I'm reading http://www.haskell.org/haskellwiki/IO_inside#inlinePerformIO and it shows a passage of code: write :: Int -> (Ptr Word8 -> IO ()) -> Put () write !n body = Put $ \c buf@(Buffer fp o u l) -> if n <= l then write' c fp o u l else write' (flushOld c n fp o u) (newBuffer c n) 0 0 0 where {-# NOINLINE write' #-} write' c !fp !o !u !l = -- warning: this is a tad hardcore inlinePerformIO (withForeignPtr fp (\p -> body $! (p `plusPtr` (o+u)))) `seq` c () (Buffer fp o (u+n) (l-n)) I got as far as the second line, looking things up in this index http://hackage.haskell.org/package/base-4.6.0.1/docs/doc-index.html But “write !n body = ⋯” I understand write is defined to be a function taking an Int and another function, but what does !n mean? I went through the 2010 Report (BTW, the PDF is useless for searching for the ! character so I used the HTML version page-by-page) and found it used as a modifier for named record fields (it says "strict" but I think it's describing non-optional). Then I went through the GHC users guide for language extensions, and found a reference in §7.2.1 without explanation but as an example where the difference between f x = let (Foo a b, w) = ..rhs.. in ..body.. and f x = let !(Foo a b, w) = ..rhs.. in ..body.. is "you must make any such pattern-match strict". Strict pattern matching is not mentioned elsewhere nor is it in the Haskell Report, so what am I missing? I suspect that the usage I'm asking about is related. The GHC users guide also mentions it again in §7.4.6 which I think is a reference to the same feature, “You can use strictness annotations, in the obvious places in the constructor type” but that's for use with a completely different extension (GADT types) I also recognize the “@” mark as naming the entire variable rather than just the parts of the pattern matched, but “buf” is not actually used anywhere, so does it mean something different, or has other effects, or what? Thanks in advance, —John
On 27/03/14 01:15, John M. Dlugosz wrote:
I'm reading http://www.haskell.org/haskellwiki/IO_inside#inlinePerformIO and it shows a passage of code:
write :: Int -> (Ptr Word8 -> IO ()) -> Put () write !n body = Put $ \c buf@(Buffer fp o u l) -> if n <= l then write' c fp o u l else write' (flushOld c n fp o u) (newBuffer c n) 0 0 0
where {-# NOINLINE write' #-} write' c !fp !o !u !l = -- warning: this is a tad hardcore inlinePerformIO (withForeignPtr fp (\p -> body $! (p `plusPtr` (o+u)))) `seq` c () (Buffer fp o (u+n) (l-n))
I got as far as the second line, looking things up in this index http://hackage.haskell.org/package/base-4.6.0.1/docs/doc-index.html
But “write !n body = ⋯” I understand write is defined to be a function taking an Int and another function, but what does !n mean? I went through the 2010 Report (BTW, the PDF is useless for searching for the ! character so I used the HTML version page-by-page) and found it used as a modifier for named record fields (it says "strict" but I think it's describing non-optional). Then I went through the GHC users guide for language extensions, and found a reference in §7.2.1 without explanation but as an example where the difference between f x = let (Foo a b, w) = ..rhs.. in ..body.. and f x = let !(Foo a b, w) = ..rhs.. in ..body.. is "you must make any such pattern-match strict". Strict pattern matching is not mentioned elsewhere nor is it in the Haskell Report, so what am I missing? I suspect that the usage I'm asking about is related.
The GHC users guide also mentions it again in §7.4.6 which I think is a reference to the same feature, “You can use strictness annotations, in the obvious places in the constructor type” but that's for use with a completely different extension (GADT types)
It is in fact a strictness annotation which forces ‘n’ when matching on it. It is an optimisation technique and if you're just starting out, I would not worry about it too much. I would advise against putting those on if you're not sure about the implications as they can make your performance worse. GHC can in fact often add those itself. I don't have any official reading for you but if you simply search for ‘Haskell strictness’ online, you should get quite a few links to Haskell wiki, the Wikibook, Stack Overflow and some blogs.
I also recognize the “@” mark as naming the entire variable rather than just the parts of the pattern matched, but “buf” is not actually used anywhere, so does it mean something different, or has other effects, or what?
It just seems that someone gave it a name and then didn't use it. It does not have any extra meaning.
Thanks in advance, —John
_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners
-- Mateusz K.
On Wed, Mar 26, 2014 at 08:15:07PM -0500, John M. Dlugosz wrote:
I'm reading http://www.haskell.org/haskellwiki/IO_inside#inlinePerformIO and it shows a passage of code:
write :: Int -> (Ptr Word8 -> IO ()) -> Put () write !n body = Put $ \c buf@(Buffer fp o u l) -> if n <= l then write' c fp o u l else write' (flushOld c n fp o u) (newBuffer c n) 0 0 0
where {-# NOINLINE write' #-} write' c !fp !o !u !l = -- warning: this is a tad hardcore inlinePerformIO (withForeignPtr fp (\p -> body $! (p `plusPtr` (o+u)))) `seq` c () (Buffer fp o (u+n) (l-n))
I got as far as the second line, looking things up in this index http://hackage.haskell.org/package/base-4.6.0.1/docs/doc-index.html
But “write !n body = ⋯” I understand write is defined to be a function taking an Int and another function, but what does !n mean? I went through the 2010 Report (BTW, the PDF is useless for searching for the ! character so I used the HTML version page-by-page) and found it used as a modifier for named record fields (it says "strict" but I think it's describing non-optional).
No, strict does not mean optional vs. non-optional. It has to do with lazy evaluation. In Haskell, by default, expressions are not evaluated until their value is actually needed. Making something strict means (to a first approximation) forcing it to be evaluated right away, whether its value will be needed or not. That is what all the exclamation points mean, both in the code you cited above and the examples you found; they can be used in a bunch of different contexts but they always have to do with strictness. Incidentally, that "IO inside" page is not particularly what I would recommend for beginners to read! It's like learning about spark plugs and oil filters on your first day of driving school.
I also recognize the “@” mark as naming the entire variable rather than just the parts of the pattern matched, but “buf” is not actually used anywhere, so does it mean something different, or has other effects, or what?
Nope, you're right, it's simply not used. Probably just an oversight on the part of whoever wrote the code. the buf@ could be deleted and it wouldn't change anything. -Brent
http://www.haskell.org/ghc/docs/7.6.3/html/users_guide/bang-patterns.html
Do you mean this?
On Thu, Mar 27, 2014 at 9:15 AM, John M. Dlugosz
I'm reading http://www.haskell.org/haskellwiki/IO_inside#inlinePerformIOand it shows a passage of code:
write :: Int -> (Ptr Word8 -> IO ()) -> Put () write !n body = Put $ \c buf@(Buffer fp o u l) -> if n <= l then write' c fp o u l else write' (flushOld c n fp o u) (newBuffer c n) 0 0 0
where {-# NOINLINE write' #-} write' c !fp !o !u !l = -- warning: this is a tad hardcore inlinePerformIO (withForeignPtr fp (\p -> body $! (p `plusPtr` (o+u)))) `seq` c () (Buffer fp o (u+n) (l-n))
I got as far as the second line, looking things up in this index < http://hackage.haskell.org/package/base-4.6.0.1/docs/doc-index.html>
But “write !n body = ⋯” I understand write is defined to be a function taking an Int and another function, but what does !n mean? I went through the 2010 Report (BTW, the PDF is useless for searching for the ! character so I used the HTML version page-by-page) and found it used as a modifier for named record fields (it says "strict" but I think it's describing non-optional). Then I went through the GHC users guide for language extensions, and found a reference in §7.2.1 without explanation but as an example where the difference between f x = let (Foo a b, w) = ..rhs.. in ..body.. and f x = let !(Foo a b, w) = ..rhs.. in ..body.. is "you must make any such pattern-match strict". Strict pattern matching is not mentioned elsewhere nor is it in the Haskell Report, so what am I missing? I suspect that the usage I'm asking about is related.
The GHC users guide also mentions it again in §7.4.6 which I think is a reference to the same feature, “You can use strictness annotations, in the obvious places in the constructor type” but that's for use with a completely different extension (GADT types)
I also recognize the “@” mark as naming the entire variable rather than just the parts of the pattern matched, but “buf” is not actually used anywhere, so does it mean something different, or has other effects, or what?
Thanks in advance, —John
_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners
On 3/26/2014 8:26 PM, yi lu wrote:
http://www.haskell.org/ghc/docs/7.6.3/html/users_guide/bang-patterns.html
Do you mean this?
Thanks! I missed it when scanning through the GHC docs.
participants (4)
-
Brent Yorgey -
John M. Dlugosz -
Mateusz Kowalczyk -
yi lu