1 patch for repository http://darcs.haskell.org/bytestring: Wed Jun 20 12:22:32 EEST 2012 Lauri Alanko * Add unsafePackMallocCStringLen New patches: [Add unsafePackMallocCStringLen Lauri Alanko **20120620092232 Ignore-this: 2d16fea6dbfcee6245604acbe2a5c4dc ] { hunk ./Data/ByteString/Unsafe.hs 39 unsafePackCString, -- :: CString -> IO ByteString unsafePackCStringLen, -- :: CStringLen -> IO ByteString unsafePackMallocCString,-- :: CString -> IO ByteString + unsafePackMallocCStringLen, -- :: CStringLen -> IO ByteString #if defined(__GLASGOW_HASKELL__) unsafePackAddress, -- :: Addr# -> IO ByteString hunk ./Data/ByteString/Unsafe.hs 267 len <- c_strlen cstr return $! PS fp 0 (fromIntegral len) +-- | /O(n)/ Build a @ByteString@ from a malloced @CStringLen@. This +-- value will have a @free(3)@ finalizer associated to it. +-- +-- This funtion is /unsafe/. If the original @CString@ is later +-- modified, this change will be reflected in the resulting @ByteString@, +-- breaking referential transparency. +-- +-- This function is also unsafe if you call its finalizer twice, +-- which will result in a /double free/ error, or if you pass it +-- a CString not allocated with 'malloc'. +-- +unsafePackMallocCStringLen :: CStringLen -> IO ByteString +unsafePackMallocCStringLen (cstr, len) = do + fp <- newForeignPtr c_free_finalizer (castPtr cstr) + return $! PS fp 0 len + -- --------------------------------------------------------------------- -- | /O(1) construction/ Use a @ByteString@ with a function requiring a } Context: [Add .gitignore file for GHC build. p.capriotti@gmail.com**20120307153052 Ignore-this: 755dcd78a9ec48b2ec2f3d98995b3a5f ] [Move __hscore_memcpy_src_off from base into include/fpstring.h Ian Lynagh **20120222155246 Ignore-this: f2f8fc7008f94e571ce78a33dfdb06f4 ] [Allow using older versions of the random package Duncan Coutts **20111117011128 Ignore-this: 8dc641a5670ab553609ad51796144713 ] [Fix some more import warnings Duncan Coutts **20111117005928 Ignore-this: dc08d7ddfd0af7f0d82b5a5c5132aaa3 ] [Fix an issue with char/binary Handle write ordering with ghc-6.12 Duncan Coutts **20111117005826 Ignore-this: f4721f3f50380d4dc684d2ce21001d0f For 6.12 it needs an extra flushWriteBuffer. ] [Fixes for ghc-6.10 Duncan Coutts **20111116233104 Ignore-this: bc56311805eb0c66634d3284c6ab7d75 ] [Make the testHandlePutBuilder conditional and add char8 version Duncan Coutts **20111116232920 Ignore-this: 415dba7d9209cbb7f0620c74e5ec27aa The testHandlePutBuilder does not apply to older ghc with the pre-Unicode Handles. Extra char8 test covers old ghc and should work with new too. ] [Add implementation of builder internal hPut for old or non-ghc Handle Duncan Coutts **20111116232729 Ignore-this: 56dacc6ed13971053e24dbf7c0e0004e The current hPut is specific to newer GHC with the new Unicode Handle stuff. Provide an implementation for older ghc and non-ghc compilers. ] [Fix building with ghc-7.3 and hopefully 7.4.x also Duncan Coutts **20111116220837 Ignore-this: 5a5baec25bfcefb784e70176256a1937 Based on a patch contributed by Herbert Valerio Riedel. Also fixes warnings for ghc-7.2. ] [Do not expose the BasicEncoding and other internals for this release Duncan Coutts **20111116204547 Ignore-this: 3a0f545c399c44acf4bc07ed03e9db9 We are being conservative here. There is useful functionality that we will want to expose somehow eventually, including the fixed and bounded size encodings for maximum speed of short encodings, plus the ability to do things like size-prefixed runs of data. However we will give ourselves some time to let this stuff settle down and a bit longer to think about the best way to support the more advanced & low-level bits. ] [port additional work on bytestring builder Simon Meier **20111113230154 Ignore-this: d9653d44e057498c62e61fd012c5e268 This patch adds the remainder of the work I performed on the old bytestring repository. It adds the following changes: - Faster implementation of Float and Double serialization. We exploit that there is already a buffer that we can use to convert the Float to a Word32 via the FFI. - Renaming of 'charASCII' and 'stringASCII' to 'char7' and 'string7': The new names are easier to use (as they are shorter) and follow the same convention as the 'char8' versions. - Move of variable length encodings and the chunked and size-prefixed builder transformers into a separate module: Data/ByteString/Lazy/Builder/BasicEncoding/Extras.hs The remainder of Data/ByteString/Lazy/Builder/BasicEncoding is now well-documentet and its API is considered stable. - Bugfixes to the size-prefixing builder transformer and a test case that caught the original bug. - The modules Data.ByteString.Lazy.Builder.Internal Data.ByteString.Lazy.Builder.BasicEncoding Data.ByteString.Lazy.Builder.BasicEncoding.Internal Data.ByteString.Lazy.Builder.BasicEncoding.Extras are now also exposed, but only Data.ByteString.Lazy.Builder.BasicEncoding is haddocked. They are required to provide transitional packages of blaze-html, blaze-builder, and blaze-enumerator based on the new bytestring builder. ] [Fix a few test failures due to incorrect Int64 -> Int conversion Duncan Coutts **20111115005203 Ignore-this: 2abe2b578946bcb407b6b78b776e1d08 Switch it around so we only promote Int -> Int64. ] [Add Simon's builder benchmark suite and add to .cabal file Duncan Coutts **20111113203741 Ignore-this: 661887305b8b317cd9a0196a0e4decad Makes use of cabal's new support for benchmarks ] [Add missing extension Duncan Coutts **20111113203349 Ignore-this: 59b1d1b93d4451270adbff6a85e4dbdf ] [Convert existing main test suite over to the minimal TestFramework code Duncan Coutts **20111113175155 Ignore-this: 6081a3d02265d25c61a3bf5813d2a5ef This will make it easier to switch to the test-framework package later. ] [Make the builder test suite into a cabal test-suite Duncan Coutts **20111113174805 Ignore-this: c8f2dab786e801a84f4e6fffb2b24bb5 The advantage is it makes it easier to run automatically. The disadvantage is we cannot use the nice test-framework package since cabal then thinks we've got a circular dependency since test-framework indirectly depends on bytestring. This will be solvable in future with encapsulated package dependencies, but util then we use a minimal implementation of the bit of the test-framework that we're using. ] [Add the test suite for the builder monoid Duncan Coutts **20111113163022 Ignore-this: 8cafd89cb45865dadee852a5433166e6 Again, this is Simon Meier's code ] [Update docs for chunk size constants Duncan Coutts **20111113162612 Ignore-this: 134635c118c3012135fbd01dbfdc69b1 ] [Add new Builder monoid by Simon Meier Duncan Coutts **20111113162509 Ignore-this: 1ad3cc5cbc8f84144a2c8b2c84d82ec7 The design of strict and lazy ByteStrings makes concatenation expensive. A builder monoid lets us efficiently build bytestrings by sticking bits together in an ad-hoc way. For example by pretty-printing or serialising. This is as opposed to a uniform approach using unfoldr. This is the result of a lot of work over several years by several people. This new version is by Simon Meier. It was based on the version by Jasper Van der Jeugt for the blaze-html library. The idea goes back further to the builder monoid in the binary library by Lennart Kolmodin and Ross Patterson. That was based on a builder monad by myself. The trick in all these designs is a continuation style monad/monoid for combining builders, and writing directly into buffers. The continuation trick allows it to be done in a non-persistent way, which in turn makes possible to just write directly into buffers. Simon and Jasper's implementation follows the same general approach at the previous designs but achieves better performance by paying careful attention to low level details like buffer filling strategies and minimising the number of buffer bounds checks. ] [Added tests for toStrict and fromStrict Duncan Coutts **20111107124459 Ignore-this: d0c7929a386034358915ecc790cb1b7a ] [Also export foldrChunks and foldlChunks Duncan Coutts **20111107124358 Ignore-this: 94e856897ef7ab281f4eb670c9bdd2f4 Along with fromStrict and toStrict we now match the Text API in this area. ] [Add conversion functions between lazy and (single) strict ByteStrings Duncan Coutts **20111107124352 Ignore-this: 688da877749822671f3d4f19a18ed9d API proposal and initial patch by Herbert Valerio Riedel http://article.gmane.org/gmane.comp.lang.haskell.libraries/16444 More or less unanimous support. Don points out that he and I deliberately did not include these functions originally in an attempt to discourage people from converting back and forth since it is expensive to do so. Clearly people want these functions anyway. So instead I've added the following documentation to toStrict: Note that this is an /expensive/ operation that forces the whole lazy 'ByteString' into memory and then copies all the data. If possible, try to avoid converting back and forth between strict and lazy bytestrings. ] [Fix documentation of complexity of toChunks Duncan Coutts **20111107124046 Ignore-this: 855b8095c66933d4f3b362447a831c9c ] [Bump version to 0.10.0.0 Duncan Coutts **20111107113626 Ignore-this: 6ca32307604b4f606ec20787df181e19 Some minor API changes and a number of extensions ] [Update module metadata and copyright info Duncan Coutts **20111107113610 Ignore-this: 5333efad2ec1b1032d28968de5acfff ] [A few minor doc improvements Duncan Coutts **20111107113555 Ignore-this: 417595713f2ea82a7fb77e813c01a36d ] [Update package description and tested-with list Duncan Coutts **20111107113533 Ignore-this: 46fe8d73b4d6e7fcee82a60fb23ffa47 Liberally borrowed the style of description from the text package. ] [Fix for ghc-6.8 / base-3 Duncan Coutts **20111107113400 Ignore-this: 9dc0838327115775cecb717ff5b48e3f ] [Fix warning about C FFI types Duncan Coutts **20111107113341 Ignore-this: 79f8b5ccb46537808b79eb5e3fa624a ] [CPP-conditional LANGUAGE pragmas now work since ghc-7.0.x Duncan Coutts **20111107113315 Ignore-this: 3a5ae0dc7e24c124954c2714d4e69c2 ] [Fix implementation of Lazy.groupBy Duncan Coutts **20111107113212 Ignore-this: 770b846bc6817876fab2126267823912 ] [Clean up a few minor things in the test properties Duncan Coutts **20111107113128 Ignore-this: 4e47064369d66b56503947220559b627 ] [Add more extensive tests for the various new pack and unpack functions Duncan Coutts **20111107113107 Ignore-this: a0b8335ef27279ca1e7ce66dc94fa62e ] [Include non-0 offset in instance Arbitrary Strict.ByteString Duncan Coutts **20111107113008 Ignore-this: f7f8bf83f79368bc748f4e07cff6abc And fix a test that breaks. ] [Add tests for groupBy Duncan Coutts **20111107112900 Ignore-this: a1e1619e572d8fe0db84e4d3bc0f3de3 Note that it fails for Lazy.groupBy due to a chunk boundary bug. ] [Add BangPatterns extension Duncan Coutts **20111106194117 Ignore-this: 239b498c92e32e592914d40c2e1f6187 ] [Move Eq, Ord, Monoid instances to eliminate orphans Duncan Coutts **20111106192815 Ignore-this: af00829573a746c18fd415c6454d0450 ] [Use the new {un,}pack{Bytes,Chars} functions and simplify Duncan Coutts **20111106182800 Ignore-this: 630dddaa4adaf3398c3a18eace92df4d We had a whole variety of odd pack/unpack functions. The nice thing is that with modern ghc we can use simpler implementations and get as good or better code than the old ones. ] [Drop unnecessary -funbox-strict-fields and set -fspec-constr-count Duncan Coutts **20111106181118 Ignore-this: 8050eaefdafd298b1df338ff5880284e We use the UNPACK pragma explicitly in the couple places where we use strict fields. ] [Move the IsString instance so it is not an orphan Duncan Coutts **20111106180508 Ignore-this: ca0582bb256102c9a3315313293dfd51 ] [Add proper Data class instances that actually contain the data. Duncan Coutts **20111106175737 Ignore-this: b4654ee840a451db339ed65023a811ec Same style as instances for Data.Text. ] [Add a proper Show and Read instance for lazy bytestrings Duncan Coutts **20111106165851 Ignore-this: 9c3959d661a2d1192c075bf7eeab57dc Previous one was derived which was silly. ] [Add new internal list pack and unpack functions Duncan Coutts **20111106165316 Ignore-this: 47f712808229c6253c62174e3450d73d Use them for the Show and Read instances. They should also be a tad better in terms of speed and memory use. ] [Remove unused fusion module Duncan Coutts **20111106001627 Ignore-this: 4fa7b48fbf2d89acd2ed9dfc408dadfb ] [Fix test-suite and get rid of some warnings Duncan Coutts **20111105233537 Ignore-this: 88f93c9517b8081ee6f24f4d3e7f5dc1 Based on a patch by Bas van Dijk ] [Update cabal package metadata Duncan Coutts **20111105171754 Ignore-this: ef6cf2e6d9cbf0c808ddb7d3052cf1ff Remove old homepage, add source repo and bug report addresses. ] [Add NFData instances for strict and lazy ByteStrings Bas van Dijk **20111105165050 Ignore-this: dc45153ceef774d3457118de44fa67e8 ] [Throw exception in IO for functions in IO Duncan Coutts **20111105163259 Ignore-this: e0ac1f69debe2e1295c20a4ca6d72c98 Tracked down by Gershom Bazerman ] [Specify sensible fixities for cons and snoc Duncan Coutts **20111105151635 Ignore-this: c6d764970b3922e09ca3b07898472ead As suggested by Yitzchak Gale. ] [Follow change to FFI decls: Import constructors of newtypes Ian Lynagh **20111021002852 Ignore-this: 6f44939dd54d5dd7ca5f0decc968063a ] [Use Safe Haskell if GHC >= 7.2 David Terei **20110809192450 Ignore-this: 41e8f96f377740495943bd215d27a8a9 ] [Remove old GHC flag from Cabal file Ian Lynagh **20110704185011 Ignore-this: de4da9ee1c3f1e369e7533428846dbaf The flag has been removed in GHC 7.3 ] [Remove some old done TODO items Duncan Coutts **20110517135450 Ignore-this: ae1fe46f268da4c840903814d9139c61 ] [Update the test suite to QC2 and add a tests-suite stanza to the cabal file Duncan Coutts **20110517135110 Ignore-this: 9ba994914ff9aebdf5db5d7e706453ac QC2 uses a rather different instance for Arbitrary Int which uses much bigger numbers than QC1 used. Some properties have had to be updated to use a smaller int range or they would take nearly forever to run. ] [Update maintainers' email addresses Duncan Coutts **20110517134332 Ignore-this: ef4654d3424906be1f5a792950dbd156 ] [Drop support for ghc-6.4 and 6.6 Duncan Coutts **20110517112532 Ignore-this: b8a978b2dd1d16c6f57f5f5094b59e4e ] [Bump minor version due to API additions Duncan Coutts **20110516122813 Ignore-this: 34e61d06b93c76adb2b7d5905b838fe0 ] [Export putStrLn and hPutStrLn from D.B.Lazy.Char8 Duncan Coutts **20110516122624 Ignore-this: 775831f2477cb3f69b9ebf806f45b26c and deprecate both functions in the non-Char8 modules. Functions that rely on ASCII encodings belong in the Char8 modules. ] [Add hPutNonBlocking Duncan Coutts **20110516115811 Ignore-this: 2ec469780b26a86a7492dfc87b42caa7 Based originally on a patch by David Fox Fixes ghc ticket #1070. Also update the documentation for hGetNonBlocking ] [Use explicit import list for GHC.IO to avoid build failure Max Bolingbroke **20110514221755 Ignore-this: c1825af0914b958099eda67ed583db50 ] [Bump version number to 0.9.1.10 Ian Lynagh **20110107162327 Ignore-this: 775963ffa10934c592060656106d502e ] [Make is build with nhc98. Malcolm.Wallace@me.com**20110107142432] [TAG 0.9.1.9 Duncan Coutts **20110106171112 Ignore-this: 1aa3fd2d9d791b99bc4a36cea3862bb1 ] Patch bundle hash: e4b75d1055bd352248de7839b1ec7180db5cfb53