On 29/08/2020 14:17, Barry Scott wrote:

...

On 29 Aug 2020, at 13:42, Filipp Bakanov mailto:filipp@bakanov.su> wrote:

I'd like to propose adding argmax and argmin functions to the python list. These functions return the index of a maximum / minimum element of the list. Eg:

a = [1, 4, 2, 3] print(a.argmax())  # 1 print(a.argmin())  # 0

It's a very popular request (based on stackoverflow https://stackoverflow.com/questions/16945518/finding-the-index-of-the-value-... ), and currently there is no elegant way to find it.

What do you think?

Just do this:

...

...

...

a=[1,4,2,3] min(a) 1 a.index(min(a)) 0 a.index(max(a)) 1

Barry

This has the drawback of passing twice over the list. The following doesn't, but the complexity somewhat makes Filipp's point:

...

...

min((e, i) for i, e in enumerate(a))[1] 0

I think one would want argmin() and argmax() to work with general iterables, so I wonder if the stdlib would not be a better home than list itself. I half expected it to be an itertools recipe. The advantage of a recipe is that variations such as needing the last occurrence of the minimum are easily accommodated. Jeff Allen

It's expected that python single iterating would be slower than two times C iterating. Nevertheless one time C iterating will be probably faster than a separate min + index.

...

I think one would want argmin() and argmax() to work with general iterables How is it supposed to work with set or dict or other iterables without clear order?

My last case was to find the min index of the array part. index_min could accept the starting element index, as index does. a = [1, 4, 1, 2, 3] print(a.index_min()) # 0 print(a.index_min(1)) # 2 On Sun, 30 Aug 2020 at 11:52, Barry Scott wrote:

On 30 Aug 2020, at 09:03, Jeff Allen wrote:

On 29/08/2020 14:17, Barry Scott wrote:

On 29 Aug 2020, at 13:42, Filipp Bakanov wrote:

I'd like to propose adding argmax and argmin functions to the python list. These functions return the index of a maximum / minimum element of the list. Eg:

a = [1, 4, 2, 3] print(a.argmax()) # 1 print(a.argmin()) # 0

It's a very popular request (based on stackoverflow https://stackoverflow.com/questions/16945518/finding-the-index-of-the-value-... ), and currently there is no elegant way to find it.

What do you think?

Just do this:

...

...

...

a=[1,4,2,3] min(a) 1 a.index(min(a)) 0 a.index(max(a)) 1

Barry

This has the drawback of passing twice over the list. The following doesn't, but the complexity somewhat makes Filipp's point:

...

...

...

min((e, i) for i, e in enumerate(a))[1] 0

That is 4x slower then my code for 1,000,000 items.

--------------- a.py -------------- import sys import time import random

alg = sys.argv[1] size = int(sys.argv[2])

x = [random.randint(0, 1_000_000) for _ in range(size)]

start = time.time()

if alg == 'barry':

m = x.index(min(x))

elif alg == 'jeff':

m = min((e, i) for i, e in enumerate(x))[1]

end = time.time()

print( alg, end-start, (end-start)/size ) -------------------------

Here is the output I got on my laptop.

barry 0.022754907608032227 2.2754907608032226e-08 barry 0.03325295448303223 3.325295448303223e-08 barry 0.034243106842041016 3.4243106842041016e-08 barry 0.02784109115600586 2.784109115600586e-08 jeff 0.13722586631774902 1.3722586631774904e-07 jeff 0.1359708309173584 1.359708309173584e-07 jeff 0.13658690452575684 1.3658690452575684e-07

I think one would want argmin() and argmax() to work with general iterables, so I wonder if the stdlib would not be a better home than list itself. I half expected it to be an itertools recipe. The advantage of a recipe is that variations such as needing the last occurrence of the minimum are easily accommodated.

Surely its index_min() and index_max() not argmin() and argmax().

Barry

_______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/VMSSYH... Code of Conduct: http://python.org/psf/codeofconduct/

On 30 Aug 2020, at 12:25, Filipp Bakanov wrote:

 It's expected that python single iterating would be slower than two times C iterating. Nevertheless one time C iterating will be probably faster than a separate min + index.

...

...

I think one would want argmin() and argmax() to work with general iterables How is it supposed to work with set or dict or other iterables without clear order?

My last case was to find the min index of the array part. index_min could accept the starting element index, as index does.

a = [1, 4, 1, 2, 3] print(a.index_min()) # 0 print(a.index_min(1)) # 2

Of course this could be implemented in C in python, but you are going to have to convince the python developers that it’s important enough to want to maintain the code. Is your use of such a function so performance sensitive that you need this in C? Have you considered writing this as a C extension for your own use? Barry

On Sun, 30 Aug 2020 at 11:52, Barry Scott wrote:

...

...

...

On 30 Aug 2020, at 09:03, Jeff Allen wrote:

On 29/08/2020 14:17, Barry Scott wrote:

...

...

On 29 Aug 2020, at 13:42, Filipp Bakanov wrote:

I'd like to propose adding argmax and argmin functions to the python list. These functions return the index of a maximum / minimum element of the list. Eg:

a = [1, 4, 2, 3] print(a.argmax()) # 1 print(a.argmin()) # 0

It's a very popular request (based on stackoverflow https://stackoverflow.com/questions/16945518/finding-the-index-of-the-value-... ), and currently there is no elegant way to find it.

What do you think?

Just do this:

...

>> a=[1,4,2,3] >> min(a) 1 >> a.index(min(a)) 0 >> a.index(max(a)) 1

Barry

This has the drawback of passing twice over the list. The following doesn't, but the complexity somewhat makes Filipp's point:

...

...

> min((e, i) for i, e in enumerate(a))[1] 0

That is 4x slower then my code for 1,000,000 items.

--------------- a.py -------------- import sys import time import random

alg = sys.argv[1] size = int(sys.argv[2])

x = [random.randint(0, 1_000_000) for _ in range(size)]

start = time.time()

if alg == 'barry':

m = x.index(min(x))

elif alg == 'jeff':

m = min((e, i) for i, e in enumerate(x))[1]

end = time.time()

print( alg, end-start, (end-start)/size ) -------------------------

Here is the output I got on my laptop.

barry 0.022754907608032227 2.2754907608032226e-08 barry 0.03325295448303223 3.325295448303223e-08 barry 0.034243106842041016 3.4243106842041016e-08 barry 0.02784109115600586 2.784109115600586e-08 jeff 0.13722586631774902 1.3722586631774904e-07 jeff 0.1359708309173584 1.359708309173584e-07 jeff 0.13658690452575684 1.3658690452575684e-07

I think one would want argmin() and argmax() to work with general iterables, so I wonder if the stdlib would not be a better home than list itself. I half expected it to be an itertools recipe. The advantage of a recipe is that variations such as needing the last occurrence of the minimum are easily accommodated.

Surely its index_min() and index_max() not argmin() and argmax().

Barry

_______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/VMSSYH... Code of Conduct: http://python.org/psf/codeofconduct/

On Sun, 30 Aug 2020 09:51:14 +0100 Barry Scott wrote:

That is 4x slower then my code for 1,000,000 items.

This is highly dependent on the cost of equality checks, therefore on the type being compared. Regards Antoine.

On 31/08/2020 01:14, Christopher Barker wrote:

On Sun, Aug 30, 2020 at 7:28 AM Barry mailto:barry@barrys-emacs.org> wrote:

...

How is it supposed to work with set or dict or other iterables without clear order?

see the discussion in another recent thread about making dict indexable -- which looks like it's not going to happen. Unfortunately, there are a lot of those. so no -- this should not work with general iterables, indexes don't really make sense for iterables, only Sequences.

It would be as useful as:

...

...

min(dict(b=1, c=3, a=2)) 'a'

I half expected 1 (and from argmin would expect 'b') but I see why the actual result is "consistent". Jeff