I came across a function named `wsample` in a `utils` package of my workplace recently.
The "w" in `wsample` stands for `weighted`, and it randomly selects an element from a container according to relative weights of all the elements.
In most articles and codes I saw online, a function like this is often named `weighted_random_choice`, which sounds *correct* to me.
So when I saw this `wsample` function, I considered it a improper name.
Because `wsample`makes me think of `random.sample`, which returns a list of randomly generated elements, not a element.
> I came across a function named `wsample` in a `utils` package of my
> workplace recently.
> The "w" in `wsample` stands for `weighted`, and it randomly selects
> an element from a container according to relative weights of all the
> elements.
I agree that wt_select for weighted select might be better for a sample size of 1 ;-).
> In most articles and codes I saw online, a function like this is
> often named `weighted_random_choice`, which sounds *correct* to me.
Easier to understand, at least the first time, harder to write. Be glad the author did not use 'ws' as would have once been common.
> So when I saw this `wsample` function, I considered it a improper
> name. Because `wsample`makes me think of `random.sample`, which
> returns a list of randomly generated elements, not a element.
On Oct 11, 1:33 pm, Satoru Logic <satorulo...@gmail.com> wrote:
> I came across a function named `wsample` in a `utils` package
> of my workplace recently.
> The "w" in `wsample` stands for `weighted`, and it randomly
> selects an element from a container according to relative
> weights of all the elements.
> So when I saw this `wsample` function, I considered it a improper name.
I tend to agree. I like descriptive names for functions that don't
assume familiarity. If you're unhappy with a function's name and
you're unable to modify the module it belongs to, you can always
rebind it yourself:
from utils import wsample as weighted_random_choice
If I'm then going to use that function heavily elsewhere, I _might_
rebind it, but always within the scope that is going to use the
abbreviated binding:
wrc = weighted_random_choice
<lots of lines of code using wrc>
On Wed, 10 Oct 2012 20:33:16 -0700, Satoru Logic wrote:
> I came across a function named `wsample` in a `utils` package of my
> workplace recently.
> The "w" in `wsample` stands for `weighted`, and it randomly selects an
> element from a container according to relative weights of all the
> elements.
> In most articles and codes I saw online, a function like this is often
> named `weighted_random_choice`, which sounds *correct* to me. So when I
> saw this `wsample` function, I considered it a improper name. Because
> `wsample`makes me think of `random.sample`, which returns a list of
> randomly generated elements, not a element.
You can have a sample size of one.
wsample sounds fine to me. weighted_random_choice is okay too. It depends whether you value brevity over explicitness. Explicit is good, but some names are just too long.
If this were my code base, I would probably go for weighted_sample without mentioning "random" in the name, the reasoning being that samples are almost always random so explicitly saying so doesn't help much.
On Thursday, October 11, 2012 2:29:37 PM UTC+8, Steven D'Aprano wrote:
> On Wed, 10 Oct 2012 20:33:16 -0700, Satoru Logic wrote:
> > I came across a function named `wsample` in a `utils` package of my
> > workplace recently.
> > The "w" in `wsample` stands for `weighted`, and it randomly selects an
> > element from a container according to relative weights of all the
> > elements.
> > In most articles and codes I saw online, a function like this is often
> > named `weighted_random_choice`, which sounds *correct* to me. So when I
> > saw this `wsample` function, I considered it a improper name. Because
> > `wsample`makes me think of `random.sample`, which returns a list of
> > randomly generated elements, not a element.
> You can have a sample size of one.
> wsample sounds fine to me. weighted_random_choice is okay too. It depends
> whether you value brevity over explicitness. Explicit is good, but some
> names are just too long.
> If this were my code base, I would probably go for weighted_sample
> without mentioning "random" in the name, the reasoning being that samples
> are almost always random so explicitly saying so doesn't help much.
I think if a programmer has used the built-in `random` module before, he would expect a function with "sample" in its name to return a population sequence.
If a function is to return scalar value instead of sequence, I would expect it to be named "choice".
On Wed, 10 Oct 2012 23:44:42 -0700, suzaku wrote:
> I think if a programmer has used the built-in `random` module before, he
> would expect a function with "sample" in its name to return a population
> sequence.
I have used the random module for about fifteen years, and I still write random.sample when I need to use random.choice.
In statistics, probability, and plain English, a sample can be a single item: that's why we can say "a sample" or "two samples".
> If a function is to return scalar value instead of sequence, I would
> expect it to be named "choice".
And I wouldn't. But what do I care? I'm never going to use the code you're talking about, so call it "sasquatch" if you like, it's no skin off my nose.
On Friday, October 12, 2012 10:22:16 AM UTC+8, Steven D'Aprano wrote:
> On Wed, 10 Oct 2012 23:44:42 -0700, suzaku wrote:
> > I think if a programmer has used the built-in `random` module before, he
> > would expect a function with "sample" in its name to return a population
> > sequence.
> I have used the random module for about fifteen years, and I still write
> random.sample when I need to use random.choice.
> In statistics, probability, and plain English, a sample can be a single
> item: that's why we can say "a sample" or "two samples".
Thanks for sharing your experience.
As I'm not a native speaker of English, when I learned that `random.choice` return single item,
and `random.sample` return a sequence of items, I thought that the behaviour is determined by their definitions.
