(name :under_scored) returns "under-scored"... Why?

[jedi2light]

I have the next code:

(setv sample :under_scored)
(name sample)

Expected:

"under_scored"

Actually:

"under-scored"

At the same time, Clojure's namereturns"under_scored"`. Please, fix it.

[Kodiologist]

name calls unmangle internally. name itself is of dubious value, though; we should probably just trash it.

[allison-casey]

Could we just make an exception for keywords since they should probably stay as close to their string representation as possible anyways?

[Kodiologist]

It would be better for the user to use whatever combination of str, mangle, unmangle, and (. … __name__) that they want. Again, I don't see what the value of name is to begin with.

[allison-casey]

because as far as i can tell none of those get you the underlying name of the keyword.

(str :hello_world) ; =>  :hello_world
(repr :hello_world) ; =>  HyKeyword('hello_world')
(mangle :hello_world) ; =>  hyx_XcolonXhello_world
(unmangle :hello_world) ; =>  :hello-world
(name :hello_world) ; =>  hello-world

The closest to what jedi wants is str but that doesn't remove the : and it's such a common operation that it doesn't seem right to make users do (.replace (str ':hello_world) ":" "" 1) every time or make their own function. Clojure doesn't unmangle keywords and we probably shouldn't either

[Kodiologist]

You can use sym.name or (cut sym 1).

[allison-casey]

lol didn't know that sym.name existed that changes things.
@jedi2light you should probably use (. :hello_!->world name) ; => "hello_!->world"
instead. since hy doesn't have namespaces like clojure i think i agree with kodi now. There doesn't seem much use for name besides providing an alias for (. key name)

[allison-casey]

There is one use case I'm starting to think of and that's the ability to normalize dict keys to strings easily.

(setv x {"a" 1  :b 2})

(dfor [k v] (x.items) [(name k) v]) ; => {"a" 1 "b" 2}
;; you could write that as
(dfor [k v] (x.items) [(if (keyword? k) k.name k) v]) ; => {"a" 1  "b" 2}

Don't know how valid of use case this is to keep name around, but something to think about

[Kodiologist]

Generally, it's a mistake to use a HyKeyword instead of a string as a dictionary key. It's too easy to confuse yourself, especially when you're looking at forms like (dict :foo 1).

[allison-casey]

yeah, that's why i don't use them myself, but i'm trying to think of reasons to keep name around since hy doesn't have clojure style namespaces which clojure's name handles specially. So far i'm not comming up with much 🤷‍♀️

[scauligi]

it's a mistake to use a HyKeyword instead of a string as a dictionary key

That seems to go against the intention of the HyKeyword.__call__ implementation, but I agree that keywords-as-keys gets confusing with keywords-as-keyword-arguments. Should we consider removing the (:foo mapping) feature? Or instead, we could change it to do a string lookup instead of a HyKeyword lookup (e.g. (:foo (dict :foo 1)) would produce 1).

[Kodiologist]

Probably the latter.

[peaceamongworlds]

Or instead, we could change it to do a string lookup instead of a HyKeyword lookup (e.g. (:foo (dict :foo 1)) would produce 1).

I think that string lookup would be the most convenient. It would actually allow for it to be used with other python objects like pandas dataframes etc. Using keywords outside of function arguments doesn't seem to mix well with the python ecosystem.

[allison-casey]

I like the string lookup idea. I pretty much never use the (:akey adict) type expression because keywords as dict keys is generally a bad idea anyways.

so would (:hello {:hello 1}) be a KeyError then?

[peaceamongworlds]

so would (:hello {:hello 1}) be a KeyError then?

I guess it would just be equal to (get {:hello 1} (. :hello name)), which throws a KeyError.

[jedi2light]

Remove (:some col)? Ok, will switch to another language then :/

[allison-casey]

The current proposals would still work though? it would just be a lookup for a string key.
(:hello {"hello": 1}) ; => 1 So i'm confused?

[scauligi]

And in particular, (:hello (dict :hello 1)) ; => 1 with the new proposal, which seems like a plus.

[scauligi]

Awesome, looks like there's enough buy-in so I'll start up a PR for this.