Re: Clair Grant on VMS code base

On 4/4/2025 2:00 PM, Simon Clubley wrote:

On 2025-04-03, Arne Vajhøj <arne@vajhoej.dk> wrote:

On 4/3/2025 2:15 PM, Simon Clubley wrote:

On 2025-04-01, Arne Vajhøj <arne@vajhoej.dk> wrote:

He looks at size 6.2 -> 7.2 -> 8.2 -> 9.2 and .mar vs .c 6.2 -> 9.2.
>
https://vmssoftware.com/resources/blog/2025-03-31-how-big-is-vms/

>
How the hell does DCL have 30,000+ lines of code ? :-)
(Even allowing for the fact most of it is in Macro-32)

>
Just for fun I downloaded bash and counted. 117 KLOC.
>

 Wow. I had no idea it was now that big. OTOH, bash has got a _lot_ more
functionality in it than DCL does.

6 times more?

Also, regarding those BLISS/Macro-32 figures across the various modules
in general, I wonder how much smaller the code would have been if it had
been written in C ?

>
C is a relative high LOC/FP language.
>
Most LOC/FP tables have C in the 100-125 range and macro-assembler
in the 200-250 range.
>
If we adjust for VAX instructions being more powerful/complex
than average ISA, add a random number and subtract wind speed I
will estimate the Macro-32 / C ratio to be around 1.5-1.75.
>
If you buy that then the DCL 33 KLOC would be 19-22 KLOC in C
to be compared with bash 117 KLOC.
>

 The problem with that analysis is what DCL does.
 It basically parses, validates, and executes commands it has been given.
That is something which can be implemented a lot more easily and concisely
in a HLL with abstracted data structure capabilities (which includes
even C) than an assembly language with no such capabilities.

It is not obvious to me that:
(LOC/FP for Macro-32) / (LOC/FP for C)
is a lot higher for a shell than for the average application - data
structures are not anything special for shells.
But maybe.
Arne