which describes everything you’ve done:


The trail file is an XML document which describes in detail your interactions with the mainframe. It will be the input to the next phase of generating the application, which will be the subject of the next post.

Can OpenLegacy help connect PHP with DB2 on the mainframe?

Recently, a poster to a mainframe technical discussion forum asked the question: How can we connect PHP (a popular server-side scripting language designed for web development) with DB2 on the mainframe? Treehouse Senior Software Developer Frank Griffin replied, describing how OpenLegacy could be the answer.  

You have some options here. If all you want is access to the raw DB2 data, JDBC or ODBC access will work fine for you, although you will have to write either C ODBC or Java JDBC code that can be called from PHP to do the deed.

If you already have mainframe code accessible via the network (CICS/IMSDC/TSO via TN3270, CICS via TCP/IP) that accesses your data and adds business logic to the mix, you can use the FOSS* OpenLegacy project to mate this logic to your PHP app.

The simplest approach involves using OL to navigate through the green screens of (CICS, IMS/DC, TSO, or whatever) to get to the data you want. This is done via the OSS** s3270 scripting 3270 emulator which creates an XML “trail” file which documents the interactions between yourself and the legacy application over the TN3270 connection. Once you’ve navigated through the series of screens that exposes the data you want to find, you trigger a portion of OpenLegacy which analyzes those screens, identifies logon sequences and unprotected fields on those screens which are linked to client-supplied input data, and generates Java classes which can at some later time use the “trail” file to re-drive the emulator to access the legacy application using client-supplied values for some of the input in order to obtain transaction output associated with those input values.

If your mainframe apps are better modularized, i. e. if you have separated the business logic from the display logic, and the business logic can be invoked through a CICS COBOL program which is designed to obtain its input and provide its output via a CICS COMMAREA, OL can analyze the source code and generate Java classes that invoke those programs directly without screen-scraping.

And, if what you really want is JDBC, OL can generate a series of Java classes that do all of the JDBC work for you and provide you with methods that simply provide the legacy data with no hint as to where it came from.

OL can also layer additional access software. Once the fundamental Java classes that access the legacy data are in place, OL can generate Java apps that use those classes on behalf of clients using all sorts of modern APIs, including SOA/SOAP, REST/JSON, and Mobile. All of this happens with the push of a button.

All of this is FOSS. You can download the OL code and start using it immediately, and it can do all that I’ve described, out-of-the-box. OL makes their money from selling an Enterprise Edition that includes support and some security and management pretties.

If the only access path from PHP to Java is direct invocation of Java code, you’d have to write a Java stub to interact with the OL classes, but this s going to be a *lot* simpler than trying to write JDBC or ODBC applications on your own.

*FOSS = Free Open-Source Software

**OSS = Open-Source Software


OpenLegacy is the world’s first and only light-weight, non-intrusive solution for automated legacy modernization and enterprise application integration. With its standards-based, open-source platform, OpenLegacy enables enterprises to rapidly extend legacy systems to mobile, web and cloud applications; delivering risk-free, high-impact results that solve immediate business needs.


OpenLegacy‘s standard tools rapidly extract the services and information from within legacy systems into an editable format that puts the power of integration into the enterprises’ hands without the expensive handcuffs of vendor lock in. Once a business process is exposed — which can be done in minutes — the output can automatically be transformed into stand-alone mobile, web, and cloud applications; and connected with other solutions. Most importantly, no changes are required to the legacy system in order for OpenLegacy to work — the process is risk-free.

Contact Treehouse today for more information!

Treehouse Software Partners with OpenLegacy

Treehouse Software Partners with OpenLegacy

By Wayne Lashley, Chief Business Development Officer for Treehouse Software

Mainframe legacy systems, be they Software AG Adabas/Natural, CA IDMS/ADS, CA Datacom/Ideal or VSAM/COBOL, have several attributes in common:

  • They represent an enormous investment and body of proprietary business knowledge and process that are not easily replaced;
  • They are mission-critical, reliable and well-managed;
  • As originally conceived, they do not lend themselves readily to today’s common technology standards and practices: Java, REST, RPC, Web Services and mobile and Web apps — not to mention whatever is coming next.

The last point is a principal rationale for the rise of the legacy modernization industry in recent years. Various ways and means of “modernizing” legacy applications have emerged, and these are well-known and well-documented in the industry, and even here in the Treehouse Software blog and its predecessor, the Treetimes newsletter. After a few years of shakeout and consolidation, it would seem that there is little new under the sun in terms of legacy modernization practices and practitioners. But that’s not so.

It was back in 2013 when we first encountered OpenLegacy, a new entrant in the field and one with a novel approach. Nobody else in the modernization biz seems to be offering a standards-based, open-source toolkit/platform that can open up any IBM z or i legacy environment to provide Web and mobile interfaces, Web services and APIs—without migrating the legacy application or changing its code. With OpenLegacy, the value of the legacy environment can be fully leveraged in today’s technologies — and in whatever comes next.

We’ve been watching the company and its offerings evolve over the months, and Treehouse Senior Software Developer Frank Griffin participated in a technical evaluation and went through OpenLegacy training. We were all pretty impressed—so much so that we recently inked a partnering agreement with OpenLegacy to help market and deliver their solutions in North America. In the post below, Frank discusses one aspect of how OpenLegacy can be used to open up mainframe 3270 applications. Stay tuned for more posts from Frank as our OpenLegacy journey continues.


Treehouse Software is now partnering with OpenLegacy to provide access to legacy IBM mainframe applications for a range of non-mainframe devices, including mobile phones and tablets as well as “heavier” clients like Service-Oriented Architecture (SOA) consumers.

Unlike many modernization approaches that require a commitment to migrate the legacy application in some way or at the very least require legacy application code changes, OpenLegacy adds more modern potential client populations to the existing application clients with no required changes to legacy code.

OpenLegacy is open-source, and is built using standards-based protocols and other open-source components wherever possible. The starter version is free to download, so you can get started with your testing immediately.

OpenLegacy can interact with legacy applications in several ways, but for simplicity in this initial post I’ll concentrate on just one: access to 3270-based applications.

We’ll assume that you have a 3270 application which can be made to display data required by a non-legacy client by navigating through one or more screens. The data need not all appear on a single screen, but can be spread over several screens. All you have to know is how to navigate through the application to get the data you want displayed.

OpenLegacy development starts with a developer environment which allows developers to describe:

(a) any input parameters needed to display the desired data

(b) the navigation process to arrive at a screen containing data to be captured

(c) the location on that screen of the data to be captured, as well as a description of that data

The OpenLegacy development suite runs as a plugin for the Eclipse Integrated Development Environment (IDE) or a self-contained IDE installation, which has several advantages. It allows OpenLegacy to piggyback on the rich GUI which Eclipse provides, and provides a familiar IDE environment for customer developers.

The OpenLegacy plugin opens several custom windows within Eclipse:


In the upper right, you can see a window with the current state of the 3270 emulator, which initializes at the standard initial logon screen for VTAM applications. In the upper left, you can see the directory structure which is automatically created for you when you use the OpenLegacy wizard to create a project. In project creation, you specify the target mainframe host, so OpenLegacy can open a TN3270 session to that host.

Navigation directions are entered by having OpenLegacy open a browser-based 3270 emulator which uses a captive copy of the s3270 emulator to interact with your mainframe system under OpenLegacy’s control. You use the browser-based emulator exactly as you would a 3270 terminal to log on, enter requests and input data, and do whatever a 3270 user would do to get to a screen of interest. However, because the 3270 emulator is running under OpenLegacy’s control, OpenLegacy is recording every keystroke needed for the navigation.

You can see the browser-based emulator in the following screenshot:


The browser application captures your keystrokes and feeds them to the 3270 emulator, also capturing the input and output from the emulator. These traces are referred to by OpenLegacy as “trails”, and are saved as XML files. When your project actually executes, this file is used instead of browser keystrokes to provide input to the emulator and scrape data from the emulator output.

When you’ve arrived at a screen containing data of interest, you switch from the browser window back to the OpenLegacy Eclipse GUI, and let OpenLegacy analyze the screen contents, identify the fields on the screen, and allow you to select those of interest and identify them with names and datatypes.

This navigation/selection cycle continues until you’ve identified all of the data you wish captured. At that point, you literally just push a button, and OpenLegacy will generate a complete rich web Java application which will accept whatever input is needed for the navigation through the legacy application, and then use the captive terminal emulator behind the scenes to log into the application, navigate to the screens of interest, and capture the fields of interest.

For Java developers (depending on what you’ve asked for), OpenLegacy also generates front-end web/Java code to use this application to accept input and return output using a number of mobile and web protocols, including SOA, HTTP GET/POST, or REST/JSON, making the legacy application immediately accessible to a wide range of modern devices.

Once compiled and exposed to the client community via a servlet container, these web/mobile applications accept requests in the specified input protocol, extract the input argument data, call the core Java code to drive the legacy application filling in fields and simulating keystrokes as needed, and collecting the desired output from screen fields as previously directed. The collected output is then formatted by the front-end application according to the requirements of the protocol involved, and returned as a response. Your device has no idea that it is interacting with a 3270 application, and your 3270 application has no idea that it is interacting with anything other than a 3270 terminal.

Oh, and did I mention that the Eclipse container includes a servlet container, and (since it was designed to allow developers to write/generate code and immediately compile and test it) a few additional mouse-clicks compile the generated code, create JARs and WARs and deploy them to that servlet container ? The result is that in as little as an hour from starting this process, you can be interfacing to your legacy application from your phone for testing. All you need is someone who knows how to navigate the legacy application and someone who knows how to navigate the OpenLegacy Eclipse GUI–two completely separate skill sets.

Regardless of what you are considering for the future for your legacy applications, OpenLegacy provides an immediate way to vastly increase your client population without touching a line of legacy code.

If this just seems like screen-scraping to you, look more closely. The differentiator here is the transparent creation of those front-end interfaces with no need for you to understand those technologies. Development using OpenLegacy doesn’t require you to know anything other than how to use the existing application.

In future posts, I’ll discuss other ways to access legacy applications using OpenLegacy via mainframe protocols that bypass 3270 emulation.

Contact Treehouse today for more information!