Re: my latest blog post

[Ricardo Wurmus <rekado@elephly.net>]

Hi Catonano,

> 2018-06-09 12:39 GMT+02:00 Ricardo Wurmus <rekado@elephly.net>:
>
>>
>> Catonano <catonano@gmail.com> writes:
>>
>> > Assuming that users read academic articles about programming languages in
>> > order to know your way aroung Guile is not reasonable
>>
>> I fail to see how this follows from Guile’s lack of a macro stepper.
>
>
> Then you were not following
>
> Mark indicated me a paper about the first macro stepper in history

I did follow the discussion and I did see Mark mentioning the paper.
Since Guile does not *have* a macro stepper, reading a paper about how a
macro stepper was implemented is certainly not a requirement to use
Guile.  Obviously, to implement a macro stepper it is useful to know how
to do this.  But implementing a macro stepper is not what a Guile user
like myself would ever do.

>> And I have
>> never felt the need to read computer science papers to “know my way
>> around Guile”.  (I have not studied computer science, so it isn’t that
>> my background allowed me to skip papers that are mandatory for other
>> users of Guile.)
>>
>> The sentence above seems like an exaggeration to me.
>>
>
> Ok, thans for your contribution
>
>
>> (I say this with no hostility, just with surprise.)
>>
>
> we have a loooong way to go

I sense a lot of hostility in your responses to me and I don’t know why
that is.  The first sentence reads like unwarranted sarcasm to me,
because it does not seem to relate to what I wrote; the second sounds
belittling, as if my surprise at what seems like an exaggeration to me
is evidence of how far behind I am in my efforts to ensure that Guix and
Guile a welcoming to newcomers.

I’m sorry that you don’t find anything useful in my response above,
which relates my personal experience with Guile as someone who is not a
computer scientist.  I hoped it would be a useful data point.

As I find discussions like this exceptionally draining and discouraging,
detrimental to my ability to volunteer time to free software, I will
take some time off from this discussion after trying to respond to your
other message to me in this thread.

>> > The monad accessing the daemon, how would I delve in it ?
>>
>> The Guix manual should explain it fully, while assuming that the concept
>> of a monad is known (because that’s not a Guix or Guile invention).
>
>
> And how would I know what a monad is without reading academic materials ?
>
> Maybe Haskell is a requirement in order to use Guile//Guix ?
>
> "assuming that the concept of a monad is known" is a problem.

By analogy, do you think that assuming the concept of recursion would be
a problem?  Or the concept of a closure?  Would you say that this should
be explained in the *Guix* manual?  (These are sincere questions.)

We are using both of these constructs.  We are also using delayed
evaluation in the form of streams.  Luckily, those are provided as a
Guile library and are thus documented in the *Guile* manual.

> Someone would want to _learn_ what a monad is AND how it can be implemented
> in scheme, by dissecting this Guix macro based feature

I used “,expand” in the past to see the expansion of a macro
expression.  Here’s an example with the state monad:

--8<---------------cut here---------------start------------->8---
scheme@(guile-user)> ,expand (with-monad %state-monad (return 1))
$3 = (let ((value 1))
  (lambda (state)
    ((@@ (guix monads) values) value state)))
--8<---------------cut here---------------end--------------->8---

This shows us that returning a monadic value in the state monad is the
same as returning a procedure that takes some state and returns it
alongside the value.

The Guix manual says about the things that we used:

--8<---------------cut here---------------start------------->8---
 -- Scheme Syntax: with-monad MONAD BODY ...
     Evaluate any ‘>>=’ or ‘return’ forms in BODY as being in MONAD.

 -- Scheme Syntax: return VAL
     Return a monadic value that encapsulates VAL.
--8<---------------cut here---------------end--------------->8---


Earlier, it also says something about what monads are all about:

--8<---------------cut here---------------start------------->8---
   This is where the ‘(guix monads)’ module comes in.  This module
provides a framework for working with “monads”, and a particularly
useful monad for our uses, the “store monad”.  Monads are a construct
that allows two things: associating “context” with values (in our case,
the context is the store), and building sequences of computations (here
computations include accesses to the store).  Values in a monad—values
that carry this additional context—are called “monadic values”;
procedures that return such values are called “monadic procedures”.
--8<---------------cut here---------------end--------------->8---

We’ve just seen the first part about associating “context” (in this case
some state) with values.  The second part requires a larger expression
to see how this is useful.

The Guix manual has an example right there:

--8<---------------cut here---------------start------------->8---
 -- Scheme Syntax: >>= MVAL MPROC ...
     “Bind” monadic value MVAL, passing its “contents” to monadic
     procedures MPROC...(1).  There can be one MPROC or several of them,
     as in this example:

          (run-with-state
              (with-monad %state-monad
                (>>= (return 1)
                     (lambda (x) (return (+ 1 x)))
                     (lambda (x) (return (* 2 x)))))
            'some-state)

          ⇒ 4
          ⇒ some-state
--8<---------------cut here---------------end--------------->8---

So this is an expression that somewhat uselessly computes the number 4
in a context where “some-state” is present.  We can use ,expand in the
REPL to see how this works, this time with a slightly simpler
expression that would compute 1 + 1 in some given context:

--8<---------------cut here---------------start------------->8---
scheme@(guile-user)> ,expand (with-monad %state-monad
                               (>>= (return 1)
                                    (lambda (x) (return (+ 1 x)))))
$6 = (let ((mvalue
        (let ((value 1))
          (lambda (state)
            ((@@ (guix monads) values) value state))))
      (mproc (lambda (x)
               (let ((value (+ 1 x)))
                 (lambda (state)
                   ((@@ (guix monads) values) value state))))))
  "Bind MVALUE, a value in the state monad, and pass it to MPROC."
  (lambda (state)
    ((@@ (guix monads) call-with-values)
     (lambda () (mvalue state))
     (lambda (value state) ((mproc value) state)))))
--8<---------------cut here---------------end--------------->8---

This looks similar to the previous expression (see “mvalue”).  Looking
at the body, we see that this is a procedure that takes some state,
computes a new value by passing the state to “mvalue”, and then passes
that to the next action (the last lambda in the original expression).

I don’t think that the presence of the relevant documentation in the
Guix manual and the existence of a macro-expansion facility should be
taken as evidence that we “think that the code should be kept obscure
and unaccessible”.

I would also like to state, though, that the use of the state monad is
hardly essential for the purpose of extending or contributing to Guix.
If you find a way to improve the existing documentation, you know from
our past interactions that the maintainers and developers are very
welcoming of additions or changes to the manual.

--
Ricardo