django.utils.functional.lazy issues

[Alexey Boriskin]

I've found some issues meditating on the lazy() function. I'm reporting them as one ticket; please tell if I should rather create one per issue.

• promise, returned by lazy(), has no attributes of original class: lazy() thinks only about methods
• promise, returned by lazy(), has weird behaviour if resultclass has __mod__ method – it overrides it completely
• methods of promise, returned by lazy(), lack the __doc__ attribute – docstring is lost
• if resultclass explicitly define itself as a not hashable with __hash__ = None, promises of that class may still be hashable.

Following diff, when applied, adds four failing tests illustrating that issues: ​https://gist.github.com/4582334

[Anssi Kääriäinen]

My understanding is that lazy() will just defer the call of the given callable until it is accessed. The idea is to not work as a proxy for an object. So, for that reason marking as wontfix.

If some of the issues are seen as problems in "defer call to method" definition of lazy (the __mod__ issue?), then reopen...

[Aymeric Augustin]

I'd like to take a look at the __mod__ issue.

[Alexey Boriskin]

Patch, which can be used to solve issues in the ticket: ​https://gist.github.com/4605065

[Aymeric Augustin]

The second item was fixed in the patch for #19160. The three others are fairly minor and don't have actual consequences in the current code base.

The patch fixes a few things but it isn't quite production-ready (builtin_method_names = ['__hash__', '__str__'] - ouch)

For these reasons I don't think it's useful to keep this ticket open.

---

However, as discussed on IRC, it could be a good idea to re-implement lazy() by overriding __getattribute__ rather than by copying methods around, both design-wise and performance-wise (to be benchmarked). Feel free to write a patch for this and open a new ticket!