Self-description / metadescriptions #78
Conversation
Assargadon
requested review from
konstantin-kalinin-wizn-tech and
passimens
Assargadon
force-pushed
the
self-description
branch
from
364b744
to
29c4055
Compare
Magritte/MAContainerTest.py
Outdated
| def test_searchByName(self): | ||
| desc = MAContainer() | ||
| desc += MAStringDescription(name = "first", label = "First Field") | ||
| desc += MAStringDescription(label = "nameless #2") |
There was a problem hiding this comment.
Кавычки то одинарные, то двойные. Питону всё равно, но рекомендуется использовать одинарные "в простых случаях" - они "легче" выглядят при отсмотре кода.
There was a problem hiding this comment.
Не знал.
Но что делать?
Как найти все места с кавычками?
There was a problem hiding this comment.
поиском, например,
="
= "
("
,"
, "
There was a problem hiding this comment.
так найдётся большая часть
| @classmethod | ||
| def isAbstract(cls): | ||
| return True | ||
|
|
||
| @property | ||
| def type(self): |
There was a problem hiding this comment.
Зачем это? Нигде не используется.
There was a problem hiding this comment.
Используется при сериализации метадескриптора.
То есть представь, у нас есть какой-нибудь Customer, а у него поле last_name.
Как из дескриптора поля понять, что это строка?
Вот по полю type
There was a problem hiding this comment.
Почему не по kind?
There was a problem hiding this comment.
потому что kind возвращает класс - он нужен, но его и описать и сериализовать сложно
Magritte/MADescription_selfdesc.py
Outdated
| def magritteDescription(self): | ||
| desc = MAPriorityContainer() | ||
|
|
||
| desc += MAStringDescription(label = "Kind", priority = 0, readOnly = True, accessor = MAAttrAccessor('type')) |
There was a problem hiding this comment.
Почему не attrAccessor('kind') ?
Magritte/MADescription_selfdesc.py
Outdated
|
|
||
| desc += MAStringDescription(label = "Kind", priority = 0, readOnly = True, accessor = MAAttrAccessor('type')) | ||
| desc += MAStringDescription(label = "Field name", priority = 1, accessor = MAAttrAccessor('name'), comment = "Non-human-readable name, something like 'field name' or `json key`, etc.") | ||
| desc += MAStringDescription(label = "Label", priority = 100, default = "#nolabel#", required = True, accessor = MAAttrAccessor('label')) |
There was a problem hiding this comment.
Почему не defaultLabel() ?
There was a problem hiding this comment.
Остатки ранних усилий, кажется. Посмотрю.
There was a problem hiding this comment.
- разобраться с "#nolabel#"
There was a problem hiding this comment.
Вынес задачу в отдельный тикет #81
Там целая подсистема генерирования label на основе дескриптора, мне не хочется делать её в рамках этого ПРа.
|
|
||
| object_encoder = MAObjectJsonWriter(description.reference) | ||
| collection = description.accessor.read(self._model) # TODO: replace on model.readUsing or description.read (both are not implemented yet) |
There was a problem hiding this comment.
Уже implemented
There was a problem hiding this comment.
- реализовать правильное чтение, без прямого доступа к аксессору
Looks like if you overwrite getter of the @Property, you need also overvrite setter of it - even if it not changed. If you overvrite getter only, setter becomes inaccessible.
refactoring is needed to exclude copy-pasted name-related piece of code
othervise it's pretty hard to determine which exactly descriptor has no label and therefore can't be converted to json
It's non-visible because there is, well, no good way to show it - it's an object. Why #accessor has reference of MAStringDescription I don't know - but it's how it in original I tested it by printing out the value of it, directly by `if description.name == "accessor"` - and result was OK
works with non-json values as accessors
and completely moved into MAElementDescriptor Why they have this pseudo-accessors and why description and default value was in the MADescription - I don't know. Have no good reasons for this.
…ReferenceDescription it has lasy-initialization anyway, and current implementation overrides whatever custom reference we'll set however, laizy initialization fixed, too, to not re-create defaultReference each time we access it
better support of references
additional thoughts are needed to manage them properly
previous approach was needed to overcome reference "init overwriting" bug, which was fixes
this way `children` _may_ be set in constructor's parameters smalltalk original has no constructors or factory methods for descriptions, they use chain calls of setters instead. But we use constructors, hence the modification.
Except of visual-only test we performs seferal self-health checks At least we know every description has a description and that every described field is accessible
Assargadon
force-pushed
the
self-description
branch
from
29c4055
to
a8c1110
Compare
This reverts commit 860129b.
Just in case `magritteDescription` exists but is not callable...or something
Добавляет дескрипторы для дескрипторов
Даже тесты какие-никакие есть
NB: некоторые сильно особенные поля (типа
referenceилиclasses) я помечаю (комментариями) как# === special ===и делаю их типаMAElementDescription. Типа, обозначить существование полей, беспечить доступ стандартным образом и т.п.