As a cosmic coincidence this happened to me yesterday. My goal: generate
n-long arrays of 1s and -1s picked from a uniform distribution to mimic
BLAS function ?LARNV.
Full disclosure: I know a bit about the subject, but I'm not an expert and
too lazy to gather my thoughts about it at the moment because I think I'll
need "choice" so I just want to end up there quickly.
- So I went to the docs, and picked numpy.random from the left menu. It
started with some paragraphs, ok something somehing bitgenerator ok fine
interesting.
- Started scrolling down, OK old vs new hmm what is RandomState...
nevermind replaced with something new; where is the numpy.random.rand
stuff? ... nevermind scroll more
- OK some more stuff about what's new... then a try except block. None of
this seems familiar
- Introduction section !? in the middle of the page. OK am I in the correct
page?
- Things are getting tough now I'm in the PCG64 something something... What
is all stuff about?
- OK intro is over then a section "What's new or different"... I am not
introduced sufficiently yet...
- Giving up and scrolling all the way down and up a few more
- OK go back to google, search for "numpy choice" end up in
numpy.random.choice.
- There is a warning ... it says this is old stuff see QuickStart
- click and we are back :)
This is really not a complaint, I think there is an occupational hazard
about random stuff throughout the internet. An insatiable urge to teach
people about pseudorandom processes but I don't think it is the right way
to go about it. Unless you are knee deep in this area almost none of this
matters to an average user. We had this discussion before over SciPy about
setting a seed as easy as the good ol' "numpy.random.seed" but all I am
getting is extensive details about how the engine works. I appreciate the
state-of-art implementation of any method but it can be saved for the
interested who wants to geek about it. Otherwise it would be really
beneficial if quick start is indeed quick. Currently it is a bit unquick :)
Best,
ilhan
On Fri, Oct 15, 2021 at 12:13 PM Kevin Sheppard <kevin.k.shepp...@gmail.com>
wrote:
>
>
> On Thu, Oct 14, 2021 at 7:22 PM Ralf Gommers <ralf.gomm...@gmail.com>
> wrote:
>
>>
>>
>> On Thu, Oct 14, 2021 at 7:19 PM Kevin Sheppard <
>> kevin.k.shepp...@gmail.com> wrote:
>>
>>> I think the issue in random specifically is that a raw list of
>>> available functions does not provide suitable guidance for someone looking
>>> for random variate generating function. This is because the module-level
>>> API is mostly dominated by methods of the singleton RandomState instance.
>>> Best practice going forward is to use the methods of a Generator instance,
>>> most likely provided by default_rng(). A simple API-list will not be able
>>> to provide this guidance.
>>>
>>
>> The list can be annotated with headings and one-line or one-paragraph
>> descriptions, something like:
>>
>> ```
>> ## Generator interface
>>
>> This is the recommended interface ... <etc - list methods too>
>>
>> ## Interface for unit testing and legacy code
>>
>> <explain purpose, then list all routines>
>> ```
>>
>> The complaint is very much valid here, I have made the same one before.
>> The way the page currently is written makes little sense - it addresses a
>> user transitioning from the old to the new interface, or explicitly
>> comparing the two for some reason. To a user just looking for information
>> on NumPy today, that's more confusing than helpful.
>>
>> The page also talks about "The new interface", "What's new and
>> different", "Some long-overdue API cleanup", and "Since Numpy version
>> 1.17.0" - that all belongs in a NEP, and not in the API reference docs.
>>
>> Cheers,
>> Ralf
>>
>>
>>
> I don't think the doc style there is ideal. I would still say that a
> relatively naive dump of `np.random` (something that would be everything in
> [v for v in dir(np.random) if not v.startswith("_")] would not lead to an
> idea set of docs because most of the "obvious" functions are methods of the
> legacy RandomState. A good set would need something like (excluding
> headers)
>
> default_rng
> Generator
> Generator.random
> Generator.integers
> ...
>
> <much lower, or IMO better in a a sperate page>
> Legacy Methods
> ==============
> <short explainer>
> np.random.random_sample
> np.random.randint
> RandmState
> ...
>
> IMO many (likely most) methods exposed in np.random should not be on the
> default landing page for np.random.
>
> Best,
> Kevin
>
>
>
>>> FFT has a very simple API and so a simple list make sense. Similarly,
>>> np.random before the generation was revamped, which is hy the old-style was
>>> adequate for <=1.16, but not for >=1.17
>>>
>>> Kevin
>>>
>>>
>>> On Thu, Oct 14, 2021 at 6:09 PM Paul M. <pmma...@gmail.com> wrote:
>>>
>>>> Hi Melissa,
>>>>
>>>> I think that's the right approach. Looking through the current docs, I
>>>> think the page on the FFT module is exemplary in this regard:
>>>>
>>>> https://numpy.org/doc/stable/reference/routines.fft.html
>>>>
>>>> It lists all the available functions (with links to details), and then
>>>> has a section on "Background Information", "Implementation Details", etc.
>>>> It's easy to get a quick overview of what the available functions are, and
>>>> then ease into the background info in terms of how it works.
>>>>
>>>> Cheers,
>>>> Paul
>>>>
>>>>
>>>> On Thu, Oct 14, 2021 at 12:44 PM Melissa Mendonça <meliss...@gmail.com>
>>>> wrote:
>>>>
>>>>> Hi Paul,
>>>>>
>>>>> Do you think having a page with the flat list of routines back, in
>>>>> addition to the explanations, would solve this?
>>>>>
>>>>> - Melissa
>>>>>
>>>>> On Thu, Oct 14, 2021 at 1:34 PM Paul M. <pmma...@gmail.com> wrote:
>>>>>
>>>>>> Hi All,
>>>>>>
>>>>>> The documentation of Numpy's submodules used to have a fairly
>>>>>> standard structure as shown here in the 1.16 documentation:
>>>>>>
>>>>>>
>>>>>> https://docs.scipy.org/doc/numpy-1.16.1/reference/routines.random.html
>>>>>>
>>>>>> Now the same page in the API documentation looks like this:
>>>>>>
>>>>>> https://numpy.org/doc/stable/reference/random/index.html
>>>>>>
>>>>>> While I appreciate the expository text in the new documentation about
>>>>>> how the generators work, this new version is much less useful as a
>>>>>> reference to the API. It seems like it might fit better in the user
>>>>>> manual
>>>>>> rather than the API reference.
>>>>>>
>>>>>> From my perspective it seems like the new version of the
>>>>>> documentation is harder to navigate in terms of finding information
>>>>>> quickly
>>>>>> (more scrolling, harder to get a bird's eye view of functions in various
>>>>>> submodules, etc).
>>>>>>
>>>>>> Has anyone else had a similar reaction to the changes? I teach a
>>>>>> couple of courses in scientific computing and bioinformatics and my
>>>>>> students seem to also struggle to get a sense of what the different
>>>>>> modules
>>>>>> offer based on the new version of the documentation. For now, I'm
>>>>>> referring
>>>>>> them to the old (1.70) reference manuals as a better way to get
>>>>>> acquainted
>>>>>> with the libraries.
>>>>>>
>>>>>> Cheers,
>>>>>> Paul Magwene
>>>>>> _______________________________________________
>>>>>> NumPy-Discussion mailing list -- numpy-discussion@python.org
>>>>>> To unsubscribe send an email to numpy-discussion-le...@python.org
>>>>>> https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
>>>>>> Member address: meliss...@gmail.com
>>>>>>
>>>>> _______________________________________________
>>>>> NumPy-Discussion mailing list -- numpy-discussion@python.org
>>>>> To unsubscribe send an email to numpy-discussion-le...@python.org
>>>>> https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
>>>>> Member address: pmma...@gmail.com
>>>>>
>>>> _______________________________________________
>>>> NumPy-Discussion mailing list -- numpy-discussion@python.org
>>>> To unsubscribe send an email to numpy-discussion-le...@python.org
>>>> https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
>>>> Member address: kevin.k.shepp...@gmail.com
>>>>
>>> _______________________________________________
>>> NumPy-Discussion mailing list -- numpy-discussion@python.org
>>> To unsubscribe send an email to numpy-discussion-le...@python.org
>>> https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
>>> Member address: ralf.gomm...@googlemail.com
>>>
>> _______________________________________________
>> NumPy-Discussion mailing list -- numpy-discussion@python.org
>> To unsubscribe send an email to numpy-discussion-le...@python.org
>> https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
>> Member address: kevin.k.shepp...@gmail.com
>>
> _______________________________________________
> NumPy-Discussion mailing list -- numpy-discussion@python.org
> To unsubscribe send an email to numpy-discussion-le...@python.org
> https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
> Member address: ilhanpo...@gmail.com
>
_______________________________________________
NumPy-Discussion mailing list -- numpy-discussion@python.org
To unsubscribe send an email to numpy-discussion-le...@python.org
https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
Member address: arch...@mail-archive.com
