PL/0 | Hacker News Reader

PL/0

en.wikipedia.org

73 pointstosh16 comments

To all interested in this little treasure, I transcribed the source code of the PL/0 compiler from the book "Algorithms + Data Structures = Programs", published by Nicklaus Wirth in 1976, and adapted it to run in Free Pascal.

You can find the sources here: https://github.com/dtoffe/adsp-pl0

Some interesting notes:

- The original compiler does not define nor implements "read" or "write" statements, as was the norm in later PL/0 implementations. - The source code in the book corresponds to the implementation in the CDC 6000 that Wirth had at hand back then. That machine used the CDC Display Code:

https://en.wikipedia.org/wiki/CDC_display_code

a 6 bit character code from before the ASCII times. Among its 64 characters, it includes single symbols for "<>", "<=" and ">=". In an attempt to change the code the least to make it run as original as possible, I changed those three character to #, { and }, as the symbols in the lexer are implemented using an array of char (only the assignment is treated as a special case).

- The interpreter printing out the values of every variable assignment, noted by another reader of this post, was a way of getting some information of the running program, since the compiler does not implement read or write statements.

- Following the compiler code, full of single letter variable names, was not the most exciting part.

The older I get the more I prefer Wirth syntax languages with keyworded blocks and := assignment operators, and regret how C block syntax and =/== took over.

I learned first on Pascal & Modula-2 and only picked up C later and while I appreciated its terse minimalism at the time and through the 90s, I actually don't at all now. I find it less readable.

I have always disliked the := as assignment operator convention. In these declarative languages, assignment is done frequently. There is little cognitive load to using '=' as assignment, although perhaps a bit jarring for math folk.

<- is somewhat better, but, again, for such a common operation, a single character is just more convenient. Sure, we could have editors that turn "=" into := or <- but now we're getting too fancy especially for something pedagogical.

I also don't mind the -> for C pointers; and certainly don't mind the <= >= or even == conventions (although at least today's compilers warn when they see "if (a=b) ...".

Ultimately, humans won't be writing code anymore anyway ( ;-) ?) so maybe the issue is entirely moot.

> I have always disliked the := as assignment operator convention. In these declarative languages, assignment is done frequently.

> I also don't mind the -> for C pointers

Mmm. These two opinions should be contradictory if held on principle as opposed being held out of impression.

    it = next(it);
    if ((*it)->node->op == EQ) ...

vs.

    it := next(it);
    if it.node.op = EQ ...

Eh. I don't really mind either of those except for the stupid parens after the "if" in the first case.

Technically, if you don't make assignment an expression, you can even get away with using "=" for both. And "->" exists only because structs originally weren't really typechecked; you could take any pointer and just do "->struct_field" at it, and the compiler would auto-cast.

Using '=' for both assignment and comparison is awkward when parsing incomplete code. Consider e.g.:

  j = 5;

The user starts writing (<|> is the cursor position):

  i = <|>
  j = 5;

This is a valid expression (i is a boolean). But the user probably intends to finish writing something like:

  i = 0;
  j = 5;

So in the intermediate state we would like to emit a single warning about an incomplete statement. But since it is valid as written, we instead end up warning about e.g. j being unbound.

Why were forward slashes so popular in computing product names in the 70s and 80s?

PL/0, PS/2, CP/M, etc.

It was a convention to denote a variation or version. Not sure how the trend started though.

Maybe referencing the reputation of IBM System/360?

I think it started with IBM: System/360 and /370, PS/2, OS/2, PL/I

And then Gary Kildall also seemed to like it with CP/M and PL/M, but those were after IBM had used it and I'd guess Gary was just copying IBM.

Between just those two influences you cover a huge portion of the mainframe and micro computer worlds during the 60s-80s

"The publisher of Wirth's books (Addison-Wesley) has decided to phase out all his books, but Wirth has published revised editions of his book beginning in 2004."

That is sad, but the revised editions seem to be published online.

Interesting. The article states "The compiler prints the value as a given variable changes." -- surely it means the program does, and not the compiler?

I take it to mean that the compiler inserts variable print code on variable modifications.

It might be more precisely stated something like "The language's semantics require that when a variable changes value, that change includes the side-effect of printing the new value."

The compiler produces p-code to be interpreted by an interpreter, so it is the interpreter that prints the value.

Any relation to PL/I?

I think the name was likely a tongue-in-cheek reference to PL/I being notoriously complex to implement a compiler for. Wirth designed a language for teaching that had an extremely small set of features (making its complexity much closer to "0" than "1"). It is basically a small subset of Pascal.

Reply on news.ycombinator.com