Names partial-apply and partial-reverse-apply would be my preferred

when looking at the issue with a clean mind, without having read any papers about currying (there are newborns every day!), the long versions of the proposed names are much more intuitive.

just one more 0.02,

Nikodemus Siivola nikodemus@random-state.net writes:

I'm not sure. I'm not totally opposed, but I'm not convinced either: even if it wrong, the names are "culturally correct", if you will. Names partial-apply and partial-reverse-apply would be my preferred ones if we did the renaming.

I find them too long. Long names can distract from the intent of the actual code. They're supposed to merely combine domain-level logic, but being long they make the impression they're more than that.

-T.

On Jun 8, 2009, at 09:40 , Tobias C. Rittweiler wrote:

Nikodemus Siivola nikodemus@random-state.net writes:

...

I'm not sure. I'm not totally opposed, but I'm not convinced either: even if it wrong, the names are "culturally correct", if you will. Names partial-apply and partial-reverse-apply would be my preferred ones if we did the renaming.

I find them too long. Long names can distract from the intent of the actual code. They're supposed to merely combine domain-level logic, but being long they make the impression they're more than that.

FWIW, I agree with Tobias' arguments.

I'd vote for PAPPLY (and probably by extension REVPAPPLY) if there is to be a name change.

On 6/8/09, Attila Lendvai attila.lendvai@gmail.com wrote:

...

I find them too long. Long names can distract from the intent of the actual code. They're supposed to merely combine domain-level logic, but being long they make the impression they're more than that.

if i see PAPPLY (or any other abbreviated name for that matter), my brain first maps it to PARTIAL-APPLY and then to the meaning itself.

IMHO, abbreviation is bad because it can be done in a million ways and it only introduces an extra level of indirection which is an (admittedly small) obstacle against understanding.

of course the picture is different for names that are used in every second line.

it might be only me, though. and i don't have strong feelings because names are inherently subjective and our tools should handle the personal preferences. CL + slime does not provide much in this regard, but packages and symbol aliases work.

I tend to agree on long names & short names being an orthogonal issue.

What is this "symbol aliases" you mentioned? Is it some common idiom or unannounced library of yours?

I was wondering if anyone knows of any equivalent to Python's

import some_long_external_module_name as shortname

using just the plain CL package system?

Yong.

2009/6/4 Nikodemus Siivola nikodemus@random-state.net:

...

What about papply or partial-apply? Or maybe p-apply? This way, The reversed version might be called rp-apply. Otherwise reverse-partial-apply?

I am aware that this is a common usage/mistake in CL, but Alexandria doesn't have to perpetuate this error. An unfortunate downside is breaking backwards compatibility. I'm not sure what Alexandria's policy is regarding this...

I'm happy to send patches if this renaming is agreed upon.

I'm not sure. I'm not totally opposed, but I'm not convinced either: even if it wrong, the names are "culturally correct", if you will. Names partial-apply and partial-reverse-apply would be my preferred ones if we did the renaming.

I'm slowly coming to think that calling them CURRY and RCURRY is indeed a mistake we should stop propagating.

(defun partial-apply (function &rest partial-arguments) "Returns a function that calls FUNCTION with PARTIAL-ARGUMENTS followed by the arguments it is called with." (declare (optimize (speed 3) (safety 1) (debug 1))) (let ((fn (ensure-function function))) (lambda (&rest more) (declare (dynamic-extent more)) ;; Using M-V-C we don't need to append the arguments. (multiple-value-call fn (values-list partial-arguments) (values-list more)))))

(define-compiler-macro partial-apply (function &rest partial-arguments) (let ((temps (make-gensym-list (length partial-arguments) "PARTIAL-ARG"))) `(let ,(mapcar #'list temps partial-arguments) (declare (optimize (speed 3) (safety 1) (debug 1))) (lambda (&rest more) (declare (dynamic-extent more)) (apply ,function ,@temps more)))))

(defun reverse-partial-apply (function &rest partial-arguments) "Returns a function that calls FUNCTION with the arguments it is called with followed by PARTIAL-ARGUMENTS." (declare (optimize (speed 3) (safety 1) (debug 1))) (let ((fn (ensure-function function))) (lambda (&rest more) (declare (dynamic-extent more)) (multiple-value-call fn (values-list more) (values-list partial-arguments)))))

Cheers,

-- Nikodemus

2009/8/6 Leslie P. Polzer sky@viridian-project.de:

Nikodemus Siivola wrote:

...

I'm slowly coming to think that calling them CURRY and RCURRY is indeed a mistake we should stop propagating.

Alright, but don't you agree that partial-apply and esp. reverse-partial-apply are names that are too long for such a basic function?

No. Clarity is far more important than being terse: they are no slower to read, and meaning is that much more immediately obvious. Note though that I'm not convinced either is a real win over

(lambda (b) (foo a b))

and

(lambda (a) (foo a b))

in any way. Using PARTIAL-APPLY above would be sophistry and obfuscation. They become nice only when the number of arguments is increases:

(partial-apply #'foo a b c)

rather than

(lambda (d e f) (foo a b c d e f))

in which case the lenght of the name is a non-issue:

(papply #'foo a b c)

saves a whopping 6 letters.

Cheers,

-- Nikodemus

Attila Lendvai wrote:

if people second the other opinion, then i feel that i must speak up on behalf of the non-abbreviation group.

the need for abbreviation (if any) should be satisfied by a tool (the editor), not a library.

I also vote for the longer, more descriptive name. Auto-complete is my friend.

Thanks,

Ryan Davis Acceleration.net Director of Programming Services 2831 NW 41st street, suite B Gainesville, FL 32606

Office: 352-335-6500 x 124 Fax: 352-335-6506

From: "Leslie P. Polzer" sky@viridian-project.de

Nikodemus Siivola wrote:

...

I'm slowly coming to think that calling them CURRY and RCURRY is indeed a mistake we should stop propagating.

Alright, but don't you agree that partial-apply and esp. reverse-partial-apply are names that are too long for such a basic function?

I second this.

regards, Samium Gromoff