Differences between ETL 2nd printing and Eiffel Software implementation

  1. Added classes
  2. Added classes
  3. Added keywords
  4. Added semantics
  5. Added external support

See Also: See also: Differences between standard ECMA-367 and Eiffel Software implementation

"ETL 2nd printing" refers to the book "Eiffel: The Language" (2nd printing), published by Prentice Hall.

Added classes

Added classes

  • New basic classes have been added: INTEGER_8, INTEGER_16, INTEGER_64 and WIDE_CHARACTER. INTEGER is now specified as having a 32 bits representation
  • New TUPLE, ROUTINE, PROCEDURE and FUNCTION classes required by the agent mechanism.

Added keywords

  • Precursor
  • reference: new keyword to specify that a type is used as a reference type.
  • agent: new keyword used by the agent mechanism.
  • create: Instead of using the famous exclamation mark to create an instance of a class, you can use the keyword create. Below you will find a correspondence table between the old and the new syntaxes. The old syntax is still valid, but at some points Eiffel Software will remove it from its implementation:

Old syntax New syntax
a create a
a.make create a.make
B| a create {B} a
B| a.make create {B} a.make

  • note: replacement for the keyword indexing.
  • attribute: new keyword to declare attribute body.
  • attached: new keyword to specify attached types and object tests.
  • detachable: new keyword to specify detachable types.

Added semantics

  • Generic creation
  • Expression creation: you can now create an object within an expression. For example, you want to create an object and pass it as an argument to a function. Whereas you had to create a local variable, create the object and pass it to the function, you now simply need to pass to the function the creation expression. Here is a small example:

Old method New method
<code> local a: STRING do a.make (10) f (a) end </code> do f (create {STRING}.make (10)) end
This is also very useful since it can improve the power of assertions.

  • Mutually recursive constraints: one can now write class A [H, G->H] or class B [H -> C, G -> ARRAY [H]]. As a result, the declaration A [D, E] is valid only if E is a descendant of D. Similarly, the declaration B [E, ARRAY [D]] is not valid, if E is a descendant of D.
  • Tuples
  • Agents
  • Feature access:

local value: INTEGER do value := {MY_CLASS}.value end
The previous call is valid, if and only if:

    • value is a feature representing a constant of a basic type (INTEGER, DOUBLE or CHARACTER)
    • value is a C/C++/DLL external feature
    • value is an IL static external feature

Added external support

Look at the page for C and C++ with the introduction of struct and C++ external features encapsulation.