Peter,

I applaud your excellent assistance with Raku.  Et. al. (you know the names)...

Outstanding community!

Mark

-----Original Message-----
From: Peter Pentchev <r...@ringlet.net> 
Sent: Monday, June 8, 2020 17:13
To: perl6-us...@perl.org
Subject: Re: question about the multi in method

On Mon, Jun 08, 2020 at 01:28:34PM -0700, ToddAndMargo via perl6-users wrote:
> On 2020-06-08 02:45, Richard Hainsworth wrote:
> > Ok Todd, let me have a go at this issue.
> > 
> >  From what I understand, you see 'multi' and think 'there are more 
> > than one', which leads to the question 'where are they?'
> > 
> > My understanding of 'multi' is 'there COULD be more than one', which 
> > leads to the question 'are there any?'
> > 
> > This is actually a very powerful aspect of Raku.
> > 
> > There are (as has been stated in this thread) four types of which  
> > multi = 'could be more than one' and only = 'only one' .
> > 
> > If the developer wrote 'only sub' then, no one else could write a 
> > sub with that name. For a developer of a program or a module, that 
> > might be OK, because it's unlikely anyone else would need/want to use that 
> > name.
> > But for a compiler, the probability is reversed. Someone will want 
> > to use it. That is why in many other languages there are 'reserved words'
> > which you are not allowed to use.
> > 
> > In Raku, if you develop a new type, for example say a quodion, and 
> > you want to 'add' quodions, then you CAN use + because it can be overloaded.
> > It is defined as a multi.
> > 
> > Or in your most overused example 'needle' can be rewritten to take a 
> > quodion, which may yield a spectrum of results because that's what 
> > quodions are for.
> > 
> > Since needle is defined as a 'multi', if you write a 'needle' method 
> > for quodions, the needle for strings STILL exists, and Raku will 
> > correctly call the method you wrote for quodions if a quodion is 
> > implied, but call the default needle if a string (or any other type) is 
> > used.
> > 
> > Since you have developed the quodion method entirely for your paid 
> > subscribers, the module is NOT available to the public, so you wont 
> > tell anyone about it. So it doesn't have to be in the documentation.
> > 
> > Is this making any sense to you?
> > 
> > Richard
> 
> 
> Hi Richard,
> 
> You are perfectly clear and beautifully well written.
> You did make me look up "quodion".
> 
> :-)
> 
> And I am feeling guilty because I did already know all this and it 
> looks like you spent a considerable amount of time getting into my 
> head to formulate an answer that was perfect for me.
> 
> Here is the thing.  I have no objecting to the developer calling 
> something a "multi" or an "only" depending on his needs.  That is 
> entirely "HIS" call, "NOT" mine.
> 
> And I hope the above statement ends all the well meaning, helpful 
> folks who are trying the re-explain the concept to me.  I got it a 
> LONG time ago.
> 
> My objection is when there actually exists more than one and they are 
> not experimental or otherwise hidden for some reason and are truly, 
> actually part of the distribution, then they also should be 
> documented.
> Updates to the documentation should be part of the code check in 
> process, and in all probability, are.
> 
> So, basically what I am saying is when multiples are there and are 
> open to the public, than it should be documented.  In this situation, 
> why would the developer want his hard work hidden?

OK, let me try to give you another example. If I'm too long-winded, there will 
be an abstract at the end :)

Say somebody is writing a library that should serve as the standard library for 
the C language. Now, there is a published, adopted standard for the C language 
that all C compilers and C standard libraries follow[1].

Now, the C language standard says that there may be more than one library - one 
called "libc" as "the C library" providing pretty much all of the functions, 
and another one called "libm" as "the mathematics library" providing some 
additional functions for more-or-less precise mathematical operations[2]. The 
standard says that the C library must provide a function called printf() that 
prints out all kinds of things according to a specific format, and that the 
maths library must provide a function called trunc() that pretty much rounds a 
real number down to the nearest integer that is not larger than it (so 
trunc(3.0) would return 3, trunc(4.5) would return 4, and trunc(-3.5) would 
return -4).

OK... So somebody is writing a "libc" that must provide the printf() function, 
and they suddenly realize that, in order to be able to print out real numbers 
with, say, two digits after the decimal point, one of the easiest ways is to 
have a function that rounds a real number down.
So they write such a function, let's call it printf_trunc(), and they hide it, 
they make it internal to the "libc" library, so your programs can't see it. 
Right? Everything's fine, there is an internal function that nobody can access.

So the next thing they do is turn to the maths library, and they realize that 
they have to write a trunc() function, and the lightbulb comes on:
they have *already written* a function that rounds a real number down, but for 
various reasons that *only matter to the author of the library*, not to the 
users of the library, that function needs to live in the "libc" library, not in 
the "libm" library.

OK, but they do not want to do the work twice, they do not want to have the 
same code in both libraries, do they? So they want to make the
trunc() function in the maths library call the printf_trunc() function in the 
standard C library. But... but that means... that means that
printf_trunc() may no longer be internal - it *must* be exposed, at least to 
the maths library, and the way the C language works[3], this means that *any* 
program that uses the standard C library can now see the printf_trunc() 
function. That means that anybody who takes the time to look at the source of 
the library can figure out that there is such a function and, if they feel 
adventurous, they may actually call it in their own program.

But! and this is the important point. THEY DO NOT WANT TO.

The author of a program - not the author of the library! - but the author of a 
program that uses that library, that is you, that is me, this author *does not 
want* to call a function that is not defined in the C standard. This author - 
you and me - can guess that this function is probably an internal one used by 
some of the other libraries developed along with the standard C library. But 
they don't need to
guess: they KNOW, from reading the documentation, that this function is NOT 
part of the specification of the C language. Moreover, they KNOW, by reading 
the documentation of the "libc" library, that this function is NOT listed there.

So, even though you and I can see the printf_trunc() function, we never want to 
call it.

There is a very important reason for that: it is a function that is written 
solely for the needs of the maths library, and if tomorrow the next version of 
the maths library needs this function to operate in another way, maybe add 
another argument, maybe drop an argument because something is handled 
elsewhere, the C library's printf_trunc() function *will change*. And if you or 
I had used it in our programs, our programs would break.

Then, if we go to the author of the library and complain about using the
printf_trunc() function that is not in the specification of the C language and 
that is not listed in the documentation of the "libc"
library... the author would say, and I quote, "Huh?!" They would say, "Wait... 
*why* are you using this function? How do you know about... no, wait, nevermind 
how you know about it... *why* are you using a function that is neither part of 
the language specification, nor listed in the documentation of this 
implementation of the language?"

So, you see, the hard work of the author of the library is not hidden.
It is there for us to see in the printf() function from "libc" and in the 
trunc() function from "libm". Every time we use one of these well-known and 
well-documented functions, we make use of the hard work of the author. We do 
not need to call the internal function to do that.

So, to recap:
- there are two libraries, "libc" and "libm"
- "libc" has a function that wants to round numbers down
- "libm" must have a function that wants to round numbers down
- the common function to be used by both libraries has to be placed in
  "libc", but it needs to be visible so that "libm" can use it
- somebody curious enough can look into the sources of "libc" and see
  this function
- but this somebody must realize that this is an internal, undocumented
  function that serves some purpose for the combination of "libc" and
  "libm" and that is *not* supposed to be used by anything outside these
  two libraries
- so this somebody will shrug it off, say "hey, there's some kind of
  implementation detail, there's some kind of internal function, okay,
  forget about it"

Hope that helps!

G'luck,
Peter

[1] ...and to everyone who cringed at that statement... sorry! :)
    Let's go with that for now and push the pain away :)

[2] ...and to everyone who cringed at *that* statement... sorry! :)

[3] ...let's go with "the C language" here, not get into details about
    compilers and linkers and foreign functions and stuff

--
Peter Pentchev  r...@ringlet.net r...@debian.org p...@storpool.com
PGP key:        http://people.FreeBSD.org/~roam/roam.key.asc
Key fingerprint 2EE7 A7A5 17FC 124C F115  C354 651E EFB0 2527 DF13

Reply via email to