Re: how is Haskell not robust?

On Fri, 05 Jun 2009 09:11:21 +0200, Ertugrul Söylemez <es@xxxxxxxx>
wrote:

Jon Harrop schrieb:
i think i get it that laziness can lead to big space usage. and i
guess there is some debate about how easy it is for programmers to see
"oh this code right here could lead to a space leak!"

There is no debate. I asked Simon Peyton-Jones about this in person and he
confirmed that unpredictable memory consumption is one of the biggest
practical issues with Haskell. Ertugrul claimed he knew better but proved
himself wrong.

SPJ is certainly more knowledgable than I am, but in this regard I
disagree (if he really confirmed that in this form). What is a problem
in Haskell is 'unpredicted' space behaviour, because you didn't consider
the non-strict semantics, but not "unpredictable" space behaviour. If
it was such "unpredictable" as you claim, I'd have a very hard time
writing Haskell code.

You can write code forever without seeing a problem iff you never run
the programs with data sets that leak enough to break the system.


Also even if a space leak occurs, what's wrong with that? In most cases
the cause is obvious, so just correct your code.

In some cases it is obvious. In some it is not.


If you have done that a few times, at some point you simply 'get it'
and don't produce such leaking code anymore.

The problem is that even the experts still don't 'get it'.

George
.

Relevant Pages

  • Re: how is Haskell not robust?
    ... there is some debate about how easy it is for programmers to see "oh ... this code right here could lead to a space leak!" ... There is no debate. ... Since I like lazy data structutes (IMHO, they are a nice generalization ...
    (comp.lang.functional)
  • Re: how is Haskell not robust?
    ... raould wrote: ... guess there is some debate about how easy it is for programmers to see ... "oh this code right here could lead to a space leak!" ... There is no debate. ...
    (comp.lang.functional)
  • Re: Has anbody ever tried Leak Proof fork seals?
    ... "Debate" sounds more like an invitation to argue, but, when I read ... And I posted Leak Proof's claims to see if anybody could argue ... their ideas about fork seals. ... I wanted to hear from people who had actually used Leak Proof seals. ...
    (rec.motorcycles.tech)
  • Re: (OT) Re: "claim", etc. (was Re: C Standard Regarding Null Pointer Dereferencing)
    ... something I really want to do, for a variety of reasons, not all ... that the discussion started as a public debate and should finish ... This presumes that as corporate programmers you know your jobs. ... There is a Klown named Kiki ...
    (comp.lang.c)
  • Re: mmap-related memory leak using iconv_open/close?
    ... > one of our programmers... ... it might be interesting to take a tusc system call trace of the ... The glass is neither half-empty nor half-full. ... The glass has a leak. ...
    (comp.sys.hp.hpux)