Compact Normal Form and ST

Andrew Martin andrew.thaddeus at
Mon Jun 19 13:06:39 UTC 2017

I too am struggling to find a scenario in which this causes something to
go wrong. Ending up with only one block instead of two copies of it
seems like it could only be a problem, as Edward says, "if you were
planning to use these blocks as separate allocation buffers for
subsequent modifications". But, mutable byte arrays allocated by
newByteArray# cannot escape runST, so I don't see how this could happen.
Well, I guess if you returned a frozen array, and then you thawed it
for subsequent modification outside of runST, that would be problematic,
but CSE already makes that unsafe, even without involving compact
regions. It's good to know that, for my purposes, I'm in the clear, but
I would also like to hear from Edward clarifying the original statement.

-Andrew Martin

On Mon, Jun 19, 2017 at 4:41 AM, David Feuer <david.feuer at> wrote:

> Can you explain how things could go wrong in ST, perhaps with an example?
> It's hard to see the potential problem.
> On Jun 18, 2017 11:49 PM, "Edward Z. Yang" <ezyang at> wrote:
> There are two senses of referential transparency here which should be
> considered.  First is whether or not you will get the same value results
> if you use the compact functionality in ST.  Here, the answer is yes.
> Compact normal form has very trivial semantics in this domain, and
> it would have been OK even to make compact normal forms be pure
> functions.
> Second is whether or not the performance characteristics are preserved.
> Here, the situation is different.  Most notably, pure expressions and
> invocations of the same runST block may be commoned up (via an
> optimization pass like CSE.)  In that case, what was previously two
> separate compact blocks may be commoned up into a single one.  This
> could be disaster if you were planning to use these blocks as separate
> allocation buffers for subsequent modifications.
> This motivated specializing compact to IO.  It won't segfault if you
> put it in ST, but the performance characteristics might change.
> Edward
> Excerpts from Andrew Martin's message of 2017-06-18 10:24:09 -0400:
> > In the primops file where the compact normal form functions are
> > documented (
> d94f79aaac194/compiler/prelude/primops.txt.pp#L2487),
> > I noticed that all of the functions have type signatures that constrain
> > them to only being used in IO. For example:
> >
> >     compactAdd# :: Compact# -> a -> State# RealWorld -> (# State#
> RealWorld, a #)
> >
> > I would like to know if generalizing these to allow them to work in ST
> would
> > be sound. That is, changing the type signature to:
> >
> >     compactAdd# :: Compact# -> a -> State# s -> (# State# s, a #)
> >
> > I'm not requesting that this change actually be made. I only want to
> > know if using unsafeCoerce to create the second function for my own
> > project would actually be sound.
> >
> > For those interested in knowing why I want this, it's because there are
> > situation where I'm interested in building up a giant structure in a
> > compact region, but in a way that doesn't actually require IO. I think
> > it's a pity to have to use a type signature with IO and then call
> > unsafePerformIO at the end instead of using the more constrained ST,
> > and runST, which makes it clear that I'm not doing anything observable
> > form the outside.
> >
> > As a minor bonus, the ST version of the Compact data type shouldn't need
> > the lock that the IO version does, since concurrent calls to compactAdd
> > are not possible.
> >
> > -Andrew Martin
> >
> _______________________________________________
> Libraries mailing list
> Libraries at

-Andrew Thaddeus Martin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the Libraries mailing list