I too thought about using map() equivalent to comprehension and it seems to me a good reason and perhaps there are others in favor of this choice.
In the case of the DSL “niche” of the DataFrames context, I propose the introduction of a “plus function” with the symbol “+>” which has higher precedence than “=>” 
.
We would thus have that “+>” “wins” over “=>” which “wins” over “->”
After all “+” > “=” > “-”