CobolSharp

This is a tool to extract code structure from COBOL written according to mid-80's best practices and revisualize it as more modern code structures. The purpose is to make it easier to analyze legacy code to understand what it does, extract the core business logic, and then reimplement it in modern languages.

But it is not a tool to translate Cobol code fully into a modern language, since it only considers the code, not the data. Even though the output code might look like it could be run if the tool could just translate the expressions and data structures too, the result would just be COBOL code in fancy dress. The COBOL data model is so far removed from what you'd use in C# or Python that the translated code would not be much less messy to maintain than the original code.

Future extensions to this tool could usefully start looking at the data, though. Perhaps not to translate COBOL fully into another language, but to show which sections share working storage and move local variables into functions or even inner scopes.

Example output

(This fragment shows a bit of the first example here: http://docs.oracle.com/cd/A57673_01/DOC/api/doc/OCI73/apb.htm It was necessary to change ORA-ERROR into a section for CobolSharp to process it.)

Background

In COBOL-74 an if block was best limited to a few simple statements, avoiding any nested if since that would likely end up associating an else block with the wrong if. There where no syntax for loops except for performing another section or paragraph repeatedly. On top of this the layout of each code line had restrictions that discouraged or even disallowed using indentation to indicate code structure.

As a result, what is today implemented with nested, indented code blocks was implemented with gotos, typically enforcing a paragraph naming scheme to help get some idea of the code structure.

COBOL-85 added proper code scope terminators (i.e. if ... end-if., the switch statement evaluate and inline perform loops, but much code continued to be written in mostly a COBOL-74 way.

Installation

CobolSharp requires Python 3, and a Java runtime installed in $PATH. It has been tested on Linux and Windows, and should work fine on any Unix-like system too. Graphviz is needed to plot code graphs, but not for generating code.

CobolSharp can be installed systemwide from PyPi:

pip3 install cobolsharp

Or for your own user, to avoid installing as root:

pip3 install --user cobolsharp

An executable script (or binary on Windows) called cobolsharp is installed. If you install with --user it may not be in $PATH, but in ~/.local/bin/cobolsharp.

From code

You can also clone the repository at https://github.com/petli/cobol-sharp and install from the code:

python3 setup.py install

For development setup it is recommended to use a virtualenv, e.g.:

virtualenv --python=python3 ~/test/cobolsharp
~/test/cobolsharp/bin/python setup.py develop
~/test/cobolsharp/bin/cobolsharp --help

Unit tests

There's a small test suite:

~/test/cobolsharp/bin/python setup.py test

Usage

Run cobolsharp --help to see detailed help on all command line flags.

CobolSharp can produce a number of output formats, chosen with the -f flag:

html: COBOL code and translated code in a web page (default format)
code: translated code written to a source file
full_stmt_graph: A graph of all COBOL statements
stmt_graph: A graph of all reachable COBOL statements
cobol_graph: Cobol code structure graph
acyclic_graph: Code structure graph with loops identified and broken up
scope_graph: Code scope graph where each loop scope and exit nodes have been identified
xml: Koopa XML parse tree, mainly useful during CobolSharp development

Cross-referencing code

The html format (the default) creates a standalone web page with both the original COBOL code and the translated code side-by-side.

Clicking a line in either of the code columns will scroll to and highlight the corresponding line in the other one (if there is one).

Larger code blocks (currently five lines or more) in the translated source can be folded and unfolded. There's a set of buttons to fold/unfold everything or all function levels.

perform and goto statements in the translated source have a small link button in the margin, which navigates to the definition of the referenced section or label. Browser navigation can be used to go back and forward between visited sections of the code.

The indentation level in the translated code is colour-coded, and the corresponding line in the COBOL code has the same colour. This can be turned off with checkboxes.

Line numbers and navigation buttons can be turned off to make it easy to cut-and-paste code from the page into a separate file.

Plotting graphs

The graph formats produce .dot files. They can be plotted into PNGs (and many other formats) with graphviz, e.g.:

dot -Tpng -O *.dot

Limitations

This tool will only work well for code that follows best practices on writing structured COBOL. Mainly:

There are no cross-section go to jumps.
perform is only used to pass control to a single section, not a suite of them or of individual paragraphs. I.e. a section must behave as function.

In addition the tool only understands the COBOL supported by the Koopa parser, and may not handle all statements correctly even if they are parsed well. See the project issue list for current outstanding issues, and log a new issue, preferably with example code, if you find something more.

Translation process

The COBOL code is parsed by Koopa into an XML parse tree, which is translated into a graph of statements. This is refined in several steps to identify structured code. For details on this, see the documentation strings and comments in CobolSharp/graph.py.

The final graph is then flattened into linear code. Several strategies are used to decide which representation to use, a trade-off between avoiding code jumps but also not producing very deeply nested code. For details on these strategies see the code comments in CobolSharp/analyze.py. The weightings used in these decisions can no doubt be tuned to produce better output, and any improvements on them are most welcome.

The code is flattened for a local optimization, without attempting to find the optimal representation for a larger section of the code graph. This can also be an area for improvements.

The --debug command line flag will insert comments into the translated source (overwriting any comments in the same place in the original source) explaining the decisions taken at each loop and branch statement.

But the translated code still contains gotos!

Yes, unfortunately it often will. There are pathological cases which cannot be resolved in a structured way (see e.g. test/crossedbranches.cbl), but the the trade-off mentioned above will also keep some gotos in more complex code, and maybe even add labels that aren't in the original COBOL code. This is another reason why this is mainly a tool to better understand COBOL code, not a tool to translate it fully.

License

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

Koopa

The Koopa parser generator is distributed under a BSD license. See http://koopa.sourceforge.net/ for more information.