ARexx "not not" bug

[broadblues]

Sorry guys, all prefix operators are broken:

Code: Select all

say --1
say ++1

If you take a look in the "bible":
"There may be one or more prefix operators to the left of any term."
https://books.google.de/books?hl=de&id= ... refix+plus

C'mon, our implementation is not a decision by design but simply incomplete.
It's slower, you have to parse extra parentheses.

But we don't have an implmentation of IBM Rexx we have an implemenation of ARexx, it's a seperate language, with different constructs, the time loss in dealing with brackets must be fairly trivial overall, especially when you will often end waiting for a command passed to a host via address to return.

What do you need 'not not' for anyway? It seems a bit redundant.

[xenic]

If you take a look in the "bible":

You're looking at the wrong "bible". The "bible" for ARexx is "ARexx Users Reference Manual 1.0" by William S. Hawes (the author of Amiga ARexx). It's a moot point anyway. There are no plans to ever change/update OS4 ARexx.

[nbache]

What sense does it make to invert an inverting operator? Instead of ~~0 you could write just 0. Would be even faster, no operator to parse at all.

Of course ~~0 doesn't give any benefits. But as salass00 pointed out earlier in the thread, ~~x could be used to "normalize" the value of x to a (boolean-like) 0 or 1, like it is commonly done in e.g. JavaScript.

Best regards,

Niels

[polluks]

But we don't have an implmentation of IBM Rexx we have an implemenation of ARexx, it's a seperate language, with different constructs, the time loss in dealing with brackets must be fairly trivial overall, especially when you will often end waiting for a command passed to a host via address to return.

However I have to write a bug report for Regina (http://regina-rexx.sourceforge.net) because its ARexx mode is not compatible.
The other way around seems to be more logical IMHO.

There are no plans to ever change/update OS4 ARexx.

Too late, see viewtopic.php?f=14&t=3309.

[broadblues]

But we don't have an implmentation of IBM Rexx we have an implemenation of ARexx, it's a seperate language, with different constructs, the time loss in dealing with brackets must be fairly trivial overall, especially when you will often end waiting for a command passed to a host via address to return.

However I have to write a bug report for Regina (http://regina-rexx.sourceforge.net) because its ARexx mode is not compatible.
The other way around seems to be more logical IMHO.

Well why does regina need bug report, if it supports both ~~0 and ~(~0) it's a superset of ARexx. No problem. How many rexx scripts actually run on both AmigaOS and AROS? Exact compatabilty would seem overkill, and if needed easily obtained by avoiding obscure edge cases in the syntax that stayed hidden for 30 years as you put it earlier


I would guess modify Arexx's parser would be non trivial.

[zzd10h]

"~~ also works in Rexx on IBM mainframe."

yes.

[polluks]

yes.

Thanks. Indeed, it's rather academic.
BTW Regina has "options AREXX_SEMANTICS".

[polluks]

Guess the result:

Code: Select all

say 10-1
say 10--1
say 10---1


[broadblues]

Guess the result:

Code: Select all

say 10-1
say 10--1
say 10---1


say 10+--1 is equally interesting