Adding support to operators like x[y] in infixNotation - should I send a PR?

[franji-pagaya]

Hi ,
I wanted to add x[y] operator to the SQL dialect I am parsing using pyparsing.
I have added a small fix to pyparsing infixNotation which allows left-assoc binary operator with opExpr tuple - see example below.
The chage is 5 lines long - is this a usefull addition? should I send a PR?

UNARY, BINARY, TERNARY = 1, 2, 3
expr << pp.infixNotation(
    expr_term,
    [
        (pp.oneOf("- + ~") | NOT, UNARY, pp.opAssoc.RIGHT),
        (ISNULL | NOTNULL | NOT_NULL, UNARY, pp.opAssoc.LEFT),
        ((LBRACKET, RBRACKET), BINARY, pp.opAssoc.LEFT),   ### new operator.
        (pp.oneOf("* / %"), BINARY, pp.opAssoc.LEFT),
        (pp.oneOf("+ -"), BINARY, pp.opAssoc.LEFT),
        (pp.oneOf("<< >> & |"), BINARY, pp.opAssoc.LEFT),
        (pp.oneOf("< <= > >="), BINARY, pp.opAssoc.LEFT),
        (
            pp.oneOf("= == != <> <=>")
            | IS
            | IN
            | LIKE
            | RLIKE
            | NOT_IN
            | NOT_LIKE
            | NOT_RLIKE,
            BINARY,
            pp.opAssoc.LEFT, AthenaLikeOpNode
        ),
        ((BETWEEN | NOT_BETWEEN, AND), TERNARY, pp.opAssoc.LEFT),
        (
            (IN | NOT_IN) + LPAR + pp.delimitedList(expr).setParseAction(ParamsNode) + RPAR,
            UNARY,
            pp.opAssoc.LEFT,
        ),
        (AND, BINARY, pp.opAssoc.LEFT),
        (OR, BINARY, pp.opAssoc.LEFT),
    ],
)

[ptmcg]

Sorry for taking so long to respond.

In the absence of this feature, I've had to push dereferencing of []'s up into the operand itself. In your case, this would have the effect of pushing the []'s "operator" to the very top of the list.

I'm interested in this idea, but not sure it completely works. Is an expr_term always a valid body of []'s? If it is just for numeric indexing into a sequence, you might define a pseudo-operator that only takes an arith expression in the []'s, and then add it as a unary left-assoc operator.

I'll dabble with this and see what I get. Thanks for the interesting idea!

[ptmcg]

How does this look to you, if you didn't have such a feature?

import pyparsing as pp

ident = pp.common.identifier()
number = pp.common.number()
LBRACK, RBRACK = map(pp.Suppress, "[]")

operand = ident | number
multop = pp.Char("*/%")
addop = pp.Char("+-")
signop = pp.Char("+-")

arith_expr = pp.Forward()
array_index = LBRACK + arith_expr + RBRACK

arith_expr <<= pp.infix_notation(
    operand,
    [
        (array_index, 1, pp.opAssoc.LEFT),
        (signop, 1, pp.opAssoc.RIGHT),
        (multop, 2, pp.opAssoc.LEFT),
        (addop, 2, pp.opAssoc.LEFT),
    ]
)

arith_expr.run_tests("""\
    2+2
    x/40
    x[10]/y
    """)

[ptmcg]

I'm thinking about this a bit more, and I think this would be a nice addition. Please submit a PR with your change.