Re: [Haskell-cafe] Is it safe to use unsafePerformIO here?

17 Sep 2009

      Am Donnerstag 17 September 2009 21:07:28 schrieb Cristiano Paris:
...
On Tue, Sep 15, 2009 at 11:31 PM, Daniel Fischer
 wrote:
...
...
Yeah, you do *not* want the whole file to be read here, except above for
testing purposes.
That's not true. Sometimes I want to, sometimes don't.
The "for the case of sorting by metadata" was tacitly assumed :)
...
But I want to use the same code for reading files and exploit laziness
to avoid reading the body.
...
Still, ByteStrings are probably the better choice (if you want the body
and that can be large).
That's not a problem by now.
...
To avoid reading the body without unsafePerformIO:
readBit fn
   = Control.Exception.bracket (openFile fn ReadMode) hClose
         (\h -> do
               l <- hGetLine h
               let i = read l
               bdy <- hGetContents h
               return $ Bit i bdy)
Same problem with the "withFile"-version: nothing gets printed if I
try to print out the body: that's way I used seq.
Ah, yes. The file is closed too soon.
...
I'm starting to think that the only way to do this without using
unsafePerformIO is to have the body being an IO action: simply, under
Haskell assumption, that's not possible to write, because Haskell
enforce safety above all.
Well, what about

readBit fn = do
    txt <- readFile fn
    let (l,_:bdy) = span (/= '\n') txt
    return $ Bit (read l) bdy

?

With

main = do
    args <- getArgs
    let n = case args of
                (a:_) -> read a
                _ -> 1000
    bl <- mapM readBit ["file1.txt","file2.txt"]
    mapM_ (putStrLn . show . index) $ sortBy (comparing index) bl
    mapM_ (putStrLn . take 20 . drop n . body) bl

./cparis3 30 +RTS -sstderr
2
3
CCGGGCGCGGTGGCTCACGC
CCGGGCGCGGTGGCTCACGC
         408,320 bytes allocated in the heap
           1,220 bytes copied during GC
          34,440 bytes maximum residency (1 sample(s))
          31,096 bytes maximum slop
               1 MB total memory in use (0 MB lost due to fragmentation)

./cparis3 20000 +RTS -sstderr                                             
2                                                                         
3                                                                         
AAAATTAGCCGGGCGTGGTG                                                      
AAAATTAGCCGGGCGTGGTG                                                      
       1,069,168 bytes allocated in the heap                              
         105,700 bytes copied during GC                                   
         137,356 bytes maximum residency (1 sample(s))                    
          27,344 bytes maximum slop                                       
               1 MB total memory in use (0 MB lost due to fragmentation)  

./cparis3 2000000 +RTS -sstderr
2
3
CCTGGCCAACATGGTGAAAC
CCTGGCCAACATGGTGAAAC
      80,939,296 bytes allocated in the heap
       8,925,240 bytes copied during GC
         137,056 bytes maximum residency (2 sample(s))
          45,528 bytes maximum slop
               2 MB total memory in use (0 MB lost due to fragmentation)

  %GC time      38.5%  (27.0% elapsed)

  Alloc rate    1,264,577,704 bytes per MUT second

  Productivity  61.5% of total user, 38.8% of total elapsed

./cparis3 20000000 +RTS -sstderr
2
3
CAGAGCGAGACTCCGTCTCA
CAGAGCGAGACTCCGTCTCA
     806,034,756 bytes allocated in the heap
      76,775,944 bytes copied during GC
         136,876 bytes maximum residency (2 sample(s))
          43,324 bytes maximum slop
               2 MB total memory in use (0 MB lost due to fragmentation)

  Generation 0:  1536 collections,     0 parallel,  0.35s,  0.35s elapsed
  Generation 1:     2 collections,     0 parallel,  0.00s,  0.00s elapsed

  INIT  time    0.00s  (  0.00s elapsed)
  MUT   time    0.53s  (  0.67s elapsed)
  GC    time    0.35s  (  0.36s elapsed)
  EXIT  time    0.00s  (  0.00s elapsed)
  Total time    0.88s  (  1.02s elapsed)

  %GC time      40.0%  (34.9% elapsed)

  Alloc rate    1,526,482,681 bytes per MUT second

  Productivity  60.0% of total user, 51.7% of total elapsed

Seems to work as desired.
...
Cristiano