On Mon, 31 Aug 2020 17:01:31 +0100, druck wrote:

> In any case I was born in 1968, and started programming on the BBC Micro
> in the early 80s, so luckily I avoided Cobol then and ever since.
>
COBOL wasn't actively evil (apart from the ALTER statement - and that had
vanished by the mid-80s. It just far too verbose:

  ADD A to B GIVING C ON SIZE ERROR PERFORM OVERFLOW-TRAP.

where the Java equivalent would be something like:

  try {
     c = a + b;
  }
  catch (ArithmeticException e) {
     overflowTrap(e);
  }

There's also a lot of typing involved because of the size ot variable
names and the number of them you have to define:

- fields in structures have names that aren't linked in any way to the
  structure's name.

- if a program reads a value from an input file, writes it to a database
  file and prints it in a report or displays it on screen, you'll end
  defining the variable three times and the input and display forms will
  be slightly different because they define external format.

That said, COBOL did at least always let you catch ON SIZE ERROR overflow
exceptions, which is more than could be said for other contemporary HLLS
such FORTRAN. Specifying formats for input and output fields was also
always pretty straight-forward and intuitive.

However it had some terrible features such as the ALTER verb, which
allowed a running COBOL program to self modify, making debugging hellish.

Early COBOLs would allow you to call assembler subroutines but the
language did not allow external subroutines to be written in COBOL. I
don't remember seeing this feature in any implementation before 1978.

Early COBOLs also required a program to be a single source file though it
had a COPY statement that could be used to pull in record definitions or
sections of code. Fortunately, the compilers could handle huge source
files: a program of less then 200 lines usually did nothing useful.
Programs were generally in the 1000-4000 line range though I have seen a
few in the 10,000 line range.

There was a built-in Y2K gotcha too: The only way you could get hold of
the date was with a statement like

 ACCEPT DATE-TODAY FROM SYSTEM-DATE.

where the DATE-TODAY variable *had* to be declared in working storage as:

  01 DATE-TODAY      PIC 9(6).

and the ACCEPT statement would fill it with a date formatted as 'yymmdd'.
This requirement was defined in the CODASYL Report (i.e. was part of the
language standard) and AFAIK it was not changed until after Y2K had been
and gone.


--
Martin    | martin at
Gregorie  | gregorie dot org

--- SoupGate-Win32 v1.05