While investigating the convert_R_types function in animint.js, I identified three critical issues in how outliers are processed:

Data Loss: The current logic uses parseFloat(d[v_name].split(" @ ")). In JavaScript, calling parseFloat() on an array only parses the first element and discards the rest (e.g., "100 @ 200" becomes 100 instead of [100, 200]).

Shadowing Bug: There is a general r_type == "character" check before the specific v_name == "outliers" check. Since outliers are passed as characters, the specific parsing logic is never reached.

Logical Operator: The condition currently uses a bitwise & instead of a logical &&

I have a regression test ready.

Note: I used an AI assistant to help debug the R environment setup.

I would like to work upon this issue for convert_R_types function @tdhock !