tablerow: Avoid accidental special case for constant nil cols

[dylanahsmith]

Problem

#1633 recently introduced a weird special case behaviour for a nil literal for the tablerow columns, which differs from an expression that evaluates to nil. This is because @attributes['cols'].nil? would match a nil literal expression, but wouldn't match an expression that dynamically evaluates to nil.

The included regression test demonstrates the problem, where {% tablerow i in (1..2) cols:nil %}{{ tablerowloop.col_last }}{% endtablerow %} doesn't produces the same output as {% tablerow i in (1..2) cols:var %}{{ tablerowloop.col_last }}{% endtablerow %} when var is nil.

Solution

It seems like the intended behaviour is to default 'cols' to the length of the slice being iterated over, as @attributes.key? does for the 'offset' and 'limit' attributes. So I updated the expression to default 'cols' to also use @attributes.key?.

[ggmichaelgo]

eeek! Thank you for catching this and fixing this @dylanahsmith !