BCPL octal numerical constants

Question

I've been digging into the history of BCPL due to a question I was asked about the reasoning behind using the prefix "0x" for the representation hexadecimal numbers.

In my search I stumbled upon a really good explanation of the history behind this token. (Why are hexadecimal numbers prefixed with 0x?)

From this post, however, another questions sparked:

For octal constants, did BCPL use 8 <digit> (As per specs: http://cm.bell-labs.com/cm/cs/who/dmr/bcpl.pdf) or did it use #<digit> (As per http://rabbit.eng.miami.edu/info/bcpl_reference_manual.pdf) or were both of these syntaxes valid in different implementations of the language?

I've also been able to find a second answer here that used the # syntax which further intrigued me in the subject. (Why are leading zeroes used to represent octal numbers?)

Any historical insights are greatly appreciated.

Answer 1

There were many slight variations on syntax in BCPL.

For example, while the one we used had 16-bit cells (so that x!y gave you the 16-bit word from a word address at x + y (a word address being half of the byte address), we also had a need to extract from byte address and byte values (since we were primarily creating OS and control software on a 6809 byte-addressable CPU).

Hence in addition to:

x!y - get word from byte address (x + y) * 2

we also had

x!%y - get byte from byte address (x * 2) + y
x%!y - get word from byte address x + (y * 2)
x%%y - get byte from byte address x + y

I'm pretty certain they were implementation-specific as I never saw them anywhere else. And BCPL was around long before language standards were as important as they are today.

The canonical language specification would have been the earlier one from Richards since he wrote the language (and your second document is for the Essex BCPL implementation about a decade later). But keep in mind that Project MAC was the earliest iteration - there were plenty of advancements after that as well.

For example, there's a 2013 revision of the BCPL User Guide (see Martin's home page) which specifies #b, #o and #x as prefixes for various non-decimal bases.