On Fri, Dec 1, 2017 at 10:30 AM, Chris Lattner via swift-evolution < > swift-evolution@swift.org> wrote:
On Dec 1, 2017, at 12:26 AM, Douglas Gregor <dgregor@apple.com> wrote:
*Philosophy*
Swift is, unabashedly, a strong statically-typed language.
That’s factually incorrect.
You’re going to have to explain that statement without reference to
AnyObject (we’ll discuss that case below).
Perhaps it depends on what you mean by “strong”: I interpreted that as
meaning that it provides type safety, with a level of strength akin to what
Java provides: a level that could support mobile code deployment.
Swift certainly does not provide that strong of a static type system,
because it gives people explicit ways to opt out of that.
UnsafeMutableRawPointer, unsafe bitcast, and many other facilities support
this. It also allows calling into non-type safe code, so it isn’t very
strong that way. There are also race conditions and other holes in the
type system.
All that said, I think it is correct that a subset of Swift exists that
does provide strong type safety, but particularly when bridging to C/ObjC
is involved, that quickly goes away.
More problematically for your argument: your preferred approach requires
the introduction of (something like) DynamicMemberLookupProtocol or
something like AnyObject-For-Python, so your proposal would be additive on
top of solving the core problem I’m trying to solve. It isn’t an
alternative approach at all.
I wouldn’t say that’s my preferred approach. My preferred approach
involves taking the method/property/etc. declarations that already exist in
Python and mapping them into corresponding Swift declarations so we have
something to find with name lookup. One could put all of these declarations
on some PyVal struct or PythonObject and there would be no need for
AnyObject-for-Python or DynamicMemberLookupProtocol.
You’re suggesting that the transitive closure of all Python methods and
properties be preprocessed into a single gigantic Swift PyVal type? I
guess something like that could be done.
I would be concerned because there are many N^2 or worse algorithms in
the Swift compiler would probably explode. It also doesn’t provide the
great tooling experience that you’re seeking, given that code completion
would show everything in the Python universe, which is not helpful.
Further, it doesn’t provide a *better* experience than what I’m
suggesting, it seems strictly worse. A preprocessing step prevents users
from playfully importing random Python modules into the Swift repl and
playgrounds. It also seems worse for implementors (who will get a stream
of new bugs about compiler scalability).
Whenever we discuss adding more dynamic features to Swift, there’s a
strong focus on maintaining that strong static type system.
Which this does, by providing full type safety - unlike AnyObject lookup.
You get dynamic safety because it goes into the Python interpreter; fair
enough. You get no help from your tools to form a correct invocation of any
method provided by Python.
Sure, that’s status quo for Python APIs.
IMO, this proposal is a significant departure from the fundamental
character of Swift, because it allows access to possibly-nonexistent
members (as well as calls with incorrect arguments, in the related
proposal) without any indication that the operation might fail.
The only way your claim is correct is if someone implements the protocol
wrong. What you describe is true of AnyObject lookup though, so I
understand how you could be confused by that.
AnyObject lookup still requires you to find an actual declaration with a
type signature. Yes, there are still failure cases. The dynamic type might
be totally unrelated to the class in which you found the declaration you’re
supposedly calling, which is a typical “unrecognized selector” failure and
will always be an issue with dynamic typing. The actual type safety hole
you’re presumably referring to is that the selector could be overloaded
with a different type signature, and we don’t proactively check that the
signature we type-checked against matches the signature found at runtime.
It’s doable with the Objective-C method encodings, but has never been
considered worthwhile.
To be clear, I’m not suggesting a change to AnyObject lookup. I don’t
think that it is worth changing at this point in time. My point was to
observe that the proposed DynamicMemberLookupProtocol proposal does not
suffer from these problems. It really is type / memory safe, assuming a
sane implementation.
It’s easy to fall through these cracks for any type that supports
DynamicMemberLookupProtocol—a single-character typo when using a
DynamicMemberLookupProtocol-capable type means you’ve fallen out of the
safety that Swift provides.
Since you seem to be latching on to this, I’ll say it again: the proposal
is fully type safe :-)
That said, the “simple typo” problem is fundamental to the problem of
working with a dynamically typed language: unless you can eradicate all
“fundamental dynamism” from the language, you cannot prevent this.
That said, since this is a problem inherent with these languages, it is
something people are already very familiar with, and something that
everyone using those APIs has had to deal with. This is also not our
problem to solve, and people in the Python community have been discussing
and working on it over a large part of its 25 year history. I find your
belief that we could solve this for Python better than the Python community
itself has both idealistic and a bit naive.
I’m not pretending we can fully solve the problem. I’m pointing out that
depending entirely on DynamicMemberLookupProtocol throws away information
about method declarations that is present in Python and used by Python
tooling to good effect.
I’m not opposed to going further over time, but I’d like to get started
at some point :-). I’m not in a super urgent hurry to get this in in the
next week or month or anything like that, but I also don’t want to wait
until Swift 10.
*Tooling*
The Python interoperability enabled by this proposal *does* look fairly
nice when you look at a small, correctly-written example. However,
absolutely none of the tooling assistance we rely on when writing such code
will work for Python interoperability. Examples:
* As noted earlier, if you typo’d a name of a Python entity or passed the
wrong number of arguments to it, the compiler will not tell you: it’ll be a
runtime failure in the Python interpreter. I guess that’s what you’d get if
you were writing the code in Python, but Swift is supposed to be *better*
than Python if we’re to convince a community to use Swift instead.
* Code completion won’t work, because Swift has no visibility into
declarations written in Python
* Indexing/jump-to-definition/lookup documentation/generated interface
won’t ever work. None of the IDE features supported by SourceKit will work,
which will be a significant regression for users coming from a
Python-capable IDE.
Yes, welcome to the realities of modern Python development!
Python plugins for IDEs (e.g., for Atom) provide code completion, goto
definition, and other related features. None of the Swift tooling will work
if Swift’s interoperability with Python is entirely based on
DynamicMemberLookupProtocol.
I don’t understand your rationale here. I think you agree that we need
to support the fully dynamic case (which is what I’m proposing). That
said, this step does not preclude introducing importer magic (either
through compiler hackery or a theoretical "type providers” style of
feature). Doing so would provide the functionality you’re describing.
Statically-typed languages should be a boon for tooling, but if a user
coming from Python to Swift *because* it’s supposed to be a better
development experience actually sees a significantly worse development
experience, we’re not going to win them over. It’ll just feel inconsistent.
By your argument we should ban AnyObject lookup as well, given its
inconsistency with the rest of the language.
By my argument, we should at least replace the
ImplicitlyUnwrappedOptional result with a true Optional (so one has to
acknowledge that the method shouldn’t be there). We’ve seriously considered
it, but it’s a source-breaking change, and it hasn’t seemed worth the
engineering effort to pursue it.
Ok, that would be a nice step, but doesn’t fix the type safety hole.
In any case, my proposal allows the use of strong optional results as
well, so the fate of AnyObject isn’t really bound up with it.
I don’t think that removing AnyObject dispatch entirely is possible at
this point in Swift’s lifetime. While AnyObject has become much less
prominent than it was in the Swift 1.0 days ([AnyObject] and [NSObject :
AnyObject], oh my!), there is still a significant amount of code using it
in the wild.
Completely agreed. The major advantage I see of changing it now is if
there is some small mostly-user-invisible-change that allows a dramatic
simplification to the compiler implementation.
*Dynamic Typing Features*
It’s possible that the right evolutionary path for Swift involves some
notion of dynamic typing, which would have a lot of the properties sought
by this proposal (and the DynamicCallableProtocol one). If that is true—and
I’m not at all convinced that it is—we shouldn’t accidentally fall into a
suboptimal design by taking small, easy, steps.
Given that you haven’t followed the discussion on the many threads we’ve
had on this, and haven’t proposed a workable approach to this problem, I’m
not sure upon what basis your fears and uncertainty and doubt are founded.
A few meta-comments here. First of all, following all previous threads on
a discussion is not realistic.
I understand that, but you’re also accusing the proposal of being a
suboptimal design made by looking at a series of small easy steps, instead
of the right design for the long term. I’m pointing out that it is hard to
see the rationale for that sort of claim.
This is part of the reason why we have different stages in a proposal’s
lifetime, and is the responsibility of the proposal’s authors to capture
alternatives and rationale in the proposal to make it self-contained.
Agreed. As I mentioned in my previous email, I definitely screwed that
up by not capturing this discussion in the proposal. Thank you again for
pulling this perspective to the front of the discussion so I could fix that
oversight.
This proposal went through rapid iteration in the pitch phase and has now
been elevated to a pull request to ask for formal review—you should expect
more people to come on board having not read those threads. I appreciate
that you have now captured more alternatives and rationale in the proposal.
Agreed. This is why I’ve been proactive about starting threads and
trying to keep visibility on the proposal each time there is a significant
change. I really do value the discussion and feedback (both on the
proposed direction but also the writing itself).
Second, it is absolutely reasonable to disagree with the technical
direction of a proposal without providing a complete solution to the
problem that the proposal is attempting to solve. Some problems aren’t
worth solving at all, or fully.
Ok, but at some point, if there is no alternative proposed, then a strong
opposition has the appearance of saying “we shouldn’t solve this problem”.
It was my understanding that thought that this was a worthwhile problem to
solve.
*How Should Python Interoperability Work?*
Going back to the central motivator for this proposal, I think that
providing something akin to the Clang Importer provides the best
interoperability experience: it would turn Python declarations into *real*
Swift declarations, so that we get the various tooling benefits of having a
strong statically-typed language.
This could be an theoretically interesting refinement to this proposal
but I’m personally very skeptical that this is every going to happen. I’ve
put the rationale into the alternatives section of the proposal. I don’t
explain it in the proposal in this way directly, but I believe it is far
more likely for a Pythonista transplant into Swift to rewrite their code in
Swift than it is to use Python type annotations.
I assume that this belief is based on type annotations lack of traction
in the Python community thus far?
There are many parts to this, which have to do with the ObjC<->Swift
situation being very different than the Python<->Swift situation:
1) The annotations don’t have significant traction in the Python
community.
2) The Python annotations are not as powerful as ObjC generics are, and
thus lack important expressive capability.
3) Many Python APIs are wrappers for C APIs. “Swiftizing” a Python API
in this case means writing a new Swift wrapper for the API, not adding type
annotations.
4) The Python community doesn’t care about Swift, and are not motivated
to do things to make Swift succeed.
5) There is no “clang equivalent” for Python (that I’m aware of) which
close enough to the way Clang does for us to directly use. The owners of
the existing Python compiler/interpreter implementations are not going to
be strongly motivated to change their stuff for us.
Finally, just MHO, but I don’t expect a lot of “mix and match"
Python/Swift apps to exist (where the developer owns both the Python and
the Swift code), which is one case where type annotations are super awesome
in ObjC. IMO, the most important use-case is a Swift program that uses
some Python APIs.
Sure, the argument types will all by PyObject or PyVal,
That’s the root of the problem. Python has the “fully dynamic”
equivalent of “id” in Objective-C, so we need to represent that somehow.
Even if we followed the implementation approach of the Clang importer, we
would need some way to represent this dynamic case. That type needs
features like DynamicMemberLookup or AnyObject. In my opinion, the
DynamicMemberLookup approach is better in every way than AnyObject is.
The AnyObject approach has the advantage of knowing the set of declared,
reachable APIs:
* Code completion shows all of the APIs that are possible to use via
dynamic dispatch, with their signatures so can fill in the right # of
arguments, see the names of the parameters, see documentation, etc.
* Indexing/refactoring/goto definition all point you to the declarations
that could be the targets of dynamic dispatch
The DynamicMemberLookup approach is better for cases where you don’t have
a declaration of the member you want to access. I suspect that’s not the
common case.
Your points are valid, but the advantages for Objective-C don’t obviously
translate to Swift. Note that ObjC (due to its heritage) has very long
method names that are perhaps arguably designed to not conflict with each
other often. Python doesn’t have this heritage, and it has much shorter
names, which means that we’ll get a lot more conflicts and a lot less
“safety" out of this.
AnyObject lookup also depends on a strange set of scoping heuristics that
was designed to be similar to Clang’s “header import” scope. It isn’t
clear that this approach will work in Python, given that it doesn’t have an
analogue of umbrella headers that import things that cross frameworks.
Which approach do you think is the best way to handle the untyped
“actually dynamic” case?
AnyObject already exists in the language, and it fits the untyped
“actually dynamic” case well. It does require having a declaration for the
thing you want to reference, which I consider to be important: we can
code-complete those declarations, goto-definition to see those
declarations, index/refactor/look up documentation based on those
declarations.
I’d be more inclined to push for the ImplicitlyUnwrappedOptional ->
Optional change if we did something to make AnyObject more prominent in
Swift.
I didn’t realize that you were thinking we would literally use AnyObject
itself. I haven’t thought fully through it, but I think this will provide
several problems:
1) You’re mushing all of the ObjC and Python world’s together, making the
ObjC interop worse just because you’re doing some Python stuff too.
2) You’re introducing ambiguity: does “ao = [1,2,3]” create an NSArray or
a Python array? How do string literals work? (The answer is obvious,
Python loses). Maybe there is some really complicated bridging solution to
these problems, but that causes its own massive complexity spiral.
3) You can’t realistically overload the Python operator set on AnyObject,
which means you get a worse python experience.
4) AnyObject magic is currently limited to Apple platforms. This would
bring its problems to other platforms like Linux.
There are probably other issues, but I haven’t thought through it.
but the names are there for code completion (and indexing, etc.) to work,
and one could certainly imagine growing the importer to support Python’s typing
annotations <https://docs.python.org/3/library/typing.html>\.
You’re basing this on the flawed assumption that local variables will
pervasively have types, which I can’t imagine being the case. Even on
"typable” API, I wouldn’t expect people to commonly get code completion
results for reasons now explained in the proposal.
Remember that one *does* get code completion results for member access
into an AnyObject… lots of them… but the list filters down pretty fast when
you type a few characters, and then you get a member access that’ll fill in
stubs for (say) the arguments to the method you were trying to call. But
you can’t get those code completion results without having declarations to
complete to.
Fair point, it’s unclear to me how useful this would be with python’s
style of naming, but it could work.
In truth, you don’t even need the compiler to be involved. The dynamic
“subscript” operation could be implemented in a Swift library, and one
could write a Python program to process a Python module and emit Swift
wrappers that call into that subscript operation. You’ll get all of the
tooling benefits with no compiler changes, and can tweak the wrapper
generation however much you want, using typing annotations or other
Python-specific information to create better wrappers over time.
I’d love for you to sketch out how any of this works with an acceptable
user experience, because I don't see anything concrete here.
We don’t need the basic dynamic case in the language to do this
experiment. Take the PyVal struct from the proposal. Now, write a Python
script that loads some module Foo and uses Python’s inspect
<https://docs.python.org/3/library/inspect.html> module to go find the
classes, methods, etc., and pretty-print Swift code that uses PyVal. So
this:
def add_trick(self, trick):
turns into
extension PyVal {
func add_trick(_ trick: PyVal) -> PyVal {
/* do the magic to call into Python */
}
}
Using the inspect module, you can extract parameter names, default
arguments, docstrings, and more to reflect the existing Python API as Swift
API, packed into a bridging module.
Note that we have a “flat” namespace of all Python methods on PyVal,
which is basically what you get with AnyObject today. Swift tooling will
provide code completion for member accesses into PyVal. Goto definition
will jump to the pretty-printed declarations, which could have the
docstrings formatted in comments and would show up in QuickHelp. The types
are weak (everything is PyVal), but that’s what we expect from importing a
dynamically-typed language.
As I mention above, I expect this to expose significant scalability
problems in the Swift compiler and it also defeats REPL/Playgrounds. Being
able to use the Swift REPL is really important for Python programmers.
-Chris
_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution