https://d.puremagic.com/issues/show_bug.cgi?id=8755



--- Comment #17 from monarchdodra@gmail.com 2014-03-21 01:51:41 PDT ---
(In reply to comment #14)
> >Furthermore, it also improves usability by making the seeds passed by parameter pack, instead of forcing the use of a tuple.<
> 
> OK. (Despite in a modern language tuples should be built-in, so using them should be natural, common, and syntactically very cheap. In Python/Haskell/Scala code you don't see functions that refrain from accepting a tuple).

The current most spread phobos style is to accept arguments via vararg, and to return tuples when you need to return several args, so I tried to stick to that.

I'm not sure it is possible to accept either a Tuple (that "auto expands" later), and varargs, without potentially creating some ambiguity.

> >Finally, it allows using only 1 seed, in which case, the same seed is replicated and is used for all the functions.
> 
> This is from the unittests:
> 
>     // Compute sum and sum of squares in one pass.
>     // This can be used to compute get the average and standard deviation.
>     // A single seed (0.0) is passed, but it is optional
>     // if the range is not empty.
>     r = a.fold!("a + b", "a + b * b")(0.0);
>     assert(approxEqual(r[0], 35));  // sum
>     assert(approxEqual(r[1], 233)); // sum of squares
> 
> This is ambiguous, it seems that "a + b" has a seed while "a + b * b" doesn't have a seed. So in my opinion if you give N function then you need to give 0 seeds, or one N-tuple, or N seeds. So I don't like this.

You think? It made sense to me. I'll have to ask for more input on this. That said, if we turn it down, then it would be possible to make `Tuple` and `args...` co-exist as input argument style. EG:

r = a.fold!("a + b", "a + b * b")(0.0, 0.0); //OK!
r = a.fold!("a + b", "a + b * b")(tuple(0.0, 0.0)); //OK! too!

The second one is more verbose, but I see its justified if your seed is already in the form of a tuple.

So I think you sold me on it.

> >Oh yeah, also, I made it so that when no seed is given, it is an Error to use an empty range. This is the only case of deviation, but I think having nothrow justifies it.
> 
> I am not sure this is a good idea. Throwing when you give no seed is probably acceptable. But I am not sure.

It's a deviation, but I think it's justified. It makes the code nothrow, and quite frankly, accessign an empty range is an Error, so end of story.

The only argument I'd accept in its favor is stability with reduce, but if we could redesign, I'd never accept throwing an exception in such a case.

> > "iterables" are not supported anymore.
>
> If by "iterables" you mean that fold doesn't accept opApply-based iterables then I am against this change, I have plenty of code that opApply-based and I sometimes use reduce on them.

OK. I can work them back in.

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

https://d.puremagic.com/issues/show_bug.cgi?id=8755



--- Comment #18 from bearophile_hugs@eml.cc 2014-03-21 03:46:15 PDT ---
(In reply to comment #17)

> it would be possible to make `Tuple` and
> `args...` co-exist as input argument style. EG:
> 
> r = a.fold!("a + b", "a + b * b")(0.0, 0.0); //OK!
> r = a.fold!("a + b", "a + b * b")(tuple(0.0, 0.0)); //OK! too!
> 
> The second one is more verbose, but I see its justified if your seed is already in the form of a tuple.

What I don't like is to give only 1 single scalar argument if you have N functions and then implicitly multiply the single seed N times. Similar implicit behaviours look handy, but they make the code less clear, and later usually they find some way to bite your rump.

On the other hand I think that it's uncommon to give more than one function to reduce/fold. So this whole sub-feature is could even be chopped away, simplifying fold.

How many real usages of multi-function reduce to you have in Phobos?

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

https://d.puremagic.com/issues/show_bug.cgi?id=8755



--- Comment #19 from bearophile_hugs@eml.cc 2014-03-21 04:03:46 PDT ---
See also Issue 10670

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

https://d.puremagic.com/issues/show_bug.cgi?id=8755



--- Comment #20 from monarchdodra@gmail.com 2014-03-21 05:07:58 PDT ---
(In reply to comment #18)
> What I don't like is to give only 1 single scalar argument if you have N functions and then implicitly multiply the single seed N times. Similar implicit behaviours look handy, but they make the code less clear, and later usually they find some way to bite your rump.

I've already agreed to that point, and will be removing the functionality. Furthermore, I doubt it provides any real feature: We never call reduce with so many functions that this would be justified.

> On the other hand I think that it's uncommon to give more than one function to reduce/fold. So this whole sub-feature is could even be chopped away, simplifying fold.

When you say "sub-feature", are you talking about about the "1-seed" thing, or having multiple functions/

If it's multiple functions, then I have to say that it's an opt-in feature, and it works anyways. It's not actually complicated to implement, and it buys us nothing to remove it. Why want to chop it? I've used it before. It's nice.

If it's the single seed for multiple functions, yeah.

> How many real usages of multi-function reduce to you have in Phobos?

I'm not sure it's a fair reference, because Phobos doesn't "do" anything, it just provides function. I'd be willing to bet there is a fair amount of it being used out there.

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

https://d.puremagic.com/issues/show_bug.cgi?id=8755



--- Comment #21 from bearophile_hugs@eml.cc 2014-03-21 05:32:04 PDT ---
(In reply to comment #20)

> I'm not sure it's a fair reference, because Phobos doesn't "do" anything, it just provides function. I'd be willing to bet there is a fair amount of it being used out there.

OK. One usage case I can think of is to compute the sum and count of an input range, to compute its average :-)

-- 
Configure issuemail: https://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------