keyword lookup on NamedTuple

[shaunlebron]

I believe a NamedTuple in python supports two types of lookups: attribute name or index.

Right now, I have to use (py.- nt :foo) whereas (:foo nt) does not work and returns nil. I believe this is because the jvm bridge is not able to treat a namedtuple as both a dict and a tuple. Is this possible?

Thank you!

(I deleted a followup comment about (py. nt _asdict) not working, but it was my error in not requiring "py.")

[jjtolton]

Clojure Keywords are not currently polymorphic with python __getattr__.

…

On Fri, Apr 22, 2022 at 1:23 AM shaun ***@***.***> wrote: I believe a NamedTuple <https://docs.python.org/3.3/library/collections.html#namedtuple-factory-function-for-tuples-with-named-fields> in python supports two types of lookups: attribute name or index. Right now, I have to use (py.- nt :foo) but (:foo nt) returns nil. I believe this is because the jvm bridge is not able to treat a namedtuple as both a dict and a tuple. Is this possible? Thanks! — Reply to this email directly, view it on GitHub <#202>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACPJX42LLX3M4TWEMMF45ILVGIZWZANCNFSM5UBF5LRA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jjtolton]

Did you require “py.”?

…

On Fri, Apr 22, 2022 at 1:35 AM shaun ***@***.***> wrote: Also, while trying to convert a namedtuple to a dict using ._asdict() <https://docs.python.org/3/library/collections.html#collections.somenamedtuple._asdict>, I had a peculiar run-in with the following two expressions which I presumed would be equivalent: - (py. nt _asdict) fails with "unable to resolve classname: py" - ((py.- nt _asdict)) succeeds — Reply to this email directly, view it on GitHub <#202 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACPJX43NDCFWRAR4HMB7AHLVGI3BPANCNFSM5UBF5LRA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[shaunlebron]

Thanks. Is there a way to make keywords polymorphic to getattr from the user side?

[jjtolton]

(moved to comment below b/c apparently markdown doesn't work from email)

[jjtolton]

Do you mean extending libpython-clj itself? We do have hooks into the
datafy/nav paradigm, which would be one option. Checkout the code here:
https://github.com/clj-python/libpython-clj/blob/master/src/libpython_clj2/require.clj#L352

You could probably do something like:

(defmethod pydafy 'builtins.NamedTuple [nt] 
;; code to convert named tuple -> clojure, possibly in conjunction 
 ;; with a Clojure NamedTuple protocol that extends `get` 
... ) 

note that once you've extended with pydafy,
you can then use standard Clojure datafy function. This would result in
code that might look like, at a low level,

(:foo (datafy nt))

There are lower level hooks available for performant marshalling of
data types from Python, but that would be a broader feature request
discussion.

[shaunlebron]

Thanks so much for the example, I was not familiar with datafy. Glad to know this.

And just to share here, I missed the following distinction in Python (since they’re both equivalent in JS):

• __getitem__ is for foo['bar'] = foo.get('bar')
• __getattr__ is for foo.bar = getattr(foo, 'bar')

So I understand your original response now, that keywords map to getitem, whereas py.- is to be used for getattr. I’ll just use it that way to not hide what's happening in python. Thank you!

[jjtolton]

Great point! We're always discussing the right balance to strike in how much to hide/expose. Always open to suggestions, thank you for your input!

…

On Fri, Apr 22, 2022 at 1:15 PM shaun ***@***.***> wrote: Thanks so much for the example, I was not familiar with datafy. Glad to know this. And just to share here, I missed the following distinction in Python (since they’re both equivalent in JS): - __getitem__ is for foo['bar'] = foo.get('bar') - __getattr__ is for foo.bar = getattr(foo, 'bar') So I understand your original response now, that keywords map to getitem, whereas py.- is to be used for getattr. I’ll just use it that way to not hide what's happening in python. Thank you! — Reply to this email directly, view it on GitHub <#202 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACPJX44SHSJV2VVHVJTJSQLVGLNCHANCNFSM5UBF5LRA> . You are receiving this because you commented.Message ID: ***@***.***>

[shaunlebron]

Thanks. I don't really know what I'm doing yet, but I can't believe this project exists AND works flawlessly so far. Thanks for all your work 🙏