Remove duplication of precedence in `mathics/core/parser/operators.py`

[rocky]

Mathics scannner is now getting close to complete in recording operator precedence. If information is lacking, it should be added to mathics-scanner. For example, we may need to add information about prefix-ops, postfix ops, left binary ops, etc.

This should be used instead of mathics/core/parser/operators.py and that removed or reduced.

Note that one gentle way to do this is to first start out with mathics-scanner, and then merge in information that is currently in this file. And this is also a way to tell whether things are fully done: if the set difference of what is in operators.py minus the scanner tables is not empty, we need to add more.

[TiagoCavalcante]

@rocky in mathics/core/parser/operators.py there's information about the type of the operator (e.g.: prefix, postfix, binary). How that information would be added to the YAML?

[rocky]

I can think of four possibilities:.

• add a new field for the operator kind, e.g operand-kind: prefix
• or change the single precedence field name into one of 3 or 4, e.g prefix-operator: 160, postfix-operator: 160
• change the precedence value to include the additional precedence: [160, "prefix"]
• expanding the above: change "precedence" to "operator" and include the name in the list: operator: ["Alternative, "binary", 160]

Of these the last one seems to lead to the least possibility of inconsistency. The first case allows for the possibility that an operator-kind could be specified while precedence isn't. And the second case allows for two mutually exclusive precedence names. The third case I suppose would have to be checked datatype: a tuple of two items with the first a number and the second a string of one of some categories.

In all of these possibilities, there could be default values to make it easier to convert the existing data.

[TiagoCavalcante]

There aren't many operators so it could be done all at once.

I like the 2nd possibility. We have, e.g. Increment and PreIncrement with the same symbol.

[rocky]

There aren't many operators so it could be done all at once.

I like the 2nd possibility. We have, e.g. Increment and PreIncrement with the same symbol.

Ok - implementor's choice!