Operator sections

[kazcw]

The Enso Spec defines rules for the bounds of an operator section (see last paragraph of section)--specifically, when an expression containing an elided operand is the first child of its parent expression, and the parent is not a parenthesis node, the parent expression is included in the operator section function.

The parser implements this rule, creating an OprSectionBoundary node in the appropriate position. However, the compiler has never been updated from the pre-spec behaviour, and currently ignores the OprSectionBoundary.

We should fix this inconsistency. If the spec behaviour is desired, I think switching to it would be straightforward to implement, but it would be a semantics-changing backward-incompatible change, and so should be done sooner rather than later. Alternatively if we would like to keep the current semantics, we can make the AST consistent with the behaviour by eliminating OprSectionBoundary nodes, and all the parser logic that computes their locations.

[kazcw]

If it would be helpful, I can use our syntax tooling to survey the locations in our current .enso sources that are dependent on the current behaviour or could potentially be simplified with the spec behaviour (which allows more complex expressions to be built from operator sections).

[JaroslavTulach]

the compiler ... ignores the OprSectionBoundary.

Is there an erroneous snippet showing the error?