On: One in five genetics papers contains errors thanks...

[bikenaga]

Thanks for looking all this up, particularly the stuff about spreadsheets. I wonder what real early spreadsheets like Lotus 1-2-3 or even Visicalc would do with "-3^2". I have to feel one of the early spreadsheet programmers just made a mistake, because I can't think of a practical reason for going against the mathematical convention if they were already aware of it. Or maybe it comes from bc? ... does bc predate modern spreadsheets?

(Well, I just tried "=-3^2" in an org-mode table and it gives "-9".)

[shagie → bikenaga]

You've got to go even earlier... which is where the bc link becomes interesting. It's from 1975.

Lotus 1 2 3 dates from 1983... I can't find a copy of it that is runnable.

VisiCalc would be a good one to look at at 1979. It also presents 9 https://archive.org/details/VisiCalc_1979_SoftwareArts

You've also got sc https://en.wikipedia.org/wiki/Sc_(spreadsheet_calculator) from 1981.

    docker run -it ubuntu:latest
    # apt-get update
    # apt-get install sc
    # sc
    = -3^2

And you'll see 9.00 (screen shots of those two https://imgur.com/a/L0ZvJlP and the one from VisiCalc )

This is the way its worked for a long time.

---

(edit / further thoughts)

I believe that the underlying issue is that unary - (negation) and binary - (subtraction) use the same operator and you need the unary one to have a very high precedence to avoid other problems from happening.

Consider the expression: 2^-2

Is that 0.25 or a parse error?

[bonzini → shagie]

Unary and binary - can be given different precedence, because there is never ambiguity as to whether you're in front of a unary minus or a binary minus. A binary operator is never encountered at the beginning, or after a parenthesis, or after another binary operator, therefore those are the cases where you'd have a unary - or +.

Rather, the problem is whether -2 is parsed as a numeric literal, or a unary minus followed by a numeric literal (which would only include positive numbers).

[atq2119 → shagie]

Why would the expression 2^-2 pose a problem? There is no ambiguity there that requires operator precedence to resolve.