Call it madness, but I have a fascination with trying to transcribe my C++ accounts program to COBOL.
I have an idea of a kind of interoperative facility, where I can code in a mixture of both. As a first stage, I have written a small program in Lisp which contains a record schema and an input converter. Here is the schema:
(defparameter *amount* ‘(amount dec 12 2))
(defparameter *dstamp* ‘(dstamp str 10))
(defparameter *quantity* ‘(quantity dec 12 3))
(defparameter *ticker* ‘(ticker str 6))
(defparameter *yahoo* `(,*dstamp*
(tstamp str 8)
,*ticker*
(rox dec 6 4)
(price dec 12 6)
,*amount*))
(defparameter *record-descriptor*
`(( “comm-1” (,*ticker*
(dload str 1)
(typ str 4)
(unit str 3)
(name str 22)))
(“etran-1” (,*dstamp*
(folio str 4)
,*ticker*
,*quantity*
(amount dec 12 2)
(way str 1)
(taxable str 1)
(desc str 20)))
(“leak-1” (,*dstamp*
(folio str 4)
,*ticker*
,*quantity*
(taxable str 1)
(desc str 20)))
(“nacc” ((dr-acc str 4)
(cr-acc str 4)
(acc-type str 1)
(mult dec 1 0)
(desc str 20)))
(“nb” ((desc str 20)))
(“ntran” (,*dstamp*
(dr-acc str 4)
(cr-acc str 4)
,*amount*
(desc str 20)))
(“yahoo” ,*yahoo*)
(“yahoo-1” (,@*yahoo*
(desc str 20)))
))
I have written some Lisp functions that takes in free-form input, dispatches on the type of record, and produces a fixed string line output. The next step would be to see if I could generate COBOL copybooks and C++ structures from the schema.
On the COBOL side, this should make reading in the data fairly straightforward, as it can transfer a line from input directly into the copybook structure.
I think C++ would be trickier. It doesn’t have fixed-length strings or structs that I can simply serialise and deserialise.
On the other hand, if I want to cut Lisp out of the loop, it will be more difficult to persuade COBOL to convert raw input into record format. I had solved the problem a long time ago on C++.
My COBOL is virtually non-existent, so there’s certainly a learning curve in me massaging it into behaving the way I want it to.
My Lisp is fairly weak, but to a large extent, you don’t have to know much Lisp. I found myself creating the requisite macros quickly. The two points that stood out to me is that I had to spend a long time on the formatting routines. The format statement is powerful, but somehow it did not seem powerful enough. I had to wrap it in a few convenience functions to obtain the output I wanted.
What I also noticed is that Lisp (I am using clisp) is a slow old beast. It took nearly 0.5 seconds (clock time) to process the input. My guess – and it’s a guess shaped from an anlogous experience – is that C++ would be able to process this in a tenth of the time probably at most, and I imagine COBOL would be somewhere in that region, too. The notion, as some Lispers would claim, that Lisp could be “just as fast as C++” seems fanciful.
COBOL is a strange language. It is, at once, an easier, and more difficult, language to use.
The experiment continues.
