SweepFormula: Support minus as sign in the parser (and therefore don't require parantheses)

[t-b]

•print FormulaExecutor(FormulaParser("-min([0, 1])"))[0]
  -1
•print FormulaExecutor(FormulaParser("+min([0, -1])"))[0]
  0

that is not what you expect. Using 58f24b8a (UTF_SweepFormula: Use module definition and make the functions static, 2019-09-21).

[ukos-git]

Remember that -1 is not numeric but an object

[ukos-git]

Why don‘t you just use (-1)*min(0,1)

[ukos-git]

the actual problem that you describe is simpler than your example.

min(0,-1) currently gives

{
  "min": [
    0,
    1
  ]
}

while

min(0,(-1)) gives

{
  "min": [
    0,
    {
      "-": [
        1
      ]
    }
  ]
}

and your formulas are solvable by parentheses

+min(0,(-1))

{
  "+": [
    {
      "min": [
        0,
        {
          "-": [
            1
          ]
        }
      ]
    }
  ]
}

The simplest solution is encapsulating (-[0-9]) and (+[0-9]) by parenthesis in the preparser but that would interfere with operation evaluation.

A similar approach as for the parentheses evaluation would also be possible.

[ukos-git]

what do you suggest for solving the -1 problems?

[ukos-git]

Personally, I recommend to always encapsulate -1 in brackets.

[ukos-git]

currently failing evaluations:

1+-1

{
  "+": [
    1,
    1
  ]
}

-1*2

{
  "-": [
    {
      "*": [
        1,
        2
      ]
    }
  ]
}

2+-1*3

{
  "+": [
    2,
    {
      "*": [
        1,
        3
      ]
    }
  ]
}

non-first array elements:

[1, 0, -1]

[
  1,
  0,
  1
]

-1 - 1

{
  "-": [
    1,
    1
  ]
}

• [ ] restore to the last calculation if - did not belong to an operation but was evaluated as a numeric prefix

[t-b]

[0, 1, 2] [3, 4, 5]

evaluates to

[
  0,
  1,
  2,
  [
    3,
    4,
    5
  ]
]

which is a bit suprising.

[ukos-git]

That is a bug in the parser. Should we track all issues of the parser here?

It should be of no particular relevance to our use scenario but definitively needs a fix.

[ukos-git]

I think, such checks for invalid syntax should go to the preparser. But we could probably also catch it in the parser.

The check for the above example would be that GrepString(str, "\]\s*\[") is invalid

[t-b]

@ukos-git Collecting them here sounds good!