Difference between revisions of "ECMA"
Jump to navigation
Jump to search
Hzwakenberg (talk | contribs) |
Hzwakenberg (talk | contribs) m |
||
(11 intermediate revisions by 3 users not shown) | |||
Line 1: | Line 1: | ||
− | In [http://www.ecma-international.org/publications/files/ECMA-ST/ECMA-367.pdf ECMA-367] the ECMA committee TC39-TG4 defines an Eiffel standard. Liberty is not committed to fully implement this standard, but |
+ | In [http://www.ecma-international.org/publications/files/ECMA-ST/ECMA-367.pdf ECMA-367] the ECMA committee TC39-TG4 defines an Eiffel standard. Liberty is not committed to fully implement this standard, but it already supports more of the ECMA-improvements than its predecessor did at the time. Liberty will implement those parts of ECMA that match the effective, efficient and simple design of previous versions of Eiffel. The parts requiring an extensive run-time model will not be currently implemented; those that needlessly complicate the language to comply to widespread programming conventions will be evaluated case-by-case. |
− | == |
+ | ==Liberty Eiffel offers these ECMA features already:== |
+ | * '''attribute''' keyword may be used to define an attribute including self-initialization with contracts checking. |
||
With small deviations: |
With small deviations: |
||
* Non-conforming inheritance (but using '''insert''' instead of '''inherit {NONE}'''); see [[Typing policy]] |
* Non-conforming inheritance (but using '''insert''' instead of '''inherit {NONE}'''); see [[Typing policy]] |
||
* [https://github.com/LibertyEiffel/Liberty/issues/78 Assigners] with the following deviations: |
* [https://github.com/LibertyEiffel/Liberty/issues/78 Assigners] with the following deviations: |
||
− | ** VFAC is not enforced (let the standard rules of the replacing procedure call play instead) |
+ | ** VFAC (ECMA §8.5.22) is not enforced (let the standard rules of the replacing procedure call play instead) |
− | ** assigners are inherited with proper renames (ECMA does not explicitly specify the rules |
+ | ** assigners are inherited with proper renames (ECMA does not explicitly specify the rules either way) |
* [[task:12666|Cosmetic syntax changes]]: |
* [[task:12666|Cosmetic syntax changes]]: |
||
** '''is''' is now optional |
** '''is''' is now optional |
||
Line 17: | Line 18: | ||
==These ISE features are also implemented:== |
==These ISE features are also implemented:== |
||
− | [[task:13103| |
+ | [[task:13103|These features]] are not (yet) in the ECMA standard: |
* the newer '''alias "()"''' |
* the newer '''alias "()"''' |
||
* its companion implicit tuples |
* its companion implicit tuples |
||
Line 39: | Line 40: | ||
==Not yet decided:== |
==Not yet decided:== |
||
* <code><nowiki>TYPE[G]</nowiki></code> and explicit conversion |
* <code><nowiki>TYPE[G]</nowiki></code> and explicit conversion |
||
+ | * postcondition of is_equal: same_type: Result implies same_type(other) |
||
+ | ** this essentially forbids any equal objects with different types and is inconsistent to the current implementation in Liberty. It has some nice background (e. g. symmetry is not easy to guarantee without that), but we also do not want to brake existing code |
||
See also [[Compatibility]]. |
See also [[Compatibility]]. |
Latest revision as of 19:54, 22 July 2022
In ECMA-367 the ECMA committee TC39-TG4 defines an Eiffel standard. Liberty is not committed to fully implement this standard, but it already supports more of the ECMA-improvements than its predecessor did at the time. Liberty will implement those parts of ECMA that match the effective, efficient and simple design of previous versions of Eiffel. The parts requiring an extensive run-time model will not be currently implemented; those that needlessly complicate the language to comply to widespread programming conventions will be evaluated case-by-case.
Liberty Eiffel offers these ECMA features already:
- attribute keyword may be used to define an attribute including self-initialization with contracts checking.
With small deviations:
- Non-conforming inheritance (but using insert instead of inherit {NONE}); see Typing policy
- Assigners with the following deviations:
- VFAC (ECMA §8.5.22) is not enforced (let the standard rules of the replacing procedure call play instead)
- assigners are inherited with proper renames (ECMA does not explicitly specify the rules either way)
- Cosmetic syntax changes:
- is is now optional
- alias is implemented, including alias "[]"
- note replaces indexing and can be placed at the start and end of a class, and at the start of a feature
- create replaces creation
- generic creation
These ISE features are also implemented:
These features are not (yet) in the ECMA standard:
- the newer alias "()"
- its companion implicit tuples
- if-then-else expressions
These features are planned:
- generic inheritance
- conversions - maybe with a slightly stricter interpretation
- named TUPLE elements
- void-safety (Liberty implementation will probably differ from ECMA)
- attached vs detached
- across
These ECMA features are not planned in Liberty:
- No-variant agent arguments.
These features are not in ECMA but implemented in Liberty:
- inline agents are closures
- explicit agent conformance rules (see this JOT paper)
Not yet decided:
TYPE[G]
and explicit conversion- postcondition of is_equal: same_type: Result implies same_type(other)
- this essentially forbids any equal objects with different types and is inconsistent to the current implementation in Liberty. It has some nice background (e. g. symmetry is not easy to guarantee without that), but we also do not want to brake existing code
See also Compatibility.