Backwards incompatible changes for "is_empty" in 1.6+

[GUI]

The changelog notes a fix for is_empty behavior for tables with boolean keys in 1.6.0, but the changes in 1.6+ appear to be more extensive than that. In 1.5.4, is_empty returned false for many other object types that now return true in 1.6-1.7. Examples:

• Penlight 1.5.4

  -- Numbers
  is_empty(0) -- false
  is_empty(20) -- false
  -- Booleans
  is_empty(false) -- false
  is_empty(true) -- false
  -- Functions
  is_empty(print) -- false
  -- Userdata
  is_empty(newproxy()) -- false

• Penlight 1.6.0-1.7.0

  -- Numbers
  is_empty(0) -- true
  is_empty(20) -- true
  -- Booleans
  is_empty(false) -- true
  is_empty(true) -- true
  -- Functions
  is_empty(print) -- true
  -- Userdata
  is_empty(newproxy()) -- true

This appears related to the changes made in #267, which defaults is_empty to true for all unchecked cases, rather than false. A few additional tests were added in #279 which indicate the expected value for is_empty for at least numbers and booleans is true, but I wanted to confirm if these changes were actually intentional.

I personally find the is_empty behavior in 1.5 to more align with my expectations, since I wouldn't necessarily anticipate numbers, booleans, functions, or userdata to be considered "empty." But if this change is intentional and is_empty is really only intended to be used to check strings, tables, and nil types, then should this change in behavior maybe be called out more explicitly in the CHANGELOG, since this change in behavior could have more impact than expected?

[Tieske]

thx for the report, the new behaviour also complies with the documentation.