Incorrect Order of Arithmetic Operations

[piotrbajdek]

Describe the bug
Incorrect Order of Arithmetic Operations

To reproduce
Steps to reproduce the behavior:

2 + 2 * 10
== 40
2 + 2 ** 10
== 1048576
2 * 2 ** 10
== 1048576

Expected behavior
Should be:

2 + 2 * 10
== 22
2 + 2 ** 10
== 1026
2 * 2 ** 10
== 2048

Platform version
RED: [ branch: "master" tag: #v0.6.4 ahead: 4629 date: 28-Jan-2023/11:28:13 commit: #ec0c31279023590e285fb2e1869380670ee0239f ]
PLATFORM: [ name: "Fedora Linux 37 (Container Image)" OS: 'Linux arch: 'x86_64 version: 6.1.7 build: {#1 SMP PREEMPT_DYNAMIC Wed Jan 18 17:11:49 UTC 2023} ]

[hiiamboris]

Quoting the R2 core manual which is also relevant for Red here:

The evaluation of expressions from left to right is independent of the type of operator that is used. For example:

print 1 + 2 * 3
9

And:

In REBOL, rather than requiring users to remember the precedence of operators, you only need to remember the left-to-right rule. More importantly, for advanced code such as expressions that handle expressions (in reflection, for example) you do not need to reorder terms based on precedence. The evaluation order is kept simple.

My personal view is this design just makes the interpreter much faster and simpler, as this way it operates on the data directly, not requiring an AST.

When you want the commonly used order you can use math function:

>> math [2 + 2 * 10]
== 22
>> math [2 + 2 ** 10]
== 1026
>> math [2 * 2 ** 10]
== 2048

[greggirwin]

No offense taken. Many things in Red seems strange to newcomers. There are always design tradeoffs but, having used Redbol langs for 20+ years, I find other languages terribly limiting and archaic most of the time. It's always fair to ask "Why does it work that way?" and some of the answers to what people think are bugs often bend their minds. :^) Ultimately, Red is a great language construction kit, as math shows how to build a tiny DSL for that domain.

[piotrbajdek]

There are always design tradeoffs but, having used Redbol langs for 20+ years, I find other languages terribly limiting and archaic most of the time.

But you're still on 32 bits! Just joking! 😜

[hiiamboris]

Yes, 32 bitness, and for serious math a lot is left to be desired. Also alpha stage of development. These are fair points of critique at the moment.

[piotrbajdek]

Is the Discussions section the appropriate place to ask questions about Red, or would you prefer that I inquire elsewhere?

[hiiamboris]

Ideally, there ☻

[dockimbel]

@piotrbajdek

A few comments:

• The cognitive overload of dozen(s) of precedence rules with sometimes exception(s) is immense compared to the simple left-to-right rule. It takes little time to get used to it as it is becomes quickly intuitive. I think that no Red/Rebol coder would want to come back the operators precedence-hell. Fortran also has its own unintuitive exceptions for operators precedence.
• I would argue that it is easier to make mistakes when using precedence-base rules than the single-rule approach we use.
• There are mainstream languages like Forth or Postscript that do not use precedence rules for operators but rather postfix notation. Math teachers, scientists and financial analysts since the 70's often rely on postfix-notation calculators like the HP 11C or 12C, which are both faster and easier to use than classic ones requiring parentheses to mark precedence.