BANCStar ([email protected])

The Worst Programming Environment in the World?

This is BANCStar source code:

2999,,,
8600,,,1
3000,829,6,30089
10829,2446,22245,22012
11585,15855,22002,22002
11586,15865,22002,22002
11587,15875,22002,22002
11588,15885,22002,22002
11596,15965,22002,22002
11597,15975,22002,22002
11598,15985,22002,22002
11599,15995,22002,22002
11600,16005,22002,22002
11601,16015,22002,22002
11602,16025,22002,22002
11603,16035,22002,22002
11604,16045,22002,22002
11605,16055,22002,22002
11606,16065,22002,22002
11607,16075,22002,22002
11608,16085,22002,22002
11609,16095,22002,22002
11610,16105,22002,22002
11611,16115,22002,22002
11612,16125,22002,22002
11613,16135,22002,22002
11614,16145,22002,22002
11615,16155,22002,22002
11616,16165,22002,22002
11617,16175,22002,22002
11618,16185,22002,22002
11619,16195,22002,22002
11620,16205,22002,22002
11621,16215,22002,22002
11622,16225,22002,22002
11623,16235,22002,22002
11624,16245,22002,22002
11625,16255,22002,22002
11626,16265,22002,22002
11627,16275,22002,22002
11628,16285,22002,22002
11629,16295,22002,22002
11630,16305,22002,22002
11631,16315,22002,22002
11632,16325,22002,22002
11638,16385,22002,22002
11639,16395,22002,22002
11677,22002,22002,22002
11678,22002,22002,22002
11679,22002,22002,22002
11680,22002,22002,22002
11691,22002,22002,22002
11693,22002,22002,22002
11707,22002,22002,22002
11749,22002,22002,22002
3001,351,2,
10190,3512,22002,22002
10191,3522,22002,22002
10192,3522,22002,22002
10193,3522,22002,22002
3000,192,1,
10193,1902,1912,1922
10863,3552,22002,22002
10864,3562,22002,22002
10865,3572,22002,22002
10866,3582,22002,22002
10867,3592,22002,22002
10505,3662,22002,22002
10414,3912,22002,22002
10415,3922,22002,22002
10416,3932,22002,22002
10563,4072,22002,22002
10562,4082,22002,22002
10566,4102,22002,22002,
3001,,,
31597,10001,716,108
31598,10001,785,108
31599,10001,717,108
31600,10001,710,107
31601,10001,786,108
31602,10001,715,108
31603,10001,787,108
31604,10001,714,108
31605,10001,713,108
31606,10001,712,108
31607,10001,711,108
31608,10001,765,108
31609,10001,766,108
31617,10001,767,108
31618,10001,768,108
31619,10001,769,108
31620,10001,770,108
31621,10001,771,108
31622,10001,772,108
31623,10001,776,108
31624,10001,777,108
31625,10001,778,108
31626,10001,779,108
31627,10001,780,108
31628,10001,781,108
31629,10001,782,108
31630,10001,797,108
31631,10001,763,108
31632,10001,764,108
8500,,2,
1254,2301,1,-2301

Imagine being introduced to that code your first day on the job. In February of 1990, I was. A team of five programmers maintained hundreds of files of this code in production for years (we didn't invent it; we only had to work with it). It took about two weeks to become fluent in the language. As a historical curiosity, I present here the tool I made to cope with it, which duplicated as closely as possible the highlighter-pens-and-paper method that earlier programmers used to work on the code.

This was before HTML; the output device we had was a dot matrix printer, which could do condensed type, boldface, italics, and underlining (some of those attributes will require CSS to mimic on modern browsers).

Op Codes

A complete list of BANCStar opcodes has been found (at the end of this file).

BANCStar Source Code

The file C16LNAPP.SCN (warning: scanned PDF) is a comprehensive example of some production code used in First Interstate Bank of Washington around 1991. It is not used any more, and is of some historical interest, so with consent of the participants I have published it here.

As a side project at the bank, I wrote a sophisticated formatter dubbed LIST that translated C16LNAPP.SCN into this (warning: scanned PDF). LIST produced a view on the code tailored to the programmer's need, generating cross-references, statistics, and screen shots for documentation.

More BANCStar Source Code

MM1SM1.SCN ("Main Menu 1 Sub-Menu 1") was the top-level programme in the system. When I get the LIST utility running again, I'll post the translation here.

The only prompt file I have is the 'liabilities' one, LIAB.PFL.

The LIST Utility

LIST is the programme that translates screen code into something readable for humans.

The complete source code to LIST has been recovered (version 2.2b). It was written for a now-obsolete Microsoft C compiler on MS-DOS, so it doesn't compile now, but should be easy to port to any POSIX operating system. Presently, the source code looks for a few DOS-specific #include files like dos.h and conio.h, but I don't think it uses any obsolete memory model features like near and far pointers. Sorry the Makefile is a scanned PDF.

The 1991-vintage executable still sort of works on Microsoft Windows in 2014 (in a CMD prompt window).

LIST.EXE needs two data files to work: some screen code and a prompt file. Run it like this:

LIST Screen File Formatter Version 2.2a                             Revision 11
Copyright (C) 1990, 1991 Joe Loughry. All rights reserved.

Usage:list [/prompt][/sN][/eN][/a][/d][/f][/h][/n][/x][/z] filename.ext

The options, which may appear in any order, are as follows:

     /icl      Use the prompt file ICL.PFL
     /liab     Use the prompt file LIAB.PFL
     /other    Use the prompt file OTHER.PFL
     /rates    Use the prompt file RATES.PFL
     /pNAME    Use the prompt file specified by NAME
     /sN       Start listing at line N of the source file.
     /eN       Stop listing at line N of the source file.
     /a        Print listing only; do not print List Report.
     /d        Format output for the IBM 4019 Laser Printer and send to file.
     /f        Send output to the file C:\LISTFILE.RJL -- do not print.
     /h        Display this Help message.
     /n        Do not perform syntax checking.
     /x        Print extended Reference List report. [default]
     /z        Print List report only; do not print listing.

$ list -liab -f mm1sm1.scn

Porting Notes

To port LIST to a modern POSIX operating system, at least the following changes may be required. The source code is ANSI C and I believe it has no dependencies on quirks of the real-mode Intel IA32 memory model (e.g. NEAR and FAR pointers).

1. Update the command line option processing code to use POSIX syntax (-a) instead of MS-DOS syntax (/a).

2. In a few places, MS-DOS system calls are used, e.g., waiting for a keypress, or to clear the screen.

3. Update the "printer-specific" code (it's all well-marked and bracketed by #define commands in the source) to add support for boldface, italics, condensed print, and underlining (or some other collection of visually distinctive formatting) in the output so it can be displayed as HTML.

4. Recompile with the appropriate #define commands to support HTML output.

Wikipedia page

The BANCStar programming language has its own [Wikipedia page] (http://en.wikipedia.org/wiki/BANCStar_programming_language).

My 1997 article on the language

I wrote the following article in 1997. It was published by Brian Connors in the Turing Tarpit and updated in 2000. Further samples of BANCStar source code will be collected here as they are found.

I think the world is ready to be introduced to BANCStar.

The following actual sample of BANCStar source code was taken from a production system. I'm not too worried about revealing a lot of proprietary information here, as only about ten people in the world can read this code:

8607,,,1
11547,15475,22002,22002
1316,1629,1,1649 
3001,1316,3,30078
11528,22052,22002,22002
9301,0,1528,1528
31568,10001,800,107
8560,,,1568
8550,210,,
3001,,,
3100,1316,3,30089
11547,15475,22002,22002
3001,1316,3,30089
3001,1317,3,10000
8400,,,
8550,700,801,
3001,,,
9301,0,522,522
3000,1284,3,10001
8500,,3,
8500,,5,
1547,,1,-2301

I think we had something like 1,350 files of this code, most files a few hundred lines long, though some were over 1,000 lines. The system ran the retail branch operations of a major commercial bank in the early 1990s.

Interesting features of BANCStar included:

• The only legal characters in a BANCStar program are the digits 0 through 9, comma, minus sign, and carriage return. Blank lines are not allowed. If a "." appears anywhere in the file, the compiler will crash.

• Comments are strictly prohibited.

• Control structures available include the 3000 ("conditional"), 3001 ("block conditional"), and 3101 ("reverse block conditional"), as well as 8500 ("GOTO") and 8550 ("combination GOTO").

• There is a hard limit of 2000 variables + constants in the entire system. Anything to be displayed on screen or printed on a form must be defined as a constant, thereby taking up part of this valuable space. As a result, most calculations are done with a block of twenty or so "working storage" variables, which are continually reused. Everything in the system is global. New projects always started off with the programmer searching for a handful of working storage numbers that could be "borrowed" long enough to complete the calculation, then restored to their original values before the rightful owner noticed that they were gone.

• Labels within the code are absolute, so any time a new page was added or removed, all downstream GOTOs must be found and re-targeted.

A few more details I remember:

BANCStar actually came with a "screen generator" that was supposed to be used to construct applications. But the 5.1c version of the generator was so limited that experienced programmers soon began to pry off the covers and modify directly the intermediate code that the run-time module actually executed.

When I arrived in Seattle in early 1990, I joined a team of seven programmers who by now routinely wrote directly in BANCStar machine language, completely ignoring the screen generator. The first time they showed me the language, I thought it was a joke. But within a few weeks I was reading and writing the code, with the aid of every BANCStar programmer's favourite tools: a dot-matrix printer, lots and lots of different colour highlighters, and a three-ring binder called the Prompt File, stuffed with printouts of the dozens of tables in the system, and religiously updated anytime anyone changed anything of significance. (I wasn't kidding about reusing storage; if you needed a constant integer 1000, and you could find a place where somebody else had once used that same value, you linked your code to his and hoped it never changed.)

We developed some in-house tools for programming large applications in BANCStar, and at one point attempted to interest Broadway & Seymour in them, but I think they never really believed us, that we were writing directly to their internal, low-level, undocumented machine code.