list of validator

[reece]

From the new docs:
"A validator that raises a TypeError if the initializer is called with a non-list type or with a list that contains one or more elements of a wrong type. None values are not permitted. As with instance_of, checks are perfomed using isinstance() therefore it’s also valid to pass a tuple of types."

[codecov]

Codecov Report

Merging #158 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master   #158   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files           8      8           
  Lines         551    563   +12     
  Branches      122    125    +3     
=====================================
+ Hits          551    563   +12

Impacted Files	Coverage Δ
src/attr/validators.py	100% <100%> (ø)	✅

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f232cc8...764eed9. Read the comment docs.

[codecov]

Codecov Report

Merging #158 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master   #158   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files           8      8           
  Lines         551    563   +12     
  Branches      122    125    +3     
=====================================
+ Hits          551    563   +12

Impacted Files	Coverage Δ
src/attr/validators.py	100% <100%> (ø)	✅

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f232cc8...764eed9. Read the comment docs.

[hynek]

OK so let’s do this.

But if we tackle this, I would suggest: container_of(container_type, element_type, empty_ok=True) with a bunch of convenience helpers list_of(), set_of(), tuple_of()…

[reece]

empty_ok=True means that the container can be empty, right?

What about also adding none_ok, permitting None values? Or could optional() be used for elements?

What about mappings? For example: mapping_of(container_type, key_type, value_type, empty_ok=True, none_ok=True)

[hynek]

empty_ok=True means that the container can be empty, right?

yep!

What about also adding none_ok, permitting None values? Or could optional() be used for elements?

I’d say that exactly what optional() is for. :) So yes, you’ll have to handle that case.

That reminds me: unless someone has good arguments, I don’t think we should do element_type but validator! And validator needs also to take a list. :)

What about mappings?

Baby steps. :)

[reece]

I see what you did there with the "you'll need to..." !

I appreciate attrs, but I'm not up for implementing this now. You might prefer to close this PR and migrate comments to #39.

[hynek]

No problem! This is kind of the reason, why we still don’t have it: it’s more work than it looks initially if you want to cover all bases. :-/

[wsanchez]

This feels like a bit of reinventing the type annotation system in Py3.

If I understand this correctly, the feature is validator=list_of(int).

Why not validator=instance_of(List[int])?

That is, I think I'd rather write this code:

from pathlib import Path
from typing import Iterable

from attr import attrib, attrs
from attr.validators import instance_of


@attrs()
class ThingWithPaths(object):
    _paths = attrib(validator=is_type(Iterable[Path]))

    def exists(self):
        for path in self._paths:
            if not path.exists():
                return False
        return True

thing1 = ThingWithPaths([Path("."), Path("foo")])
print("thing1 exists:", thing1.exists())

thing2 = ThingWithPaths([".", "foo"])  # validation fails here
print("thing2 exists:", thing2.exists())

So what's missing is is_type, instead of a series of foo_of validators.

Another advantage here is that if we have can find the type annotation (eg #151), then is could even be simpler: _paths = attrib(validator=typing) where the type for the validator to check is obtained from the annotations.

[wsanchez]

See #158 (comment)

[IlyaSkriblovsky]

@wsanchez, typing doesn't work for dynamic checking using isinstance:

>>> isinstance(['qwe', 'asd'], List[int])
True

typing is a declaration-only lib, it doesn't provide any means of runtime typechecking.

[wsanchez]

@IlyaSkriblovsky Yeah, I'm figuring that out, which is weak sauce but that's why I changed the code to is_type from instance_of. It wasn't meant to be a thing that already works so much as what would be more awesome. :-)

I'll add here a reference to a ticket to fix that, which you just sent to Glyph: python/typing#377

I guess it's just unfortunate not to be able to leverage typing better. From your work on cattrs, I got the impression that it was possible, but that was optimistic, I guess.

[hynek]

So my understanding is that is_type with py3 syntax is not feasible anytime soon? In any case, as much as I loath Python 2, I’m sure a Python 3-only option would not stop people opening an issue and asking for this every week. :)

So I guess we still need a solution for all the legacy-impaired.

Closing this at the request of the reporter tho.