lua-users home
lua-l archive

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index]




On Monday, April 7, 2014, Dirk Laurie <dirk.laurie@gmail.com> wrote:
2014-04-07 21:51 GMT+02:00 Hisham <h@hisham.hm>:
> On 7 April 2014 16:38, Dirk Laurie <dirk.laurie@gmail.com> wrote:
>> There has been more negative reaction, in stronger language too,
>> to every point raised in favour of keeping the functions, from PUC-Rio
>> than from any of the users elsewhere. The one point that is hard to
>> argue against, that the saving in executable size is less that 0.1%,
>> has been ignored.
>
> Countless other features have also been proposed in the past that
> would amount to 10-line patches and they have been solemnly ignored by
> the Lua team (including personal favorites of mine which I shall not
> name here lest this thread starts drifting off again). That's why _I_
> didn't even touch the "executable size" argument; I suppose that when
> they proposed the removal, they were already aware of how much code it
> was.

There is a substantial difference between not adding a proposed feature
and actually removing an existing feature of approximately the same size.
In the first case, you are merely disappointing the expectations that some
optimists had. In the second, you are running the risk of breaking something.

I'm not saying that breaking existing programs is something to be avoided
at all costs. When an existing feature cannot coexist peacefully with a new
and better feature, of course it should be removed with some regret.

But removing sinh, cosh and tanh (and not even with regret) in no way
improves Lua except the insignificant savings in size and the service
paid to the ideological goal of smallness. Is that really worth the risk of
breaking something?

Is this a real question or an _expression_ of frustration? 

In my own work, when I'm designing an API and find that the design doesn't look like the interface, I might refactor and often for no good reason other than it doesn't "feel right." I'm not suggesting, even a little, that this is the reason here.

But the value of design and how one might prioritize it _can_ come down to a very personal decision.

I read this thread and I keep concluding that "math" is a library that stands out as being especially "application" and not "language" oriented. That is: it supports peoples ability to apply lua to a set of problems, not help it generally work within a system. 

It's possible that this is the source of the debate.

The maxim goes: "If it ain't broke, don't fix it."
Not: "If it ain't broke, then break it."